springer proceedings in physics 111 springer proceedings in physics 90 Computer Simulation Studies in Condensed-Matter Physics XV Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 91 The Dense Interstellar Medium in Galaxies Editors: S. Pfalzner, C. Kramer, C. Straubmeier, and A. Heithausen 92 Beyond the Standard Model 2003 Editor: H.V. Klapdor-Kleingrothaus 93 ISSMGE Experimental Studies Editor: T. Schanz 94 ISSMGE Numerical and Theoretical Approaches Editor: T. Schanz 95 Computer Simulation Studies in Condensed-Matter Physics XVI Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 96 Electromagnetics in a ComplexWorld Editors: I.M. Pinto, V. Galdi, and L.B. Felsen 97 Fields, Networks, Computational Methods and Systems in Modern Electrodynamics A Tribute to Leopold B. Felsen Editors: P. Russer and M. Mongiardo 98 Particle Physics and the Universe Proceedings of the 9th Adriatic Meeting, Sept. 2003, Dubrovnik Editors: J. Trampetić and J. Wess 99 Cosmic Explosions On the 10th Anniversary of SN1993J (IAU Colloquium 192) 100 Lasers in the Conservation of Artworks LACONA V Proceedings, Osnabrück, Germany, Sept. 15–18, 2003 Editors: K. Dickmann, C. Fotakis, and J.F. Asmus 101 Progress in Turbulence Editors: J. Peinke, A. Kittel, S. Barth, and M. Oberlack 102 Adaptive Optics for Industry and Medicine Proceedings of the 4th International Workshop Editor: U. Wittrock 103 Computer Simulation Studies in Condensed-Matter Physics XVII Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 104 Complex Computing-Networks Brain-like and Wave-oriented Electrodynamic Algorithms Editors: I.C. Göknar and L. Sevgi 105 Computer Simulation Studies in Condensed-Matter Physics XVIII Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 106 Modern Trends in Geomechanics Editors: W. Wu and H.S. Yu 107 Microscopy of Semiconducting Materials Proceedings of the 14th Conference, April 11–14, 2005, Oxford, UK Editors: A.G. Cullis and J.L. Hutchison 108 Hadron Collider Physics 2005 Proceedings of the 1st Hadron Collider Physics Symposium, Les Diablerets, Switzerland, July 4–9, 2005 Editors: M. Campanelli, A. Clark, and X. Wu 109 Progress in Turbulence 2 Proceedings of the iTi Conference in Turbulence 2005 Editors: M. Oberlack et al. 110 Nonequilibrium Carrier Dynamics in Semiconductors Proceedings of the 14th International Conference, July 25-29, 2005, Chicago, USA Editors: M. Saraniti, U. Ravaioli 111 Vibration Problems ICOVP 2005 Editors: E. Inan, A. Kiris Volumes 64–89 are listed at the end of the book. Editors: J.M. Marcaide and K.W. Weiler The Seventh International Conference on Vibration Problems ICOVP 2005 05-09 September 2005, stanbul, Turkey Edited by Esin nan stanbul, Turkey and Technical University, stanbul, Turkey İ İ ş İ İ ı şI k University,ı İstanbul Ahmet K rı A C.I.P. Catalogue record for this book is available from the Library of Congress. Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com Printed on acid-free paper No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. ISBN-10 1-4020-5400-9 (HB) ISBN-13 978-1-4020-5400-6 (HB) ISBN-10 1-4020-5401-7 (e-book) ISBN-13 978-1-4020-5401-3 (e-book) © 2007 Springer All Rights Reserved ICOVP-2005 CONTENTS Preface xv Adanur S, Soyluk K, Dumanoğlu A A, Bayraktar A (RC) Asynchronous and antisynchronous effects of ground motion on the stochastic response of suspension bridges 1 Akkaş N (Special Talk) European union framework programme 7 building the Europe of knowledge 7 Akköse M, Adanur S, Bayraktar A, Dumanoğlu A A (RC) Dynamic response of rock-fill dams to asynchronous ground motion 9 Aksoğan O, Choo B S, Bikçe M, Emsen E, Reşatoğlu R (RC) A comparative study on the dynamic analysis of multi-bay stiffened coupled shear walls with semi-rigid connections 15 ˘ Free vibrations of cross-ply laminated non-homogeneous composite truncated conical shells 21 Akyüz U, Ertepınar A (RC) Symmetric and asymmetric vibrations of cylindrical shells 27 Aldemir Ü, Aydın E (RC) An active control algorithm to prevent the pounding of adjacent structures 33 Aldemir Ü, Güney D (RC) Vibration control of non-linear buildings under seismic loads 39 Aslanyan A G, Dronka J, Mishuris G S, Selsil Ö (RC) Vibrations of damaged 1D-3D multi-structures 45 Aşçı N, Uysal H, Uzman Ü (RC) Sizing of a spherical shell as variable thickness under dynamic loads 51 iv i Aksogan O, Sofiyev A H, Sofiyev A (RC) viii CONTENTS Aydoğdu M, Taşkın V (RC) Vibration analysis of simply supported functionally graded beams 57 Baltacı A, Sarıkanat M, Yıldız H (RC) Damping effects of rubber layer in laminated composite circular plate during forced vibration 63 Banerjee M M (GL) A critical study on the application of constant deflection contour method to nonlinear vibration of plates of arbitrary shapes 69 Banks S P, Salamcı M U (RC) Nonlinear wave equations and boundary control using visco elastic dampers 79 Birlik G, Sezgin Ö C (RC) Effect of vibrations on transportation system 85 Bonifasi-Lista C, Cherkaev E (RC), (presented by Cherkaev A) Identification of bone microstructure from effective complex modulus 91 Boyac H (RC) Beam vibrations with non-ideal boundary conditions 97 Chakrabarti B K, Chatterjee A (GL) A two-fractal overlap model of earthquakes 103 Cherkaev A, Cherkaev E, Slepyan L (GL) Dynamics of structures with bistable links 111 Çelebi M (GL) Real-time seismic monitoring of the new Cape Girardeau (MO) bridge and recorded earthquake response 123 Demir F, Türkmen M, Tekeli H, Çırak İ (RC) Earthquake response of masonry infilled precast concrete structures 137 Demiray H (GL) Nonlinear waves in fluid-filled elastic tubes: a model to large arteries 143 Doğan V (RC) The nonlinear axisymmetric vibrations of circular plates with linearly varying thickness under random excitation 151 Doğan V, Kırca M (RC) Dynamic analysis of a helicopter rotor by Dymore program 157 ı CONTENTS ix Ecker H (GL) 163 Erdem A Ü (RC) A formulation of magnetic-spin and elasto-plastic waves in saturated ferrimagnetic media 175 Erdik M, Apaydın N (GL) Earthquake response of suspension bridges 181 Fedorov V A, Smolskiy S M, Mizirin A V, Shtykov V V, Fomenkov A V, Kaplunov S M (RC, not presented) The microwave sensor of small moving for the active control of vibrations and chaotic oscillations modes 195 Fesenko T N, Foursov V N (RC, not presented) Forced oscillations of tube bundles in liquid cross-flow 205 Filippenko G V, Kouzov D P (RC) The exact and approximate models for the vibrating plate partially submerged into a liquid 213 Göğüş M T, Tayşi N (RC) A computational tool based on genetic algorithm for determining optimum shapes of vibrating planar and space trusses 219 Güler C, Akbarov S D (RC) On the dynamical stress field in the pre-stretched bilayered strip resting on the rigid foundation 225 Güler K, Celep Z (RC) Dynamic response of a rectangular plate-column system on a tension- less elastic foundation 231 İtik M, Salamcı M U, Ülker F D (RC) Vibration suppression of an elastic beam via sliding mode control 237 Ivanova J, Bontcheva N, Pastrone F, Bonadies M (RC) Thermoelastic stress analysis for linear elastic bodies 243 Parametric stiffness excitation as a means for vibration suppression , Özakça M x CONTENTS Kan C, Urey H, Şenocak E (RC) Discrete and continuous mathematical models for torsional vibration of micromechanical scanners 249 Kaplunov S M, Makhutov N A, Solonin V I, Shariy N V (RC, not presented) Physical modeling of stationary and impulsive processes for large- scaled constructions of fluid elastic systems 255 Karmakar B, Biswas P, Kahali R, Karanji S (RC) Thermal stresses and nonlinear thermal deformation analysis of shallow shell panel 265 Karmakar B, Karanji S B, Kahali R, Biswas P (RC) Nonlinear thermal vibrations of a circular plate under elevated temperature 271 Kaya M O, Özdemir Ö (RC) section composite Timoshenko beam by using the differential trans- form method 279 Kırış A, İnan E (RC) Estimation of microstretch elastic moduli by the use of vibrational data 285 ¨ Vibrations of a circular membrane subjected to a pulse 291 Kumbasar N (RC) A method of discrete time integration using Betti’s reciprocal theorem 297 Mehta A (GL) Bridges in vibrated granular media 305 Michelitsch T M, Askes H, Wang J, Levin V M (RC) Solutions for dynamic variants of Eshelby’s inclusion problem 317 Mondal S (RC) Note on large deflections of clamped elliptic plates under uniform load 323 Movchan A B, Haq S, Movchan N V (GL) Localised defect modes and a macro-cell analysis for dynamic lattice structures with defects 327 Flexural-torsional coupled vibration analysis of a thin-walled closed Korfalı O, Parlak I B (RC)˙ CONTENTS xi Mukherjee A, Samadhiya R (RC) Acoustic wave propagation through discretely graded materials 337 Nordmann R (GL) Use of mechatronic components in rotating machinery 345 M (GL) Computational limits of FE transient analysis 357 Oskouei A V, Aksoğan O (RC) The effect of non-linear behavior of concrete on the seismic response of concrete gravity dams 371 ¨ ¨ ¨ Calculation of microwave plasma oscillations in high temperature superconductors 377 Özer A Ö, İnan E (RC) medium with finite difference method 383 Özer M, Alışverişçi F (RC) Dynamic response analysis of rocking rigid blocks subjected to half- sine pulse type base excitations 389 Özeren M S, Postacıoğlu N, Zora B (RC) A new spectral algorithm for 3-D wave field in deep water 395 Pes̆ek L, Pu ◦ st L (RC) Experimental and numerical assessment of vibro-acoustic behavior of rubber-damped railway wheels 403 Raamachandran J (GL) Charge simulation method applied to vibration problems 409 Singh S, Patel B P, Sharma A, Shukla K K, Nath Y (GL) Nonlinear stability and dynamics of laminated composite plates and shells 415 S̆kliba J, Sivc̆a′ k M, Skarolek A (RC) On the stability of a vibroisolation system with more degrees of freedom 429 Okrouhlı́k One-dimensional wave propagation problem in a nonlocal finite Ozdemir Z G, Aslan O, Onbaşlı U (RC) xii CONTENTS The stability and vibration of conical shells composed of SI3N4 and SUS304 under axial compressive load 437 Dynamic buckling of elasto-plastic cylindrical shells under axial load 443 Sönmez Ü (RC) Dynamic buckling analysis of imperfect elastica 449 Svoboda R, S̆kliba J, Matĕjec R (RC) Specification of flow conditions in the mathematical model of hy- draulic damper 455 Tanrıöver H, Şenocak E (RC) Nonlinear transient analysis of rectangular composite plates 463 Taşçı F, Emiroğlu İ, Akbarov S D (RC) On the “resonance” values of the dynamical stress in the system comprises two-axially pre-stretched layer and half-space 469 Tayşi N, Göğüş M T, Özakça M (RC) Optimization of vibrating arches based on genetic algorithm 475 Teymür M (RC) Propagation of long extensional nonlinear waves in a hyper-elastic layer 481 Tondl A, Nabergoj R, Ecker H (RC) Quenching of self-excited vibrations in a system with two unstable vibration modes 487 Topal U, Uzman Ü (RC) Free vibration analysis of laminated plates using first-order shear deformation theory 493 Turhan Ö (RC) The generalized method as an alternative tool for com- plete dynamic stability analysis of parametrically excited systems: application examples 499 Turhan Ö, Bulut G (RC) Coupling effects between shaft-torsion and blade-bending vibrations in rotor-blade systems 505 Bolotin Sofiyev A H, Deniz A (RC) Sofiyev A, Schnack E (RC) CONTENTS xiii Yahnioğlu N, Akbarov S D (RC) Forced vibration of the pre-stretched simply supported strip contain- ing two neighbouring circular holes 511 Yavuz M, Ergüven M E (RC) Free vibration of curved layered composite beams 519 Yeliseyev V V, Zinovieva T V (RC) Vibrations of beam constructions submerged into a liquid 525 Yıldırım V (RC) Vibration behavior of composite beams with rectangular sections considering the different shear correction factors 531 Yüksel H M, Türkmen H S (RC) Air blast-induced vibration of a laminated spherical shell 537 Author Index 543 List of Participants 547 The names of the authors who actually delivered the lectures, talk, or communications at the con- ference are underlined. Abbreviations: GL = general lectures, RC = research communications PREFACE The Seventh International Conference on Vibration Problems (ICOVP-2005) took place in Şile Campus of Işık University, İstanbul, Turkey, between the dates 5- 9 September 2005. First ICOVP was held during October 27-30, 1990 at A.C. College, Jalpaiguri under the co-Chairmanship of two scientists, namely, Profes- sor M. M. Banerjee from the host Institution and Professor P. Biswas from the sister organization, A. C. College of Commerce, in the name of “International Conference on Vibration Problems of Mathematics and Physics”. The title of the Conference was changed to the present one during the third conference. The Conferences of these series are: 1. ICOVP-1990, 20-23 October-1990, A.C. College, Jalpaiguri- India 2. ICOVP-1993, 4-7 November 1993, A.C. College, Jalpaiguri- India 3. ICOVP-1996, 27-29 November 1996, University of North Bengal, India 4. ICOVP-1999, 27-30 November 1999, Jadavpur University, West Bangal, India 5. ICOVP-2001, 8-10 October 2001, (IMASH), Moscow, Russia 6. ICOVP-2003, 8-12 September 2003, Tech. Univ. of Liberec, Czech Republic 7. ICOVP-2005, 5-9 September 2005, Işık University, Şile, İstanbul, Turkey The General Lecturers of ICOVP-2005 have been personally invited by the Inter- national Scientific Committee, which this time comprised the following members: Nuri AKKAŞ (Turkey), Yalçın AKÖZ (Turkey), Orhan AKSOĞAN (Turkey), Fikret BALTA (Turkey), M. M. BANERJEE (India), Victor BIRMAN (USA), Paritosh BISWAS (India), Bikas K. CHAKRABARTI (India), Hilmi DEMİRAY (Turkey), Ali Ünal ERDEM (Turkey), Ragıp ERDÖL (Turkey), Aybar ERTEPINAR (Turkey), K. V. FROLOV (Rusia), Avadis HACINLIYAN (Turkey), xv xvi PREFACE Richard B. HETNARSKI (USA), Koncay HÜSEYİN (Canada), Esin İNAN (Chair- person, Turkey), S. M. KAPLOUNOV (Russia), Faruk KARADOĞAN (Turkey), Nahit KUMBASAR (Turkey), J. MAZUMDAR (India), Yalçın MENGİ (Turkey), Nikita MOROZOV (Russia), Natasha MOVCHAN (UK), Abhijit MUKHERJEE (India), Jiri NAPRSTEK, (Czech Republic), J. RAAMACHANDRAN (India), G. A. ROGERSON (UK), J. S̆KLIBA (Czech Republic), T. R. TAUCHERT ( USA), Mevlut TEYMÜR (Turkey), A. TONDL (Czech Republic), Şenol UTKU (USA), F. VERHUST (The Nederlands), H. I. WEBER (Brazil), Vebil YILDIRIM (Turkey). lecture in ICOVP-2005 are: Professor M. M. BANERJEE (Asanson, India) Professor B. CHAKRABARTI (Saha Institute, Calcutta, India) Professor A. CHERKAEV (Utah University, Salt Lake City, USA) Professor M. ÇELEBİ (USGS, Menlo Park, CA, USA) Professor H. DEMİRAY (Işık University, İstanbul, Turkey) Professor H. ECKER (Institute for Machine Dynamics, Vienna, Austria) Professor M. ERDİK (Bogazii University, İstanbul, Turkey) Professor A. MEHTA (National Centre for Basic Sciences, Calcutta, India) Professor Y. NATH (NIT, New Delhi, India) Professor R. NORDMANN (Technical University, Darmstadt, Germany) Professor M. OKROUHLIC (Academy of Science, Dolejskova, Czech Republic) Professor J. RAAMACHANDRAN (IIT, Madras, India) The Chairperson, Esin İnan (Işık University) has been supported by the Local Committee including her colleagues O. Aksoğan (Çukurova University) and H. Demiray (Işık University). As with the earlier Conferences of the ICOVP series, the purpose of ICOVP-2005 was to bring together scientists with different backgrounds, actively working on vibration problems of engineering both in theoretical and applied fields. The main objective did not lie, however, in reporting specific results as such, but rather in joining different languages, questions and methods developed in the respective disciplines and to stimulate thus a broad interdisciplinary research. Judging from the lively discussions, the friendly, unofficial and warm atmosphere, both inside and outside Conference rooms, this goal was achieved. The General Lecturers (GL) who kindly accepted our invitaton and delivered a Professor A. B. MOVCHAN (University of Liverpool, Liverpool, UK) PREFACE xvii The following broad fields have been chosen by the International Scientific Com- mittee to be of special importance for the ICOVP-2005: TOPIC 1. Mathematical models of vibration problems in continuum mechanics, TOPIC 2. Vibration problems in non-classical continuum models and wave me- chanics, TOPIC 3. Vibrations due to solid / liquid phase interaction, TOPIC 4. Vibration problems in structural dynamics, damage mechanics and composite materials, TOPIC 5. Analysis of the linear / non-linear and deterministic / stochastic vibra- tions phenomena, TOPIC 6. Vibrations of transport systems, TOPIC 7. Computational methods in vibration problems and wave mechanics, TOPIC 8. Vibration problems in earthquake engineering, TOPIC 9. Vibration of granular materials, TOPIC 10. Active vibration control and vibration control in space structures, TOPIC 11. Vibration problems associated with nuclear power reactors. Other topics concerned with vibration problems, in general, were open as well, but it was understood that the bulk of presentations were within the above fields. All of the lecturers were carefully “nominated” by the International Scientific Committee, so as to illustrate the newest trends, ideas and the results. Altogether there were 81 active participants from 11 different countries, who presented 13 “general lectures (GL)” 1 Special Talk and 59 “research commu- nications (RC)”, of which 49 were oral presentations and 10 were posters. Each GL was 50 minutes long, including questions and discussion, while each oral RC was 20 minutes long, including questions and discussion. The posters were exhibited throughout the 5 days of the Conference. There was ample time and op- portunity for private discussions. Many private scientific meetings and interactions took place. Hopefully these will lead to new collaborations and other research developments in the coming years. It is indicated in the Table of Contents (TOC) the character of each presentation, i.e., GL or RC. In the case of more than one author, the name of the presenting author is underlined in the TOC. At the end of this volume, there appears a list of all the presenting participants, along with addresses, both postal and e-mail. An Author Index (Al) is also included to the end of this volume; all the manuscript authors appear in it along with the page number(s) where their article(s) begins. The presenting author is always indicated by underlining their names. Authors not actually participating in or present at the symposium are marked in the Author xviii PREFACE lndex (AI) by an asterisk. All lectures delivered at the Conference are recorded in this volume with the full text. It is real pleasure to express our sincere gratitude to the people and organizations for their contributions, help and support for this Conference. In the first place, I have to mention here that it would not be possible to organize this symposium without the support of the Işık University. The full support, understanding and encouragement of Prof. Ersin Kalaycıoğlu, the Rector of the Işık University, were made the life easy for us. Secondly the support of the Faculty of Arts and Sciences was immeasurable and the staff of the Dean’s office and my secretary, Ms. Filiz Özsobacı were really worked very hard. I am very grateful to all of them. On the other hand, I would like to express my deep gratitude to the members of The International Scientific Committee for their valuable and vitalizing ideas, comments, suggestions, and criticism on the scientific program of the Conference. It is with great pleasure and gratitude that we acknowledge the support of The Technical and Scientific Research Council of Turkey -TÜBİTAK- which made possible to publish this Proceeding and also to give a small contribution to some of the General Lecturers who could not get any financial support from their own Institutes and to cover some part of their travel and local expenses of the partici- pants coming from India, Russia and the former Soviet Union Countries. Finally, we would like to send our cordial thanks to all lecturers for their excellent presentations and careful preparation of the manuscripts. We are looking forward to come together at 8th ICOVP conference, which will tentatively take place in India in 2007. Esin İnan, Chairperson Arıköy, İstanbul, January 2006 1 (eds.), Vibration Problems ICOVP 2005, ASYNCHRONOUS AND ANTISYNCHRONOUS EFFECTS OF GROUND MOTION ON THE STOCHASTIC RESPONSE OF SUSPENSION BRIDGES Süleyman Adanur1, Kurtuluş Soyluk2, A. Aydın Dumanoğlu3 and Alemdar Bayraktar1 1 , Trabzon, Turkey 2 Deparment of Civil Engineering, Gazi University, Ankara, Turkey 3Grand National Assembly of Turkey, Ankara, Turkey Abstract. In this paper, the stochastic dynamic responses of a suspension bridge subjected to asyn- chronous and antisynchronous ground motions are performed. Asynchronous and antisynchronous dynamic analyses are carried out for various wave velocities of the traveling earthquake ground motion. As an example Bosporus Suspension Bridge, built in Istanbul, is chosen. Filtered white noise (FWN) ground motion model modified by Clough and Penzien is used as ground motion model. The intensity parameter for the FWN model is obtained by equating the variance of this model to the variance of the two thirds of the S16E component of Pacoima Dam record of 1971 San Fernando earthquake, applied in the vertical direction. Mean of maximum displacements and internal forces obtained from the considered analyses are compared with each other. Key words: suspension bridge, stochastic response, asynchronous ground motion, antisynchronous ground motion, filtered white noise (FWN) ground motion model 1. Introduction Under the effect of dynamic loading one of the uncertainties in the structural analysis arises from the dynamic loading itself to which the structure is subjected. Because dynamic effects like earthquake motions are random there is a need to a process taking into account the uncertainty of the dynamic loading in the analysis. The analysis due to random loading is defined as the stochastic analysis. In long-span structures, like suspension bridges, dynamic ground motion will arrive to the support points at different times. At this time period, the content as well as the phase of the motion is likely to be changed depending on the distance between support points and the local soil conditions. If an earthquake ground motion applied to one support propagates with finite velocity is in phase with the motions applied at the other supports, the analysis that takes into account this variation is defined as asynchronous dynamic analysis. In the opposite case, when c . © 2007 Springer. ş 1–6. E. Inan and A. Kır Ü Deparment of Civil Engineering, KT ı 2 ADANUR, SOYLUK, DUMANOĞLU, BAYRAKTAR a ground motion applied to one support propagates with finite velocity is out of phase relative to the motions applied at the other supports, this analysis is defined as antisynchronous dynamic analysis. Since the wave passage effect of ground motions between the supports of life- line structures has drawn the attention of researchers, stochastic and deterministic analyses of various structural configurations have been analysed. The objective of this paper is to determine the stochastic response of a suspension bridge, which has not been analysed comprehensively subjected to asynchronous and antisyn- chronous ground motions together. For this purpose, the stochastic responses of suspension bridges subjected to antisynchronous ground motion as well as asyn- chronous ground motion are investigated. Mean of maximum displacements and internal forces obtained from the considered analyses are compared with each other. 2. Random vibration theory In the random vibration theory, the variance of the total response component is expressed as (Harichandran et al., 1996) σ2z = σ 2 zs + σ 2 zd + 2Cov(zs, zd) (1) where, σ2zs and σ 2 zd are the pseudo-static and dynamic variances, respectively and Cov(zs, zd) is the covariance between the pseudo-static and dynamic response components. In the stochastic analysis depending on the peak response and standard de- viation of the total response, the mean of maximum value (µ) can be written as µ = pσz (2) where p is a peak factor and σz is the standard deviation of the total response (Der Kiureghian and Neuenhofer, 1991). 3. Ground motion model for random vibration response The cross-power spectral density function of the accelerations at the support points l and m is expressed as (Hawwari, 1992), S v̈glv̈gm(ω) = γlm(ω)S v̈g(ω) (3) where γlm(ω) is the coherency function describing the variability of the ground acceleration processes for support degrees of freedom l and m as a function of STOCHASTIC RESPONSE OF SUSPENSION BRIDGES 3 frequency ω, and S v̈g(ω) is auto-power spectral density function of the ground acceleration. Spatial variability of the ground motion is characterised with the coherency function in frequency domain. This function is dimensionless and complex valued. Recently Der Kiureghian (Der Kiureghian, 1996) proposed a general composite model of the spatial seismic coherency function in the following form γlm(ω) = γlm(ω) iγlm(ω) wγlm(ω) s = γlm(ω) i exp[i(θlm(ω) w + θlm(ω) s)] (4) where γlm(ω)i, γlm(ω)w and γlm(ω)s characterise the incoherence, the wave-passage and the site-response effects, respectively. In this study all the spatial effects are disregarded except the wave-passage effect and the coherency function is written as γlm(ω) = γlm(ω) w = exp[i(θlm(ω) w)] (5) θlm(ω) = −τω = − dLlm V ω (6) where τ is the arrival time of the ground motion between support points l and m, V is the apparent propagation velocity and dLlm is the projection of dlm on the ground surface along the direction of propagation of seismic waves. The auto-power spectral density function of the ground acceleration charac- terizing the earthquake process is assumed to be of the following form of FWN ground motion model modified by Clough and Penzien (Clough and Penzien, 1993) S v̈g(ω) = S 0[ ω4l + 4ξ 2 l ω 2 l ω 2 (ω2l − ω2)2 + 4ξ 2 l ω 2 l ω 2 ][ ω4 (ω2f − ω2)2 + 4ξ 2 fω 2 fω 2 ] (7) where S 0 is the amplitude of the white-noise bedrock acceleration, ωl, ξl and ω f , ξ f are the resonant frequency and damping ratio of the first and second filters, respectively. In this study firm soil type proposed by Der Kiureghian and Neuen- hofer (Der Kiureghian and Neuenhofer, 1991) is used. The filter parameters for this soil type are ωl = 15.0 rad/s, ξl = 0.6, ω f = 1.5 rad/s, ξ f = 0.6. The amplitude of the white-noise bedrock acceleration (S 0) is obtained for this soil type by equating the variance of the ground acceleration to the variance of S16E component of Pacoima Dam acceleration records of 1971 San Fernando earthquake. The calculated value of the intensity parameter for the firm soil type is S 0(firm)=0.021338 m2/s3. 4. Example In this study, as an example the Bosporus Suspension Bridge (Brown and Parsons, 1975) built in Turkey and connects Europe to Asia in Istanbul is selected. The 4 ADANUR, SOYLUK, DUMANOĞLU, BAYRAKTAR bridge has flexible steel towers of 165m high, inclined hangers and a steel box- deck of 1074m main span, with side spans of 231m and 255m on the European and Asian sides, respectively, supported on piers. The horizontal distance between the cables is 28m and the roadway is 21m wide, accommodating three lanes each way. The roadway at the mid-span of the bridge is approximately 64m above the see level. Two-dimensional finite element model of Bosporus Suspension Bridge with 202 nodal points, 199 beam elements and 118 truss elements is considered for the analyses. While the deck, towers and cables are represented by beam elements, the hungers are represented by truss elements. The selected finite element model of the bridge is represented by 475 degrees of freedom. As apparent wave velocities of the ground motion, V=1000 m/s, 2000 m/s and infinite wave velocities are used. It is assumed that the vertical ground motion is propagating from the European side to the Asian side. The analyses are obtained for 2.5 percent damping ratio and for the first 15 modes. The stiffening effects of the cables caused by the dead load are also accounted for in the analyses. Mean of maximum total response values are calculated for the asynchronous and antisynchronous ground motions. Mean of maximum total response values of the deck and tower are shown in Fig. 1, respectively. It is shown that the response values calculated for the asynchronous ground motion are generally larger than the values calculated for the antisynchronous ground motion and the values increase with decreasing wave velocities in the analyses for the asynchronous and anti- synchronous ground motions. It is also shown that the infinite wave velocity case in the analysis of the antisynchronous ground motion produces nil displacement value at the middle of the deck because of the out of phase motion of the applied forces at the supports. 5. Conclusions In this study, the stochastic responses of a suspension bridge are calculated con- sidering the antisynchronous ground motion as well as the asynchronous ground motion. The response values are obtained and compared with each other. The results at the deck and tower obtained from the asynchronous ground motion case are generally larger than those of the response values obtained from the antisyn- chronous ground motion case. Because of the pseudo-static effects, the response values increase with decreasing wave velocities for both ground motions. STOCHASTIC RESPONSE OF SUSPENSION BRIDGES 5 (a) Deck vertical displacement (b) Deck bending moment (c) Deck shear force 0 2 4 6 8 Displacement (cm) 0 50 100 150 200 T o w er h ei g h t (m ) Asynchronous V=infinite V=2000 m/s V=1000 m/s Antisynchronous V=infinite V=2000 m/s V=1000 m/s -600 -450 -300 -150 0 150 300 450 600 Bridge span (m) -100 -75 -50 -25 0 D is p la ce m en t (c m ) Asynchronous V=infinite V=2000 m/s V=1000 m/s Antisynchronous V=infinite V=2000 m/s V=1000 m/s (a) Tower horizontal displacement (b) Tower bending moment 0 25000 50000 75000 100000 Bending moment (kNm) 0 50 100 150 200 T o w er h ei g h t (m ) -600 -450 -300 -150 0 150 300 450 600 Bridge span (m) 0 5000 10000 15000 20000 25000 B en d in g m o m en t (k N m ) -600 -450 -300 -150 0 150 300 450 600 Bridge span (m) 0 200 400 600 800 S h ea r fo rc e (k N ) 0 500 1000 1500 2000 Shear force (kN) 0 50 100 150 200 T o w er h ei g h t (m ) (c) Tower shear force Figure 1. Mean of maximum total response values of the deck and tower 6 ADANUR, SOYLUK, DUMANOĞLU, BAYRAKTAR References Brown W. C., Parsons M. F. (1975) Bosporus Bridge, Part I, History of design, Proc. Instn Civ. Engrs, Part 1 58 505-532. Clough R. W., Penzien J. (1993) Dynamics of Structures, Second Edition, McGraw Hill, Inc., Singapore. Der Kiureghian A., Neuenhofer A. (1991) A response spectrum method for multiple-support seismic excitations. Report No. UCB/EERC-91/08, Berkeley (CA), Earthquake Engineering Research Center, College of Engineering, University of California. Der Kiureghian A. (1996) A coherency model for spatially varying ground motions, Earthquake Engineering and Structural Dynamics 25 99-111. Harichandran R. S., Hawwari A., Sweiden B. N. (1996) Response of long-span bridges to spatially varying ground motion, Journal of Structural Engineering 122 476-484. Hawwari A. R. (1992) Suspension bridge response to spatially varying ground motion, Ph.D. Thesis, Michigan State University, Michigan. 7 EUROPEAN UNION FRAMEWORK PROGRAMME 7 BUILDING THE EUROPE OF KNOWLEDGE Nuri Akkaş Department of Engineering Sciences, METU, 06531 Ankara, Turkey 1. Introduction In March 2000, the Lisbon European Council set the goal of becoming by 2010 “the most competitive and dynamic knowledge-based economy in the world, ca- pable of sustainable economic growth with more and better jobs and greater social cohesion”. This was called the Lisbon Strategy. The project of creating a European Research Area (ERA) was endorsed as a central element of the Lisbon Strategy to achieve this goal. However, EU still invests too little in R & D. In 2003, top 500 private R & D spenders in EU decreased their R & D investment by 2̃.0%. Top 500 private R & D spenders outside EU increased their R & D investment by 3̃.9%. Overall R &D investments are as follows: EU: 1.96%; US: 2.59%; S. Korea: 2.91%; Japan: 3.12%. ERA is implemented through so-called Framework Programmes (FP). FP7 is proposed on the basis of a doubling of funds and the duration is 7 years (2007-13). FP7 will fund R& D projects of immediate in- dustrial relevance & needs of industry. Projects will include both public research institutions and private companies (PPP). FOUR MAJOR COMPONENTS OF EUROPEAN RESEARCH IN FP7: I. Cooperation: COLLABORATIVE RESEARCH COMPONENT Support transnational cooperation in 9 themes: 1. Health 2. Food, agriculture and biotechnology 3. Information and communication technologies 4. Nanosciences, Nanotechnologies, Materials and new Production Technologies 5. Energy 6. Environment and Climate Change 7. Transport and Aeronautics 8. Socio-economic sciences and the humanities 9. Space and Security Research Vibration Problems ICOVP 2005, c© 2007 Springer. – . 7 8(eds.), . şE. Inan and A. Kırı 8 AKKAŞ II. Ideas: FRONTIER RESEARCH COMPONENT III. People: HUMAN POTENTIAL COMPONENT IV. Capacities: RESEARCH CAPACITY COMPONENT FOUR TOOLS TO IMPLEMENT EU SUPPORT IN THE ABOVE 9 THEMES: 1) Collaborative projects and networks. 2) Joint Technology Initiatives (JTI). 3) Coordination of national research programs. 4) International Cooperation (INCO): - Open all activities carried out in 9 themes to researchers and institutions from all 3rd countries. - Implemented on the basis of cost-shared funding. Bilateral agreements (US, Canada, Japan, Russia, Brazil, China, India, etc). - Targeting developing countries in fields of their particular needs (health, agriculture, energy, etc). - Specific actions aiming at reinforcing research capacities of Candidate Countries to the EU. The following questions will be answered: What are Technology Platforms? What are their Rationale, Characteristics, Approach, Aims, Criteria for Estab- lishment, and Expected Results? What are the existing TPs? What are Joint Technology Initiatives (JTI)? JTI topics are proposed by Commission for a limited number of key technolo- gies, considering industrial impact, feedback to public, national support, added value of European coordination. What are the areas already identified by Commission? 9 DYNAMIC RESPONSE OF ROCK-FILL DAMS TO ASYNCHRONOUS GROUND MOTION Mehmet Akköse1, Süleyman Adanur1, Alemdar Bayraktar1 and A. Aydın Dumanoğlu2 1 Department of Civil Engineering, KT , 61080, Trabzon, Turkey 2Grand National Assembly of Turkey, 06543, Ankara, Turkey Abstract. In this study, dynamic response of rock-fill dams to asynchronous ground motion is investigated. The Keban Dam, constructed in Elazığ, Turkey, is chosen as a numerical example. In the asynchronous dynamic analysis, wave velocities of 500 m/s, 1000 m/s, 2000 m/s and infinite are used for the travelling ground motion. Stresses are obtained for the wave velocities, and compared with each other. It is observed that the propagation velocity of the ground motion greatly influences the response of the rock-fill dam. Key words: rock-fill dams, asynchronous ground motion 1. Introduction In the classical dynamic analysis, it is assumed that the ground motion is uniform and has an infinite wave velocity so that same ground motion affects all support points of structure at the same time. In reality, however, the ground motion has finite wave velocity, so it will arrive to support points at different times (Sanchez- Sesma, 1987) and the effect of finite wave velocity results in various arrival time to support points. Because structures such as dams, nuclear power stations, sus- pension bridges and cable-stayed bridges are hundreds of meters long, ground motions will arrive from one support to the other in a few seconds. If the accelera- tion records are applied with different arrival times, the displacement components occurring with the movement of the support points will not be the same at every point on the structure. Because of the finite velocity of ground motion, the support points will move relatively towards each other. In addition to the dynamic dis- placements, quasi-static displacements will take place on the structure due to this movement (Clough and Penzien, 1993). When the ground motion is considered to be travelling with finite velocity, the equation of motion, therefore, has to be written in terms of total displacements that have quasi-static and dynamic compo- nents. In the classical dynamic analysis, no extra inertial forces will come from the first component of the displacement. So the first component is subtracted from the Vibration Problems ICOVP 2005, c© 2007 Springer. – 9 14. Ü (eds.), . şE. Inan and A. Kırı 10 AKKÖSE, ADANUR, BAYRAKTAR, DUMANOĞLU total displacements and only the dynamic displacements are considered. Analysis that includes the travelling effect of the ground motion is called the asynchronous dynamic analysis (Dumanoglu and Severn, 1984; Bayraktar et al., 1998; Soyluk and Dumanoglu, 2000). 2. Formulation of asynchronous dynamic analysis The formulation of asynchronous dynamic analysis is given in the previous studies in detail (Dumanoglu and Severn, 1984; Bayraktar et al., 1998). Therefore, in this study, equation of motion of the system to asynchronous ground motion is shortly presented. The mentioned equation of motion is given by Mü + Cu̇ + Ku = F (1) where M, C and K are mass, damping and stiffness matrices, respectively. ü, u̇, and u are total acceleration, velocity, and displacement vectors, respectively. F is external load vector. In the asynchronous dynamic analysis, total displacement consists of the sum of quasi-static and dynamic components, and can be expressed as u(t) = ∑ j r ju jg(τ j, t) + ∑ i ΦiYi(τ j, t) (2) where r j is the jth displacement shape function due to unit displacement assigned to ground degree of freedom; u jg is the jth ground displacement at the support points; τ j is the arrival time of the jth ground motion at a specific support point; Φi is the modal vector for mode i and Yi is the modal amplitude for mode i. 3. Numerical example In this study, the Keban dam constructed in Elazığ, Turkey is chosen as a numer- ical example to investigate the dynamic response of a rock-fill dam to asynchro- nous ground motion. The finite element mesh of the dam is shown in Fig. 1. The Keban Dam is 163m high from river bed. The crest has a maximum length of 1097m. The main purpose of the dam is to regulate river flow and supply energy. In the finite element mesh of the dam, there are 326 nodes and 286 quadrilateral elements. The structure is treated as a plane strain problem. The interaction of the rock-fill dams with the foundation rock and the reservoir has generally neglected (Priscu et al., 1985). In this study, the interaction with the reservoir is ignored, but not the foundation rock. Materials in the dam can be grouped in three main categories: compacted rock-fill placed at various lifts, the impervious clay core flanked by transition filters and a concrete core at the bottom of the dam (Akay and Gülkan, 1975). The properties of these materials are as follows. For the compacted rock-fill, elasticity modulus E = 1.632x1010 N/m2, mass density ρ = 2120.29 DYNAMIC RESPONSE OF ROCK-FILL DAMS 11 Figure 1. Finite element mesh of Keban Dam kg/m3, and Poisson’s ratio ν = 0.36. For the impervious clay core, elasticity mod- ulus E = 1.015x1010 N/m2, mass density ρ = 2089.70 kg/m3, and Poisson’s ratio ν = 0.45. For the concrete core, elasticity modulus E = 2x1010 N/m2, mass density ρ = 2446.48 kg/m3, and Poisson’s ratio ν = 0.15. Also, the foundation rock is taken into account in the study. Its elasticity modulus, mass density and Poisson’s ratio are taken as 1.379x1010 N/m2, 2689.09 kg/m3, and 0.24, respectively. The materials used are assumed to be linearly elastic, homogenous and isotropic. The program MULSAP (Dumanoglu and Severn, 1984) is employed in the response calculations. The E-W component of the Erzincan Earthquake, March 13, 1992, Erzincan, Turkey is chosen as ground motion and given in Fig. 2. The component considered is applied in the upstream-downstream direction. In the asynchro- nous dynamic analysis of the Keban Dam, wave velocities of 500 m/s, 1000 m/s, 2000 m/s and infinite are used for travelling ground motion. Infinite velocity case corresponds to classical dynamic analysis. Figure 2. The E-W component of the Erzincan Earthquake, March 13, 1992, Erzincan, Turkey 12 AKKÖSE, ADANUR, BAYRAKTAR, DUMANOĞLU 0 100 200 300 400 500 Lateral Distance (m) 0 2000 4000 6000 8000 10000 y y ( k N /m 2 ) 500 m/s 1000 m/s 2000 m/s Infinite 0 100 200 300 400 500 Lateral Distance (m) 0 1000 2000 3000 zz ( k N /m 2 ) 500 m/s 1000 m/s 2000 m/s Infinite 0 100 200 300 400 500 Lateral Distance (m) 0 1000 2000 3000 4000 y z (k N /m 2 ) 500 m/s 1000 m/s 2000 m/s Infinite Figure 3. Horizontal, vertical and shear stresses at I-I section for wave velocities of 500 m/s, 1000 m/s, 2000 m/s and infinite of asynchronous ground motion The absolute maximum horizontal, vertical and shear stresses at section I-I shown in Fig. 1 are presented in Fig. 3. The stresses are given at the centroid of the elements. As expected, all the stress components generally increase with decreas- ing velocity of the earthquake waves. This situation is clearly seen in horizontal stresses. But, this situation couldn’t be said fully for vertical and shear stresses. It is thought that this arose from horizontal quasi-static displacements caused by asynchronous horizontal ground motion. In order to investigate the variation of the frequency content of the stresses, time-histories of only horizontal stresses at element A shown in Fig. 1 were plotted in Fig. 4 for wave velocities of 500 DYNAMIC RESPONSE OF ROCK-FILL DAMS 13 0 3 6 9 12 15 18 21 Time (s) -6000 -4000 -2000 0 2000 4000 6000 y y ( k N /m 2 ) Velocity of 500 m/s 0 3 6 9 12 15 18 21 Time (s) -6000 -4000 -2000 0 2000 4000 6000 y y ( k N /m 2 ) Velocity of 1000 m/s 0 3 6 9 12 15 18 21 Time (s) -6000 -4000 -2000 0 2000 4000 6000 y y ( k N /m 2 ) Velocity of 2000 m/s 0 3 6 9 12 15 18 21 Time (s) -6000 -4000 -2000 0 2000 4000 6000 y y ( k N /m 2 ) Velocity of infinite Figure 4. The time-histories of horizontal stresses at the element A for wave velocities of 500 m/s, 1000 m/s, 2000 m/s and infinite of asynchronous ground motion 14 AKKÖSE, ADANUR, BAYRAKTAR, DUMANOĞLU m/s, 1000 m/s, 2000 m/s and infinite of asynchronous ground motion. It is seen from Fig. 4 that both amplitude and frequency contents of the time-histories of the stresses are considerably affected by decreasing velocity of the earthquake waves. 4. Conclusions In this study, dynamic response of a rock-fill dam to asynchronous ground motion is investigated by using the finite element method. In the asynchronous dynamic analysis, wave velocities of 500 m/s, 1000 m/s, 2000 m/s and infinite are used for the travelling ground motion. As decreasing velocity of the earthquake waves, all the stress components generally increase. It is also understood from the time- histories of the horizontal stresses that both amplitude and frequency contents of the time-histories of the stresses are considerably affected by decreasing velocity of the earthquake waves. References Akay H. U., Gülkan P. (1975) Earthquake Analysis of Keban Dam, Fifth European Conference on Earthquake Engineering, Istanbul, Turkey 1-3 Paper No:40. Bayraktar A., Dumanoğlu A. A., Calayir Y. (1998) Asynchronous Dynamic Analysis of Dam- Reservoir-Foundation Systems by The Lagrangian Approach, Computers and Structures 58 925-935. Clough R. W., Penzien J. (1993) Dynamics of Structures, Second Edition, McGraw-Hill Book Company, Singapore. Dumanoglu A. A., Severn R. T. (1984) Dynamic Response of Dams and Other Structures to Differential Ground Motions, Proc. Instn. Civ. Engrs., Part 2 77 333-352. Priscu R., Popovici A., Stematiu D., Stere C. (1985) Earthquake Engineering for Large Dams, Second Edition, Editura Academiei, Bucureti. Sanchez-Sesma F. J. (1987) Site Effects on Strong Ground Motion, Soil Dynamics and Earthquake Engineering 6 124-132. Soyluk K., Dumanoglu A. A. (2000) Comparison of Asynchronous and Stochastic Dynamic Responses of A Cable-Stayed Bridge, Engineering Structures 22 435-445. 15 A COMPARATIVE STUDY ON THE DYNAMIC ANALYSIS OF MULTI-BAY STIFFENED COUPLED SHEAR WALLS WITH SEMI-RIGID CONNECTIONS O. Aksoğan1, B.S. Choo2, M. Bikçe3, E. Emsen1 and R. Reşatoğlu1 1 Department of Civil Engineering, Çukurova University, 01330 Adana, Turkey 2School of the Built Environment Faculty of Engineering & Computing, Napier University, Edinburgh, EH10 5DT, UK 3Department of Civil Engineering, University of Mustafa Kemal, 31024 Hatay, Turkey Abstract. The present work considers dynamic analysis of multi-bay shear walls having any number of stiffening beams resting on rigid foundations. Key words: coupled shear wall, dynamic analysis, Newmark method 1. Introduction To carry out a quick predesign of coupled shear walls, an elegant method called Continuous Connection Method (CCM) has been widely used. In this method, the connecting beams are assumed to have the same properties and spacing along the entire height of the wall (or in a particular section of the wall). The discrete system of connecting beams is replaced by continuous laminae of equivalent stiffness (Rosman, 1964). For applying the method to the dynamic analysis of coupled shear walls, the structure is considered as a discrete system of lumped masses with their lateral displacements as the only degrees of freedom. Then, to find the stiffness matrix, a unit loading is applied for each freedom and the deflection forms are found by the CCM (Aksoğan et al., 2003). Substituting this stiffness matrix and the previously obtained mass matrix in the free vibration equation, the natural frequencies and the corresponding mode vectors are obtained. The forced vibration analysis under time dependent loading has been carried out using the mode superposition technique and the Newmark method (Clough and Penzien, 1993). Vibration Problems ICOVP 2005, c© 2007 Springer. – 15 20.(eds.), . şE. Inan and A. Kırı 16 AKSOĞAN, CHOO, BİKÇE, EMSEN, REŞATOĞLU 2. Analysis In the coupled shear wall seen in Figure 1, the bending stiffness and the shear forces in the discrete connecting beams are replaced by equivalent continuous functions. Figure 1. A multi-bay shear wall with stiffening beams The flexibility matrix is found by applying a unit force in the horizontal direction at the height of each lumped mass, one at a time. The horizontal displace- ments found from each unit load- ing case constitute a column of the flexibility matrix, the inverse of which yields the stiffness ma- trix. The main assumption of the CCM renders the horizontal dis- placements equal at the ends of a connecting beam which is a natural result of the well accepted rigid diaphragm assumption. The relations among the shear force functions, q j−1,i and q j,i, and the cor- responding contributions to the axial forces in the piers, Q j−1,i and Q j,i, can be written from the vertical force equilibrium (Figure 2). Consequently, Figure 2. The vertical forces on an isolated region of the shear wall Q0,i = Qm+1,i = q0,i = qm+1,i = 0 dQ j,i dx = −q j,i (1) where q j−1,i and q j,i are the shear force functions and the unknown functions Q j−1,i and Q j,i, each, are the sums of the shear forces starting from the top in a span between two neighboring piers. Defining Macaulay’s brackets by MULTI-BAY STIFFENED COUPLED SHEAR WALLS 17 < x − x′ >n= ( x − x′ )n and < x − x′ >0= 1 f or x > x′ < x − x′ >n= 0 and < x − x′ >0= 0 f or x ≤ x′ (2) for a unit force at height Hp, the moment-curvature relation, yields EIi d2yi dx2 =< Hp − x >1 − m∑ j=1 Q j,iL j j = 1, 2, ...,m, i = 1, 2, ..., n (3) For the compatibility of vertical displacements at the midpoints of the connecting beams (similarly for stiffening beams): L j dyi dx − hia2j 2Ccbi q j,i − hia3j 12EIc j,i q j,i − 1E { 1 A j,k } xk∫ xk+1 (Q j,k − Q j−1,k)dx + { 1 A j+1,k } xk∫ xk+1 (Q j,k − Q j+1,k)dx − 1E n∑ k=i+1 { 1 A j,i } x∫ xi+1 (Q j,i − Q j−1,i)dx + { 1 A j+1,i } x∫ xi+1 (Q j,i − Q j+1,i)dx − δ0 = 0 j = 1, 2, . . . ,m, i = 1, 2, . . . , n (4) The terms of the compatibility equation (4) are the relative vertical displace- ments of the two sides of the midpoints due, respectively, to the bending of the piers, the relative rotation of the beams with respect to the piers, the bending of the connecting beams due to the shear forces, the axial deformations of the piers in section i, the axial deformations of the parts of the piers between section i and the foundation and the relative vertical displacements in the foundation. Differentiating (4) with respect to x, using equations (1, 3) for j = 1, 2, ...,m, i = 1, 2, ..., n, k = 1, 2, ...,m [ γ2j,i Q ′′ j,i ] m×1 − [ α2jk ] m×m [ Q j,iL j ] m×1 = − < Hp − x >1, xi+1 ≤ x ≤ xi, (5) γ2j,i = EIi L j ( hia2j 2Ccbi + hia3j 12EIc j,i ) , α2jk = ( 1 − IiL jL j+1A j+1,i ) ( j = k − 1), α2jk = ( 1 + Ii L2j ( 1 A j,i + 1A j+1,i )) ( j = k), α2jk = 1( j < k − 1), α2jk = ( 1 − IiL jL j−1A j,i ) ( j = k + 1), α2jk = 1( j > k + 1) (6) 18 AKSOĞAN, CHOO, BİKÇE, EMSEN, REŞATOĞLU The coupled set of differential equations (5) is solved by the matrix orthogonal- ization method (Meirovitch, 1980), using a variable transformation, Q j,iL j = R j,i j = 1, 2, ...,m, i = 1, 2, ..., n (7) γ21,i L1 0 . 0 0 γ22,i L2 . 0 . . . . 0 0 . γ2m,i Lm ︸����������������︷︷����������������︸ A R′′1,i R′′2,i . R′′m,i ︸��︷︷��︸ R′′ + −α211 −α 2 12 . −α 2 1m −α221 −α 2 22 . −α 2 2m . . . −α2m1 −α 2 m2 . −α 2 mm ︸��������������������������︷︷��������������������������︸ B R1,i R2,i . Rm,i ︸��︷︷��︸ R = − < Hp − x >1 − < Hp − x >1 . − < Hp − x >1 ︸�����������������︷︷�����������������︸ Me (8) where A and B are m × m and R and Me are m × 1 dimensional matrices. The homogeneous part of this matrix equation, which is an eigenvalue problem in the following form A R′′ + B R = 0 (9) is solved, the eigenvectors yielding the transformation matrix T . Since the coef- ficient matrices A and B are constant, (8) can be diagonalized, for a new variable W, using R = T W, R′′ = T W′′, A T W′′ + B T W = Me (10) Multiplying both sides of the last equation in (10) by the transpose of T , à W′′ + B̃ W = M̃e (11) where à and B̃ are diagonal vectors and (11) is uncoupled. The solution of (11) is ( j = 1, 2, ...,m, i = 1, 2, ..., n) W j,i = C j,iCosh √ B̃ j j,i à j j,i x + D j,iS inh √ B̃ j j,i à j j,i x + 1 B̃ j j,i < Hp − x >1 (12) The integration constants in (12) are found from vertical force equilibrium and compatibility at the ends (including the level of the unit force). Finding Q j,i from (7), (10) and (12) and plugging in (3), integration with respect to x yields (i = 1, 2, ..., n) yi = 1 EIi ∫ ∫ < Hp − x > 1 − m∑ j=1 Q j,iL j dx dx + Hix + Gi (13) MULTI-BAY STIFFENED COUPLED SHEAR WALLS 19 The constants in (13) are determined from the conditions for the displacements and slopes on the boundaries. Using the unit horizontal displacement expressions, the flexibility, and thereof, the stiffness matrices are found. Substituting the mass and stiffness matrices in the free vibration equation, the natural frequencies, and therefrom, the pertinent mode shapes can be found. The equation of motion has been written as M Ÿ + C Ÿ + K Y = P(t) (14) and solved numerically by using the Newmark method (Clough and Penzien, 1993). 3. Numerical results Figure 3. Shear wall with three rows of openings To verify the present method a coupled shear wall with three rows of openings (Figure 3), has been solved under the effect of a dynamic loading P(t) of 1000 kN at the top for 4 seconds. The maximum top displace- ments of the shear wall were determined, for the undamped and damped (5% damping ratio) cases by the present method and by SAP2000 (Wilson, 1997) and the results were compared (Table 1). The variation of the top displacement for the undamped and damped cases is presented in Figure 4. The example structure has been also solved consider- ing a varying beam-wall connection stiffness against rotation and the results are compared in Figure 5. 4. Conclusions Since the present method proves to be accurate enough, it should be preferred for two reasons. Firstly, the data preparation is much easier than those of the equiv- alent frame and finite element methods. Besides, it renders possible quick trials of many different cases of a structure for optimization purposes. Secondly, the computation time of the present method is much less than those of the other two. Hence, the present method can be used effectively for predesign or dimensioning purposes. 20 AKSOĞAN, CHOO, BİKÇE, EMSEN, REŞATOĞLU TABLE I. The comparison of the maximum top displacements Damping Present SAP2000 (m) ratio study Frame % Finite element % CCM (m) method diff. method diff. 0% 0.1828 0.1832 0.21 0.1841 0.69 5% 0.1638 0.1643 0.33 0.1651 0.81 Figure 4. Variation of top displacement in undamped and damped cases Figure 5. Variation of maximum top dis- placement with connection stiffness References Aksoğan O., Arslan H. M., Choo B. S. (2003) Forced Vibration Analysis of Stiffened Coupled Shear Walls Using Continuous Connection Method, Eng. Struct. 25 499-506. Clough R. W., Penzien J. (1993) Dynamics of Structures, McGraw-Hill Inc., USA. Meirovitch L. (1980) Computer Methods in Structural Dynamics, Sijthoff and Norrdhoff, Nether- lands. Rosman R. (1964) Approximate Analysis of Shear Walls Subject to Lateral Loads, Journal of the American Concrete Institute 61 717-732. Wilson E. L. (1997) SAP2000 Integrated Finite Element Analysis and Design of Structures 1-2, C&S Inc., USA. 21 FREE VIBRATIONS OF CROSS-PLY LAMINATED NON-HOMOGENEOUS COMPOSITE TRUNCATED CONICAL SHELLS ˘ 1 2 3 1Department of Civil Engineering, Çukurova University, Adana, Turkey 2Technical Sciences Department of Kazakh Branch of Teachers Institute, Azerbaijan 3Department of Civil Engineering, SD , Isparta, Turkey Abstract. In this study, the free vibration of cross-ply laminated non-homogeneous orthotropic truncated conical shells is studied. At first, the basic relations have been obtained for cross-ply lam- inated orthotropic truncated conical shells, the Young’s moduli and density of which vary piecewise continuously in the thickness direction. Applying Galerkin method to the foregoing equations, the frequency of vibration is obtained. Finally, the effect of non-homogeneity, the number and ordering of layers on the frequency is found for different mode numbers, and the results are presented in tables and compared with other works. Key words: free vibration, laminated composite, conical shell 1. Introduction Laminated structural elements composed of non-homogeneous materials with dif- ferent elastic properties are frequently used in contemporary engineering applica- tions. These materials have properties that vary as a function of position in the body. Non-homogeneous materials can frequently be found in nature as well as in man-made structures. However, typically non-homogeneous materials seem to be those with elastic constants varying continuously in different spatial directions. Continuous non-homogeneity is a direct generalization of homogeneity in theory; besides, material non-homogeneity becomes essential and must sufficiently be considered in a number of practical situations (Lomakin, 1976). In recent years, noteworthy studies have been carried out on the stress-strain relations and vibrations of membranes and shells made of non-homogeneous ma- terials (Massalas et al., 1981 and Gutierrez et al., 1998). Leissa (1973) made a comprehensive study on the vibrations of plates and gave all previous studies in summary. Most of the recent works on the vibration Vibration Problems ICOVP 2005, c© 2007 Springer. – 21 26. and Ali SofiyevOrhan Aksogan , Abdullah H. Sofiyev Ü (eds.), . şE. Inan and A. Kırı 22 AKSOĞAN, SOFİYEV, SOFİYEV Figure 1. Geometry and the cross-section of a conical shell with N layers problems of cylindrical shells made of composite materials cover the vibration problems of composites (Lam and Loy, 1995, and Liew et al., 2002). In the vibration problems, the coverage of non-homogeneity of constituent materials renders the problem a bit more complicated. Moreover, in the case of laminated shells the difficulties increase and vibration characteristics need to be determined by a dependable computation method. In recent years, some studies have been carried out on the vibrations of cylindrical shells (Aksoğan and Sofiyev, 2000). The free vibration problem of laminated truncated conical shells, made of non-homogeneous cross-ply layers, has not been studied yet. The aim of the present work is to study the vibration of an orthotropic com- posite conical shell, the Young’s moduli and density of which vary piecewise continuously with the coordinate in the thickness direction. 2. Basic equations Consider a circular conical shell as shown in Fig. 1, which is assumed to be thin, laminated and composed of N layers of equal thickness δ of non-homogeneous orthotropic composite materials perfectly bonded together. A set of curvilinear coordinates (ς, θ, S ) is located on the middle surface. R1 and R2 indicate the radii of the cone at its small and large ends, respectively, 2h is the thickness, L is the length and γ is the semi-vertex angle. Furthermore, the ς-axis is always normal to the moving S -axis, lies in the plane generated by the S -axis and the axis of the cone, and points inwards. The θ-axis is in the direction perpendicular to the S − ς plane. The axes of orthotropy are parallel to the curvilinear coordinates S and θ. The Young’s moduli and densities of all layers are defined as continuous functions of the thickness coordinate ς as (Aksoğan and Sofiyev, 2000): [ E(k+1)S (ς̄) , E (k+1) θ (ς̄) , G(k+1) (ς̄) ] = ϕ̄(k+1)1 (ς̄) [ E(k+1)0S , E (k+1) 0θ , G (k+1) 0 ] , ρ(k+1) (ς̄) = ρ(k+1)0 ϕ̄ (k+1) 2 (ς̄) , −1 + kδ̄ ≤ ς̄ ≤ −1 + (k + 1) δ̄, δ̄ = δ/h, k = 0, 1, 2, ..., (N − 1) (1) whereE(k+1)0S and E (k+1) 0θ are the Young’s moduli in the S and θ directions for the layer k+1, respectively, G(k+1)0 is the shear modulus on the plane of the layer k+1, CROSS-PLY LAMINATED NON-HOMOGENEOUS COMPOSITE 23 and ρ(k+1)0 is the density of homogeneous orthotropic material of layer k+1, δ = 2h/N is the thickness of the layers. Additionally, ϕ̄(k+1)α (ς̄) = 1 + µϕ (k+1) α (ς̄) , α = 1, 2 (2) where ϕ(k+1)α (ς̄) are continuous functions giving the variations of the Young’s moduli and densities of the layers, satisfying the condition ∣∣∣∣ϕ(k+1)α (ς̄) ∣∣∣∣ ≤ 1, and µ is a variation coefficient satisfying 0 ≤ µ < 1. The middle surface ς = 0 is located at a layer interface for even values of N, whereas, for odd values of N the middle surface is located at the center of the middle layer. After lengthy computations, the free vibration and compatibility equations of a non-homogeneous laminated conical shell, in terms of a displacement function w and a stress function Ψ are obtained as follows: c12 ∂ 4Ψ ∂S 4 + c11+2c12−c22S ∂3Ψ ∂S 3 + ( cot γ S − c21 S 2 ) ∂2Ψ ∂S 2 + c21 S 3 ∂Ψ ∂S + c21 S 4 ∂4Ψ ∂θ41 + c11−2c31+c22 S 2 ∂4Ψ ∂S 2∂θ21 + 2(c31−c11) S 3 ∂3Ψ ∂S ∂θ21 + 2(c11−c31+c21) S 4 ∂2Ψ ∂θ21 − c24 S 4 ∂4w ∂θ41 − c14+c23+2c32 S 2 ∂4w ∂S 2∂θ21 + 2(c14+c32) S 3 ∂3w ∂S ∂θ21 − 2(c14+c32+c24) S 4 ∂2w ∂θ21 − c13 ∂ 4w ∂S 4 + c23−c14−2c13 S ∂3w ∂S 3 + c24 S 2 ∂2w ∂S 2 − c24 S 3 ∂w ∂S + cot γ S ∂2Ψ ∂S 21 −ρ̃h∂2w ∂t2 = 0, (3) b11 S 4 ∂4Ψ ∂θ41 + 2b31+b21+b12 S 2 ∂4Ψ ∂S 2∂θ21 − 2(b31+b21) S 3 ∂3Ψ ∂S ∂θ21 + 2(b31+b21+b11) S 4 ∂2Ψ ∂θ21 + b11 S 3 ∂Ψ ∂S − b11 S 2 ∂2Ψ ∂S 2 + b21+2b22−b12S ∂3Ψ ∂S 3 + b22 ∂ 4Ψ ∂S 4 − − b14 S 4 ∂4w ∂θ41 + 2b32−b13−b24 S 2 ∂4w ∂S 2∂θ21 + 2(b24−b32) S 3 ∂3w ∂S ∂θ21 + 2(b32−b24−b14) S 4 ∂2w ∂θ21 − − b14 S 3 ∂w ∂S + ( b14 S 2 + cot γ S ) ∂2w ∂S 2 + b13−b24−2b23 S ∂3w ∂S 3 − b23 ∂ 4w ∂S 4 = 0 (4) where t is time, ρ̃, ai j , bi j and ci j (i, j = 1, 2, 3, 4 and k1 = 0, 1, 2) are the parameters depending on the material properties and the characteristics of the laminated conical shell. 3. The solution of basic equations Consider a cross-ply laminated truncated conical shell with simply supported edge conditions. The solutions for equation (3) take the following form (Aksoğan and Sofiyev, 2000): w = S S 1 f (t) sin β1ξ sin β2ϕ (5) where f(t) is the time dependent amplitude and the following definitions apply: β1 = mπ ξ0 , β2 = n sin γ , ξ0 = ln S 2 S 1 , ξ = ln S S 1 (6) 24 AKSOĞAN, SOFİYEV, SOFİYEV Function (5) satisfies the following geometrical boundary conditions: w(S 1, θ) = w(S 2, θ) = 0; ∂2w(S 1, θ) ∂S 2 = ∂2w(S 2, θ) ∂S 2 = 0 (7) Substituting Eq. (5) into Eq. (3) F1 = (K1 sin β1ξ + K2 cos β1ξ + K3e −ξ sin β1ξ + K4e −ξ cos β1ξ) f (t) sin β2ϕ (8) where Ki(i = 1, 2, 3, 4) are the parameters depending on the material properties and the characteristics of the laminated truncated conical shell. Substituting Eqs. (5) and (8) into Eq. (4) and by applying Galerkin’s method, the following equation is obtained: ∂2 f (t) ∂t2 + Λ ρ̃h f (t) = 0 (9) where the following definition applies: Λ = (J1Γ1 + J2Γ2 + J3Γ3 + J4Γ4 + Γ8)/Γ9 (10) in which Ji(i = 1, 2, 3, 4) and Γi(i = 1, 2, 3, 4) are the parameters depending on the material properties and the characteristics of the laminated conical shell. The following expression is obtained from Eq. (9) for the dimensionless fre- quency parameter of free vibration: Ω = ωR2 √ (1 − ν2)ρ/E (11) where ω = √ Λ/(ρ̃h). 4. Numerical computations and results To verify the present analysis, our results are compared with those presented by Lam and Loy (1995) and Liew et al. (2002) for the homogeneous cross-ply lam- inated cylindrical shells of lamination scheme (0o/90o/0o). To be able to make a comparison with cross-ply laminated cylindrical shells, γ ≈ 0o must be substituted for the semi vertex angle and R2 ≈ R1 ≈ R must be assumed. The comparisons for the shell with the following material properties (Lam and Loy, 1995 and Liew et al., 2002): E(k+1)0S = 2.5E (k+1) 0θ , E (k+1) 0θ = 7.6 × 10 9N/m2, ν(k+1)S θ = 0.26, ν(k+1)θS = 0.104, ρ (k+1) 0 = 1643kg/m 3 and γ ≈ 0o are presented in Table 1. It is obvious that agreement is achieved for all cases. Numerical computations, for cross-ply laminated conical shells have been car- ried out using expression (11) and the results are presented in Table 2. In this table, the minimum values of the frequency parameter are given for varying values of CROSS-PLY LAMINATED NON-HOMOGENEOUS COMPOSITE 25 TABLE I. Comparisons of non-dimensional frequency parameter Ω = ωR √ (1 − ν212)ρ/E22 with (0o/90o/0o) for a simply supported laminated cylindrical shell (m=1, R2/h=1000, L/R=5) n Lam and Loy (1995) Liew et al.(2002) Present 3 0.0551 0.0551 0.0553 4 0.0338 0.0338 0.0337 5 0.0258 0.0258 0.02556 6 0.0259 0.0259 0.02557 TABLE II. Variations of non-dimensional frequency parameter Ω = ωR √ (1 − ν212)ρ/E22 for different numbers and ordering of layers and different semi-vertex angles γ for a simply supported cross-ply laminated conical shell (m=1, R2/h=1000, Lsinγ/R2=0.25) Ω(ϕ(k+1)α (ς̄) = ς̄,α = 1, 2) N Ordering γ = 30◦ γ = 30◦ γ = 45◦ γ = 45◦ of layers µ = 0 µ = 0.9 µ = 0 µ = 0.9 1 (0◦)1 0.909(17) 0.830(20) 0.581(16) 0.535(18) 2 (0◦/90◦)2 0.869(18) 0.822(20) 0.573(16) 0.539(18) 2 (90◦/0◦)2 0.884(19) 0.816(20) 0.574(17) 0.539(17) 3 (0◦/90◦/0◦)3 0.927(19) 0.840(20) 0.586(17) 0.534(18) 3 (90◦/0◦/90◦)3 0.870(18) 0.809(19) 0.583(15) 0.543(17) Ω(ϕ(k+1)α (ς̄) = ς̄2,α = 1, 2) 1 (0◦)1 0.909(17) 1.091(16) 0.581(16) 0.694(16) 2 (0◦/90◦)2 0.869(18) 1.035(17) 0.573(16) 0.681(15) 2 (90◦/0◦)2 0.884(19) 1.057(18) 0.574(17) 0.683(16) 3 (0◦/90◦/0◦)3 0.927(19) 1.111(18) 0.586(17) 0.701(16) 3 (90◦/0◦/90◦)3 0.870(18) 1.035(17) 0.583(15) 0.691(15) Ω(ϕ(k+1)α (ς̄) = ς̄3,α = 1, 2) 1 (00)1 0.909(17) 0.882(18) 0.581(16) 0.565(17) 2 (00/900)2 0.869(18) 0.857(19) 0.573(16) 0.563(17) 2 (900/00)2 0.884(19) 0.861(19) 0.574(17) 0.563(17) 3 (00/900/00)3 0.927(19) 0.899(19) 0.586(17) 0.571(17) 3 (900/00/900)3 0.870(18) 0.849(18) 0.583(15) 0.570(16) Note: The values in parentheses are the wave numbers. the vertex angle, different circumferential wave numbers, and different numbers and ordering of layers. The Young’s moduli and density variation function of the materials of the layers are assumed to be power functions which, together with the other composite material properties, are given as follows (Liew et al., 2002): 26 AKSOĞAN, SOFİYEV, SOFİYEV ϕ(k+1)α (ς̄) = ς̄ q(q = 1, 2, 3), α = 1, 2 E(k+1)0S = 2.5E (k+1) 0θ , E(k+1)0θ = 7.6 × 10 9N/m2, ν(k+1)S θ = 0.26, ρ (k+1) 0 = 1643 kg/m 3 It is easily observed that, as the number and ordering of the layers change, the non-dimensional frequency parameter takes its maximum value at different wave numbers. For the homogeneous case (µ = 0) and for the non-homogeneous case ϕ(k+1)α (ς̄) = ς̄q (q = 1, 2, 3) and (µ = 0.9), as the semi-vertex angle γ increases, the value of the non-dimensional frequency parameter decreases. It has been observed that the effect of the Young’s moduli and density variation on the non-dimensional frequency parameter is highest for being parabolic and lowest for being cubic. 5. Conclusions In this study, the free vibration of cross-ply laminated non-homogeneous orthotro- pic truncated conical shells is studied. At first, the basic equations have been obtained for cross-ply laminated orthotropic truncated conical shells, the Young’s moduli and density of which vary piecewise continuously in the thickness direc- tion. Applying Galerkin method to the foregoing equations, the frequency para- meter of vibration is obtained. Finally, the effect of non-homogeneity, the number and ordering of layers on the frequency is found for different mode numbers, and the results are presented in tables and compared with other works. References Aksoğan, O., and Sofiyev, A. H. (2000) The Dynamic Stability of a Laminated Non-homogeneous Orthotropic Elastic Cylindrical Shell under a Time Dependent External Pressure, Int. Con. on Modern Practice in Stress and Vib. Anal. Nottingham, UK, 349-360. Gutierrez, R. H., Laura, P. A. A., Bambill, D. V., Jederlinic, V. A., and Hodges, D. H. (1998) Axisymmetric Vibrations of Solid Circular and Annular Membranes with Continuously Varying Density, J. Sound and Vib. 212, 611-622. Lam, K. Y., and Loy, C. T. (1995) Analysis of Rotating Laminated Cylindrical Shells by Different Thin Shell Theories, J. Sound and Vib. 186, 23-35. Leissa, A. W. (1973) Vibration of Shells, NASA SP 288. Liew, K. M, Ng, T. Y., and Zhao, X. (2002) Vibration of Axially Loaded Rotating Cross-ply Laminated Cylindrical Shells via Ritz Method, J. Eng. Mecs. 128, 1001-1007. Lomakin, V. A. (1976) The Elasticity Theory of Non-homogeneous Materials (in Russian), Moscow, Nauka. Massalas, C., Dalamanagas, D., and Tzivanidis, G. (1981) Dynamic Instability of Truncated Conical Shells with Variable Modulus of Elasticity under Periodic Compressive Forces, J. Sound and Vib. 79, 519-528. Volmir, A. S. (1967) The Stability of Deformable Systems, Moscow, Nauka. 27 2. SYMMETRIC AND ASYMMETRIC VIBRATIONS OF CYLINDRICAL SHELLS Uğurhan Akyüz and Aybar Ertepınar Department of Civil Engineering, Earthquake Engineering Research Center, Middle East Technical University, Ankara, Turkey Abstract. The stability of cylindrical shells of arbitrary wall thickness subjected to uniform radial tensile or compressive dead-load traction is investigated. The material of the shell is assumed to be a polynomial compressible material which is homogeneous, isotropic, and hyperelastic. The governing equations are solved numerically using the multiple shooting method. The loss of sta- bility occurs when the motions cease to be periodic. The effects of several geometric and material properties on the stress and the deformation fields are investigated. Key words: hyperelastic, compressible, cylindrical shell 1. Introduction The earlier research on compressible, hyperelastic solids were mostly limited to determining a suitable model representing the behavior (Ericksen, 1955) - (Blatz and Ko, 1962), and to analyzing deformations and stresses in bodies of different geometries (Levinson and Burgess, 1971) - (Carroll and McCarthy, 1995). The objective of this work is to investigate the stability and free vibrations of cylindrical shells of arbitrary wall thickness subjected to external uniform, tensile or compressive dead load traction. The material of the shell is assumed to be compressible, isotropic, homogenous and hyperelastic. Under radial surface tractions, the shell undergoes first a finite static deformation, the vibrational be- havior and the associated stability characteristics of the pre-stressed shell are then investigated. 2. The analysis of cylindrical shells Consider a long, circular cylindrical shell of arbitrary wall thickness. The inner and the outer radii in the undeformed and the deformed configurations are re- spectively denoted by r1, r2, and R1,R2. Let the cylindrical coordinates (R, θ, z) be identified with θi in the deformed state B. Then, a mass point P, whose coordinates are (R(r), θ, z) in the deformed state B, is originally at (r, θ, z) in the undeformed Vibration Problems ICOVP 2005, c© 2007 Springer. – 27 3(eds.), . şE. Inan and A. Kırı 28 AKYÜZ, ERTEPINAR state B0. Referring the undeformed and the deformed configurations to the fixed rectangular coordinates xi and Xi, respectively (Figure 3.1), one has x1 = r cos θ, x2 = r sin θ, x3 = z (1) and X1 = R(r) cos θ, X2 = R(r) sin θ, X3 = z (2) The non-zero components of the metric tensor of the undeformed state B0 (gi j) and deformed state B (Gi j) are ( g11, g22, g33 ) = ( R′2, 1/r2, 1 ) , ( G11,G22,G33 ) = ( 1, 1/R2, 1 ) (3) In equation (3) a prime denotes differentiation with respect to r. The three strain invariants I1, I2, and I3 of the deformation field are I1 = R ′2 + R2 r2 + 1, I2 = R ′2 + R2 r2 + R2R′2 r2 , I3 = R2R′2 r2 (4) For bodies made of homogeneous, isotropic and hyperelastic materials, the com- ponents of the stress tensor τi j are given by τi j = 2 √ I3 ∂W ∂ I1 gi j + 2 √ I3 ∂W ∂ I2 Bi j + 2 √ I3 ∂W ∂ I3 Gi j (5) where Bi j = I1g i j − girg jsGrs (6) The associated boundary conditions for the shell, which is assumed to be free of tractions on its inner surface and subjected to a uniform tensile or compressive dead load on its exterior surface, are given as τ11(R1) = q1 = 0, τ 11(R2) = q2 = ∓q ( r2 R2 )2 (7) It is now further assumed that the strain energy density function W has the form W = µ 2 [ f (I1 − 3) + (1 − f ) ( I2 I3 − 3 ) + 2 (1 − 2 f ) ( √ I3 − 1 ) + (2 f + β) ( √ I3 − 1 )2] (8) which has been proposed by Levinson and Burgess (Levinson and Burgess, 1971) and has been named ‘polynomial compressible material’ by the authors. In equa- tion (8) µ is the shear modulus of the material for vanishingly small strains, f is a material constant whose value lies between zero and unity, and β is expressed as β = 4ν − 1 1 − 2ν (9) VIBRATIONS OF CYLINDRICAL SHELLS 29 where ν is the Poisson’s ratio for the material as the deformations become van- ishingly small. It is noted that, for highly elastic rubbers and rubber-like materials f=0 while for solid natural and synthetic rubbers f=1. The only non-zero equation of equilibrium for this finitely deformed state is the one in the radial direction which is given by ∂τ11 ∂R + τ11 − R2τ22 R = 0 (10) The shell is now exposed to a secondary dynamic displacement field described by w1 = u (R (r) , θ, t) w2 = Rv (R (r) , θ, t) w3 = 0 (11) to investigate the existence of small, free, radial vibrations about the finitely de- formed state. The formulation of this state is based on the theory of small deformations superposed on large elastic deformations1. The incremental metric tensors and the incremental stresses are given by G∗i j = wi, j + w j,i − 2Γ r i jwr, G ∗i j = −GirG jsG∗rs (12) τ∗i j = gi jΦ∗ + Bi jΨ∗ + B∗i jΨ + G∗i j p + Gi j p∗ (13) where Φ∗ = 2√ I3 ∂2W ∂I21 I∗1 + 2√ I3 ∂2W ∂I1∂I2 I∗2 + 2√ I3 ∂2W ∂I1∂I3 I∗3 − Φ 2I3 I∗3 Ψ∗ = 2√ I3 ∂2W ∂I1∂I2 I∗1 + 2√ I3 ∂2W ∂I22 I∗2 + 2√ I3 ∂2W ∂I2∂I3 I∗3 − Ψ 2I3 I∗3 p∗ = I3 ( 2√ I3 ∂2W ∂I1∂I3 I∗1 + 2√ I3 ∂2W ∂I2∂I3 I∗2 + 2√ I3 ∂2W ∂I23 I∗3 ) + p 2I3 I∗3 B∗i j = ( gi jgrs − girg js ) G∗rs (14) and Γri j are the Christoffel symbols of the second kind, and a comma denotes differentiation with respect to the following subscript. The incremental strain in- variants I∗1 , I ∗ 2, and I ∗ 3 are given by I∗1 = 2R ′2w1,R + 2r2 (w2,θ +Rw1) , I ∗ 3 = 2 R2R′2 r2 ( w1,R + 1R2 w2,θ + 1 R w1 ) I∗2 = ( 2R′2 + 2R ′2R2 r2 ) w1,R + ( 2 r2 + 2R ′2 r2 ) (w2,θ +Rw1) (15) The non-zero incremental equations of motion are ∂τ∗11 ∂R + ∂τ∗21 ∂θ + 1 R ( τ∗11 − R2τ∗22 ) + ( 2w1,RR + 1R2 w2,Rθ ) τ11 + w1,θθ τ22 + 1 R2 ( Rw1,R − 2R w2,θ −w1 ) ( τ11 + R2τ22 ) = ρẅ1 (16) 1 A detailed discussion of the theory has been given in reference (Green and Zerna, 1968) 30 AKYÜZ, ERTEPINAR ∂τ∗12 ∂R + ∂τ∗22 ∂θ + 3 Rτ ∗12 + 1 R2 ( w2,RR + 2R2 w2 − 2 R w2,R ) τ11 + ( w1,Rθ − 2R2 w2,θθ + 3 R w1,θ + 1 R w2,R − 2 R2 w2 ) τ22 = ρ R2 ẅ2 (17) in radial and in circumferential directions respectively. In equations (16) and (17) ρ is the mass density of finitely deformed body and a dot denotes differentiation with respect to time. The current mass density ρ is related to the mass density ρ 0 of the natural state by ρ = √ g √ G ρ0 (18) using the principle of conservation of mass. In equation (18), g and G denote, respectively, the determinants ∣∣∣gi j ∣∣∣ and ∣∣∣Gi j ∣∣∣. For this secondary state, the boundary conditions, which are obtained from the requirement that the secondary surface tractions vanish, are τ∗11 − G∗11τ11 = 0 @ R = R1 & R = R2 τ∗12 − G∗12τ11 = 0 @ R = R1 & R = R2 (19) The solution of equations (16) and (17) may be assumed to be of the form u = ∞∑ n=0 R∗1n (R) cos (nθ) e iωt, v = ∞∑ n=0 R∗2n (R) sin (nθ) e iωt (20) where ω is the frequency of vibrations about the finitely deformed state, n is the circumferential mode number, and R∗1n and R ∗ 2n are unknown functions of R which, in turn, is a function of r. The case n = 0 corresponds to pure radial vibrations about the prestressed state. The case n = 1 includes rigid body motions and is left out of discussion with the understanding that the rigid body motions of the shell are prevented. The higher modes corresponding to n ≥ 2 are considered in what follows. The governing equations of both states are non-dimensionalized by introduc- ing r̄ = rr2 , R̄ = R r2 , R̄∗1n = R∗1n r2 , R̄∗2n = R∗2n r2 , τ̄11 = τ 11 µ , q̄ = q µ , ū = u r2 , ω̄ = ω √ ρ0 r22 µ (21) For the finitely deformed state, no closed form solution of the highly non-linear system of equations, equation (10), seems possible. To solve these equations nu- merically, the boundary value problem is converted to an initial value problem by using the multiple shooting method. The solution of finitely deformed state equation of equilibrium together with the corresponding boundary conditions is obtained by using a FORTRAN code called BVPSOL and developed by Deuflhard and Bader (Deuflhard and Bader, 1982). Equations of motion, equations (16) and (17) governing the secondary state are coupled linear ordinary differential equations. Since no closed form solution VIBRATIONS OF CYLINDRICAL SHELLS 31 a. f = 0.0 and ν = 0.25 b. f = 1.0 and ν = 0.49 Figure 1. Natural frequencies of vibrations versus shell thickness is available for the finitely deformed state, a numerical approach must be used to obtain the solution of this state also. The method of complementary functions is used to transform the linear boundary value problem to an initial value problem. 3. Discussion of the results In this study, the vibrational characteristic and the loss of stability of hollow circular cylindrical shells are investigated using the theory of finite elasticity in conjunction with the theory of small deformations superposed on large elastic deformations. For breathing motions (i.e. when n = 0), and asymmetric vibrations (i.e. when n ≥ 2), several examples are worked out numerically to study the effects of material constants f and ν, the outer stretch ratio (λ = R̄2/r̄2) and the thickness ratio of the shell (χ = r̄1/r̄2) on the non-dimensionalized natural frequencies of vibrations ω̄, and the critical outer stretch ratio, λcr, which is defined as the stretch ratio of the outer surface when the frequency approaches to zero. In particular, the behaviors of the foam rubber and the slightly compressible hyperelastic materials represented by polynomial model are investigated. When the applied dead load traction is equal to zero the shells are unstrained, and the natural frequencies of the vibrations about the natural state of the shell can be calculated. In Figure (1), the frequencies of vibrations about the unstrained shells are given as a function of shell thickness χ for a foam rubber, and a polyno- mial material with f=1.0 and ν=0.49. It is seen that for the asymmetric vibrations the frequencies increase as the shell becomes thicker while for breathing motion the frequencies increase as the shell becomes thinner. It is also observed that, for the asymmetric vibrations higher harmonics correspond to higher frequencies for a given shell thickness. Figure (2) show the variation of the frequency ω̄ as a function of the radial stress on the outer surface, for a foam rubber with with f=0.0, ν=0.25, and for a polynomial material with f=0.9, ν=0.49, respectively, having a shell 32 AKYÜZ, ERTEPINAR a. f = 0.0, ν = 0.25 and χ = 0.80 b. f = 0.9, ν = 0.49 and χ = 0.80 Figure 2. The effect of harmonic mode number n on frequencies versus outer radial stress thickness χ = 0.80. Four different modes, corresponding to the harmonic numbers n = 0, 2, 3, 4, are shown both in the tension and in the compression region. It is seen that, in the tension region, the lowest frequency of vibrations depends on the harmonic number n, and the state of the initial strain. A similar behavior is also observed for natural frequencies (see Figure 1). In the early stage of the inflation region, n = 2 is found to be the governing mode of vibration, and the shell behaves as a hardening system. But, as the tensile dead load traction increases, the breathing mode, n = 0, becomes the governing mode. In this stage of the inflation process, the shell behaves as a softening system, and as the limiting pressure τ̄11max is reached, the shell fails in radial expansion without bound. When the shell is subjected to an increasing compressive dead load traction, the behavior mentioned above is reversed. The hardening effect is now associated with the breathing mode, while the asymmetric modes, n = 2, 3, 4, have a softening effect on the system. It is also seen that, in the compression region, n = 2 is the governing mode of vibration. References Blatz P. J. and Ko W. L. (1962) Application of Finite Elasticity Theory to the Deformation of Rubbery Materials, Trans. Soc. Rheo. 6, 223. Carroll M. M. and McCarthy M. F. (1995) Conditions on Elastic Strain Energy Function, Zangew Math. Phys. 46, S172. Deuflhard P. and Bader G. (1982) SFB 123, Tech. Rep. 163, University of Heildelberg. Ericksen J. L. (1955) Deformations Possible in Every Compressible, Isotropic, Perfectly Elastic Body, J. Math. Phys. 34, 198. Green A. E. and Zerna W. (1968) Theoretical Elasticity, 2nd Edition, Oxford. Levinson M. and Burgess I. W. (1971) A Comparison of Some Simple Constitutive Equations for Slightly Compressible Rubber-Like Materials, Int. J. Mech. Sci. 13, 563. Mooney M. (1940) A Theory of Large Elastic Deformations, J. Appl. Phys. 11, 582. 33 AN ACTIVE CONTROL ALGORITHM TO PREVENT THE POUNDING OF ADJACENT STRUCTURES Ünal Aldemir and Ersin Aydın Faculty of Civil Engineering, İstanbul Technical University, İstanbul, Turkey Abstract. Since the adjacent structures can collide due to their out-of-phase vibration, structural pounding is an additional problem except for the other harmful damage that occurs during the earthquake vibrations. In this study, a simple active control algorithm which results in an asymptot- ically stable system is proposed for adjacent structures. To be able to prevent a possible pounding, proposed control is calculated based on the measured relative displacement of the structures. The structural responses of the example structure subjected to El Centro ground motion are investigated for optimal passive control and proposed control and compared to uncontrolled case. It is shown by numerical results that the proposed active control algorithm can reduce the response of the adjacent structures under the earthquake forces and prevent pounding. Key words: active control, pounding, output state feedback control 1. Introduction Earthquakes cause many harmful effects in the structures. A significant amount of seismic energy can be absorbed with supplemental energy dissipation systems. Different supplemental energy dissipation systems have been proposed to pre- vent harmful effects of the strong ground motions. In addition to these passive systems, active and semi-active systems have also been employed by (Bakioglu and Aldemir, 2001) and (Aldemir, 2003). Buildings in a crowded city are often built closely to each other because of the limited availability of land. In most cases, these buildings have been built without link elements between each other. The interaction of adjacent buildings has been repeatedly identified as a common cause of damage. The possibility of added active or passive control devices to link adjacent structures has been improved in recent decades. This application takes advantages of the interaction between adjacent structures which is expected not only to prevent the pounding problem of the adjacent structures but also to reduce seismic response of the adjacent structures. Structural pounding has been stud- (Athanasiadou et al., 1994). After strong earthquakes the reason of numerous field damages and collapses was pounding. Especially, in the big cities, damage Vibration Problems ICOVP 2005, c© 2007 Springer. – 33 3 ied and reported widely in literature (Anagnostopoulos, 1995), (Bertero, 1986), 8.(eds.), . şE. Inan and A. Kırı 34 ALDEMİR, AYDIN and harmful effects have occured because of pounding in the cente! r of cities. In the Mexico City earthquake in 1985 had attributed to pounding a big part of the observed damage (Anagnostopoulos, 1995 ), (Bertero, 1986). In addition strength building may be benefit to weak building provided that pounding will not cause any serious local damage from which failure could be initiated (Bertero, 1986), (Athanasiadou et al., 1994). An innovative control approach was proposed which can be realized by means of installing an interaction element between two parallel structures (Zhu et al., 2001). The possibility of using dissipative links and semi active devices to reduce the response of adjacent buildings under the wind excitation was studied (Klein et al., 1972). The two adjacent buildings were modeled by a couple of uniform shear beams coupled by a single flexible and damped link and obtained the optimal stiffness and damping in the link in case of primary structures under the wind excitation (Gurley et al., 1991). 2. Formulation of the problem Without Control The interaction control strategy takes advantage of the configuration and in- teraction of two different structural systems (Primary-Auxiliary) to reduce the vi- bration response of primary-auxiliary structure (P-A structure) during the ground motion. In this study, only response of the P-A structures is reduced the simplest form as shown in Fig. 1. The mass of P-structure is larger than the mass of A- structure. This may cause different vibration modes between P and A structures. P and A structures can be simplified to two single-degree-of-freedom-systems without control element subject to ground excitation as shown in Fig. 1. Without control element, equation of motion is given as follows: mpx .. p + cpx . p + kpxp = −mpx .. g (1) max .. a + cax . a + kaxa = −max .. g (2) where mp, ma are first modal mass of the P and A structures, kp, ka are stiffness coefficients of the P and A structures, cp, ca are damping coefficient of the P and A structures, xp, xa represent relative displacement of the P and A structures with respect to ground, x .. g represents horizontal ground acceleration. The new variable z is defined as: z = xa − xp (3) where z represents relative displacement between adjacent structures. First and second order differentiation can be written as in following form: xa = z + xp =⇒ x . a = z . + x . p =⇒ x .. a = z .. + x .. p (4) POUNDING OF ADJACENT STRUCTURES 35 xa caka xp mamp )t(xg cpkp Figure 1. The simplest mechanical model of adjacent structure without control. If eq. (4) is used in eq. (2), eq. (2) can be written depending the displacement of primary structures and relative displacement between the adjacent structures as follows: ma(x .. p + z .. )! + ca(x . p + z . ) + ka(xp + z) = −max .. g (5) Eqs. (1) and (5) are written in matrix form as follows: [ x .. p z .. ] = −M−1C [ x . p z . ] − M−1K [ xp z ] + M−1 [−mp −ma ] H x .. g (6) In state space form, eq. (6) is written as X . = AX + Ex .. g (7) in which X = xp z x . p z . where, A = [ 02x2 I2x2 −M−1K −M−1C ] E = [ 02x1 M−1H ] (8) Optimal Passive Control A passive energy dissipation element is implemented between two structures, as shown in Fig. 2. The optimum parameters of the implemented passive device (Zhu et al., 2001) kopt = kp µ(1 − β12) (1 + µ)2 copt = mpωp (1 + µ) √ µ(1 + β12)2 (1 + µ)(µ + β12) (9) 36 ALDEMİR, AYDIN Up kop cop xa caka xp mamp )t(xg cpkp Figure 2. The simplest mechanical model of adjacent structure optimal passive control where β1 = ωa ωp µ = mp ma . Then, the passive control force, Up = −koptz − coptz . (10) Control force vector U U = [−Up Up ] = −G1p [ xp z ] − G2p [ x . p z . ] (11) where G1p = [ 0 −kopt 0 kopt ] , G2p = [ 0 −copt 0 copt ] under the effects of U, eq. (6) is rewritten as [ x .. p z .. ] = −M−1(C + G2p) [ x . p z . ] − M−1(K + G1p) [ xp z! ] + M−1Hx .. g (12) in eq. (6), K and C are replaced by (K + G1p) and (C + G2p), respectively. Output State Feedback Control Active force actuators are placed on each structure as shown in Fig. 3. It is assumed that the required control energy to generate the calculated forces is available and time delay and force saturation effects are neglected. u1 and u2 are given as u1 = −g1xp − g2x . p (13) u2 = −g3xa − g4x . a (14) Using z = xa − xp eq. (14) is u2 = −g3z − −g3xp − g4z . − g4x . p (15) POUNDING OF ADJACENT STRUCTURES 37 u1 u2 xa caka xp mamp )t(xg cpkp Figure 3. The simplest mechanical model of adjacent structure active control Then, the control force vector U for this case is U = [ u1 u2 ] = −G1a [ xp z ] − G2a [ x . p z . ] (16) where G1a = [ g1 0 g3 g3 ] , G2a = [ g2 0 g4 g4 ] is positive definite matrices. For active case, the stiffness and damping matrices K and C of uncontrolled structure in eq. (6) are replaced by (K + G1a) and (C + G2a), respectively. 3. Numerical examples Example adjacent structures are chosen as shown in Fig 1. Dynamic parameter of the structures are given as ma = 0.75105 kg ka = 2.087106 N/m ca = 3.165104 kg/s mp = 1.5105 kg kp = 1.67107 N/m cp = 6.33104 kg/s G1a and G2a are selected as G1a = 106 [ 1 0 1 1 ] G2a = 106 [ 1 0 1 1 ] . For three cases, the response of adjacent structures is calculated with time history analysis under El Centro earthquake. It is shown in Fig. 4 uncontrolled relative displacement is excessive and pounding may occur depending on distance between structures. So passive control may decrease the relative displacement, but it may not be suffi- cient. Proposed active control gives the best performance and prevents pounding. 38 ALDEMİR, AYDIN -0.18 -0.16 -0.14 -0.12 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0 2 4 6 8 10 12 14 Time (sec) R e la ti v e D is p la c e m e n t Z ( m ) Without Control Passive Control Output State Feedback Control Figure 4. Relative displacement time history for three cases 4. Conclusions If the dynamic characteristics of the adjacent structures are different, pounding may occur because of their out-of-phase vibration according to distance of ad- jacent structures. In this study, a simple active control algorithm is proposed to prevent pounding for adjacent structures. Output state feedback control is calcu- lated based on the measured relative displacement of the structures. The results of proposed active control method to prevent pounding of adjacent structures show that the control performance of active control is better than the optimal passive control and uncontrolled case. References Aldemir U. (2003) Optimal control of structures with semiactive-tuned mass dampers, Journal of sound and vibration 266 847-874. Anagnostopoulos S. A. (1995) Earthquake induced pounding, State of the art Proceedings of the 10th European Conference on Earthquake Engineering 2 897-905. Athanasiadou C. J., Penelis G. G. and Cappos A. G. (1994) Seismic response of adjacent buildings with similar or different dynamic characteristics, Earthquake spectra 10. Bakioglu M, Aldemir U. (2001) A new numerical algorithm for sub-optimal control of earthquake excited linear structures, Int. Journal of Numer. Meth. Engng. 50 2601-2616. Bertero V. V. (1986) Observations on structural pounding Proceedings of The International Conference on Mexico Earthquakes, ASCE 264-287. Gurley K., Kareem A., Bergman L. A., Johnson E. A. and Klein R. E. (1994) Coupling tall buildings for control of response to wind, Proc. Sixth Int. Conf. on Structural Safety and Reliability (ICOSSAR) AA Balkema Publishers, Rotherdam 1553-1560. Klein R. E., Cusano C. and Stukel J. (1972) Investigation of a method to stabilize wind induced oscillations in large structures, ASME 72 WA/UUT-H. Zhu H., Wen Y. and Iemura H. (2001) A study on interaction control for seismic response of parallel structures, Computers and Structures 79 231-242. 39 VIBRATION CONTROL OF NON-LINEAR BUILDINGS UNDER SEISMIC LOADS Ünal Aldemir and Deniz Güney Faculty of Civil Engineering, İstanbul Technical University, İstanbul, Turkey Abstract. An optimal control law is derived using a simple integral type quadratic functional. The resulting control scheme is applied to seismically excited non-linear buildings modeled as lumped mass shear frame structures. The non-linearity is reflected by the non-linear stiffness restoring force. Active tendon actuators are considered as control devices. The performance of the proposed control is compared to those of the uncontrolled structure and the passive base isolation. It is shown by numerical simulation results that the proposed control is capable of suppressing the uncontrolled seismic vibrations of the non-linear structures. Key words: non-linear structures, vibration control, optimal control 1. Introduction Passive and active control systems have found increasing applications to alleviate the seismic hazards in civil engineering structures. Passive single and multiple mass dampers have been installed in civil structures to mitigate the wind or earth- quake induced vibrations (Chen and Wu, 2001), (Park and Reed, 2001), (Aldemir et al., 2001). The structure having a passive damper or a base isolation system is first constructed, and then it waits for the force to occur. Then, the structure responds to the induced forces without any modification in performance char- acteristics of the manufactured elements during the time history of the exciting force. Although the passive devices are simple and yet completely applicable in civil structures, they have limitations in achieving desired performance measures primarily due to the highly uncertain nature of the wind and seismic disturbances. Consequently, many active vibration control methodologies for civil structures have been explored in recent years (Aldemir et al.,, 2001), (Yang et al., 1987). In this study, a new simple integral type quadratic functional between the successive control instants is used in deriving the optimal control force. The proposed con- trol is applied to non-linear structures incorporating a full active tendon system. Numerical simulation results indicate that the proposed control can significantly Vibration Problems ICOVP 2005, c© 2007 Springer. – 39 reduce the earthquake- induced vibrations of non-linear structures. 44.(eds.), . şE. Inan and A. Kırı 40 ALDEMİR, GÜNEY 2. Control algorithm To give a brief overview of the active control concept, the governing equations of motion for a general multistory shear-beam lumped mass linear building structure are MX .. (t) + CX . (t) + KX(t) = D1 f (t) + D2U(t), t ∈ [to, t1]; X(to) = X . (to) = 0 (1) where X(t) = (x1, ..., xn)Tis the n-dimensional response vector denoting the rela- tive displacement of the each story unit with respect to the ground; the superposed dot represents the differentiation with respect to time; M is the (nxn)-dimensional diagonal constant mass matrix with diagonal elements mi= mass of i th story (i = 1, ..., n); C and K are the (nxn)-dimensional viscous damping and the stiffness matrices, respectively; D1 is the (nx1)-dimensional location matrix of excitation and given by DT1 = −(m1, ...,mn); D2 is the (nxr)-dimensional location matrix of r controllers; U(t) is the r-dimensional active control force vector and described as UT (t) = (u1(t), ..., ur(t)) and scalar function f (t) is the one dimensional earthquake acceleration. In the state space, Eq. (1) becomes Z . (t) = AZ(t) + BU(t) + D f (t), t ∈ [to, t1]; Z(to) = 0 (2) A = [ 0nxn I −M−1K −M−1C ] ; B = [ 0nxr M−1D2 ] ; D = [ 0nx1 −η ] ; Z = [ X X . ] (3) such that 0nxmis the (nxm)-dimensional zero matrix; I is the (nxn)-dimensional identity matrix; η = (1, ..., 1)T is the n-dimensional vector. In this study, to deter- mine the active control force, we introduce the integral type quadratic functional It,h(U) = ∫ t+h t [ZT (s)QZ(s) + UT (s)RU(s)]ds → min (4) with two parameters, t ≥ to and h > 0 where Q is the (2nx2n)-dimensional positive semi-definite symmetric weighting matrix and R is the (rxr)-dimensional positive definite symmetric weighting matrix; h is the time interval between the successive control instants. It is clear that the problem of minimizing the functional (4) under the condition (2) is an optimal control problem. The solution of this problem is given in detail by (Aldemir et al., 2001) as follows; U(t) = −hN−10 R −1BT QZ(t) (5) Z(t) = [I + h2 2 BN−10 R −1BT Q]−1[y(t − h) + h 2 D f (t)] (6) No = I + h2 4 R−1BT QB] (7) VIBRATION CONTROL UNDER SEISMIC LOADS 41 y(t − h) = eAhZ(t − h) + (h 2 )eAh[BU(t) + DF(t)] (8) 3. Optimal control of non-linear structures To treat the nonlinear problem of a structural system, Eq. (1) takes the form MX .. (t) + FD(t − h) + C∗(t − h)[X . (t) − X . (t − h)] + FS (t − h) + K∗(t − h) [X(t) − X(t − h)] = D1 f (t) + D2U(t) t ∈ [to, t1]; X(to) = X . (to) = 0 (9) where both FD(t) and FS (t) are general non-linear functions of X . (t) and X(t); K∗(t−h) and C∗(t−h) are the influence coefficient matrices of which the elements represent the tangent stiffness and damping at time (t − h). Using the Wilson-θ numerical integration procedure and the previously defined state vector concept with the assumption of linear viscous damping (FD(t − h) = CX . (t − h)), the solution of Eq. (9) can be expressed as follows: Z(t) = y∗(t − h) + A1 f (t) + A2U(t) (10) in which y∗(t − h) = A3Z(t − h) + A4y∗(t − h) = A3Z(t − h) + A4FS (t − h) + A5 f (t) + A6U(t − h) (11) In Eqs. (9)-(11), A j ( j = 1, ..., 6) are vectors or matrices defined as follows A1 = θ−2 [ T1 (3/h)T1 ] ; A2 = θ−2 [ T2 (3/h)T2 ] ; A3 = [ I θ−2ET3 0 I + θ−2ET4 ] ; A4 = θ−2 [ E T5 E T6 ] ; A5 = −θ−2 [ E T7 E T8 ] D1; A6 = −θ−2 [ E T7 E T8 ] D2; (12) in which I is an (nxn) identity matrix, θ is Wilson-θ constant and chosen as 1.37, and T1 = −ED1; T2 = −ED2; T3 = (6/h)M+ [3(θ− 1)I − D1]C+ h(θ2 − 1)K∗(t− h); T4 = (6/h)I + S2]C − 3K∗(t − h); T5 = −(3I + S1); T6 = −(6/h)I − S2; T7 = 2I + S1; T8 = (3/h)I + S2; E = [(6/(θh)2 M + (3/(θh)C + K∗]−1; S1 = [h(1.5θ − 1)C+ 0.5h2(θ2 − θ)K∗]M−1; S2 = [3(θ− 1)C+ hθ(θ− 1.5)K∗]M−1 Then, using the same procedure, the optimal control force and the corresponding state vectors for 42 ALDEMİR, GÜNEY m8 m7 m2 m1 m8 m7 m2 m1 Figure 1. Structural model a)Active tendon system b) Base isolated system the non-linear structure are derived as follows; U(t) = −2N−10 R −1 AT1 QZ(t) and Z(t) = [I + 2A1N−1o R−1 A T 1 Q] −1[y∗(t − h) + A2 f (t)] 4. Numerical examples In order to evaluate the effectiveness and the performance of the proposed con- trol algorithm, simulations for the controlled and uncontrolled response of the investigated non-linear structures are conducted. The seismic input to the example structures is ErzincanNS (Erzincan Earthquake; Dec 26, 1939). An 8-story build- ing shown in Fig. 1.a that exhibits bilinear elastic-plastic behavior is considered and a full active tendon control system in which the control is implemented by means of actuators exerting forces at each floor is used. It is assumed that the columns of the building are massless and the mass of the structure is concen- trated at floor levels. Each story has a mass of 155.6 ton, and a linear viscous damping coefficient of 1097.7 kN sec/m. In the study, 2 Cases are considered. In Case 1, the non-linearity for the 8-story structure is reflected by the stiffness restoring force FS [X(t)]. The yield displacement of each floor is 2 cm while the pre- and post-yield stiffness coefficients of each floor are k1=153333 kN/m and k2=45333 kN/m, respectively. Uncontrolled hysteresis loops for the shear forces fs1 , ..., fs8 in each floor of the non-linear structure are shown in Fig. 2. Without any control system, it is clear that the deformation of the unprotected structure is excessive and that yielding occurs in the first six floors. Hence, to suppress the excessive deformations, a full active tendon control system is implemented with the weighting matrices. Controlled hysteresis loops for the shear forces are shown in Fig. 3. It indicates that how effective the active tendon system with the proposed control is in reducing the deformations since all interstorey drifts are within the elastic range. This result can be seen obviously also in Fig. 3 in terms of the reduction in displacements. In Case 2, 8-storey structure subjected to Erzincan NS earthquake is isolated by rubber bearings as shown in Fig. 1.b. The superstructure is assumed to be linear elastic with identical story elastic stiffness coefficient of 153333 kN/m while the base isolation material exhibits bilinear elastic-plastic VIBRATION CONTROL UNDER SEISMIC LOADS 43 -10 -5 0 5 -10000 -5000 0 5000 fs 1 (k N ) -10 -5 0 5 -10000 -5000 0 5000 fs 2 (k N ) -8 -6 -4 -2 0 2 4 -10000 -5000 0 5000 fs 3 ( kN ) -6 -4 -2 0 2 -5000 0 5000 fs 4 ( kN ) -4 -3 -2 -1 0 1 2 -5000 0 5000 fs 5 ( kN ) -3 -2 -1 0 1 2 -4000 -2000 0 2000 fs 6 ( kN ) -2 -1.5 -1 -0.5 0 0.5 1 -4000 -2000 0 2000 fs 7 ( kN ) -1 -0.5 0 0.5 -2000 -1000 0 1000 fs 8 ( kN ) Relative displacement (cm)Relative displacement (cm) Figure 2. Uncontrolled hysteresis loops for the shear forces -2 -1 0 1 2 -5000 0 5000 fs 1 ( kN ) -2 -1 0 1 2 -4000 -2000 0 2000 fs 2 ( kN ) -1.5 -1 -0.5 0 0.5 1 1.5 -4000 -2000 0 2000 fs 3 ( kN ) -1.5 -1 -0.5 0 0.5 1 -2000 0 2000 fs 4 ( kN ) -1 -0.5 0 0.5 1 -2000 0 2000 fs 5 ( kN ) -1 -0.5 0 0.5 1 -2000 -1000 0 1000 fs 6 ( kN ) -0.6 -0.4 -0.2 0 0.2 0.4 -1000 0 1000 fs 7 ( kN ) -0.3 -0.2 -0.1 0 0.1 0.2 -500 0 500 fs 8 ( kN ) Relative displacement (cm)Relative displacement (cm) Figure 3. Active tendon system controlled hysteresis loops for the shear forces behavior. Properties of the base isolation system are: elastic stiffness of the base isolators kb1=158.1 kN/m; post-elastic stiffness of the base isolators kb2=30.1 kN/m; damping coefficient cb= 18.9 kN sec/m; mass of base isolation mb=160 ton and the yielding deformation (xb)y=10 cm. Uncontrolled time response of base isolation and the hysteresis loop of shear force in the isolators are shown in Fig. 4. It is obvious that excessive deformations occur in the isolators and the isolation system is not within the elastic range. To prevent a possible instability failure due to the excessive base deformations, active tendon system is implemented. Controlled base isolation response and the corresponding hysteresis loop for the base shear force are shown in Fig. 5. It is observed from Fig. 5 that if the active tendon forces are regulated based on the proposed control algorithm, the excessive isolation deformation is significantly reduced so that it works within the elastic range and instability failure can be prevented. 44 ALDEMİR, GÜNEY -2 -1 0 1 2 -5000 0 5000 fs 1 ( kN ) -2 -1 0 1 2 -4000 -2000 0 2000 fs 2 ( kN ) -1.5 -1 -0.5 0 0.5 1 1.5 -4000 -2000 0 2000 fs 3 ( kN ) -1.5 -1 -0.5 0 0.5 1 -2000 0 2000 fs 4 ( kN ) -1 -0.5 0 0.5 1 -2000 0 2000 fs 5 ( kN ) -1 -0.5 0 0.5 1 -2000 -1000 0 1000 fs 6 ( kN ) -0.6 -0.4 -0.2 0 0.2 0.4 -1000 0 1000 fs 7 ( kN ) -0.3 -0.2 -0.1 0 0.1 0.2 -500 0 500 fs 8 ( kN ) Relative displacement (cm)Relative displacement (cm) Figure 4. Interstorey drifts with and without control 0 2 4 6 8 10 -30 -20 -10 0 10 20 30 40 x b i (c m ) time (sec) -40 -20 0 20 40 -20 -15 -10 -5 0 5 10 15 20 25 F s b i (k N ) relative displacement (cm) 0 2 4 6 8 10 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 x b i (c m ) time (sec) -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 4 5 F s b i (k N ) relative displacement (cm) Figure 5. Uncontrolled/controlled time response of base isolation and the hysteresis loop 5. Conclusions A simple integral type quadratic functional is introduced for the purpose of sup- pressing the seismic vibrations of nonlinear buildings. It is demonstrated by sim- ulation results that the application of the proposed control to non-linear structures can prevent the structural failures and keep the uncontrolled structures within the elastic range. References Aldemir U., Bakioglu M., Akhiev S. S. (2001) Optimal control of linear structures, Earthquake Engineering and Structural Dynamics 30 835-851. Chen G. D., Wu J. N. (2001) Optimal placement of multiple tuned mass dampers for seismic structures, Journal of Structural Engineering ASCE 127 1054-1062. Park J., Reed D. (2001) Analysis of uniformly and linearly distributed mass dampers under harmonic and earthquake excitation, Engineering Structures 23 802-814. Pourzeynali S., Datta T. K. (2002) Control of flutter of suspension bridge deck using TMD, Wind and Structures 5 407-422. Rasouli S. K., Yahyai M. (2002) Control of response of structures with passive and active tuned mass dampers, Structural Design of Tall Buildings 11 1-14. Yang J. N., Akbarpour A., Ghaemmaghami P. (1987) New Optimal Control Algorithms for Structural Control, Journal of Engineering Mechanics ASCE 113 1369-1386. 45 VIBRATIONS OF DAMAGED 1D-3D MULTI-STRUCTURES ∗ Adolf G. Aslanyan1, Janusz Dronka2, Gennady S. Mishuris2 and Özgür Selsil3 1Druzhba St. 5, kv. 355, Khimki, Moscow Region, 141400, Russia 2Rzeszów Univ. of Tech., Dept. of Mathematics, W. Pola 2. 35-959, Rzeszów, Poland 3The Univ. of Liverpool, Dept. of Mathematical Sciences, L69 7ZL, Liverpool, UK Abstract. The objective of the paper is to study the vibrations of 1D-3D nondegenerate multi- structures, which are partly damaged. Such an analysis is useful to quickly asses the condition of a structure which has been in use over a long period of time. We look at three different positions of damage: top, middle or bottom of one leg. We calculate six eigenfrequencies associated with rigid body motion of the 3D structure and derive simple analytical asymptotic formulae for these eigenfrequencies. Hence, we analyse the effect of weakening of a leg. Accuracy of the asymptotic formulae is also checked with 3D finite element computations. Key words: multi-structures, partial damage, asymptotics, eigenfrequencies 1. Introduction A number of papers on eigenvalue problems posed for 2D-3D and 1D-3D multi- structures have been published in recent years (Aslanyan et al., 2002, 2003, 2003). These papers were motivated by the asymptotic study of static and spectral prob- lems associated with 1D-3D multi-structures (Kozlov et al., 1999). More recently the transition between nondegeneracy and degeneracy in 1D-3D multi-structures has been studied (Aslanyan et al., 2005). To our best knowledge, the only study on inhomogeneity in multi-structures is associated with high contrast mass densities between the 3D part and legs of a degenerate 1D-3D multi-structure (see the last chapter of Kozlov et al., 1999). The current problem is an extension to the study of structural damage on the vibrations of 1D-3D multi-structures (Aslanyan et al., 2005). We model the legs to have piecewise homogeneous sections. For the sake of simplicity, we consider only one of the legs to have this property. 2. General considerations for inhomogeneous nondegenerate multi-structures Here, we first formulate an eigenvalue problem of 3D linear elasticity, as described in Kozlov et al., 1999, and discuss the general asymptotics for the skeleton (pile structure) of a 1D-3D nondegenerate inhomogeneous multi-structure. ∗ supported by the EU Marie Curie grant (ToK scheme), contract number MTKD-CT-2004- 509809 Vibration Problems ICOVP 2005, c© 2007 Springer. .– 45 50(eds.), . şE. Inan and A. Kırı 46 ASLANYAN, DRONKA, MISHURIS, SELSİL 2.1. PROBLEM FORMULATION We consider the following eigenvalue problem: −Lu = ρω2u, x ∈ Ω ∪Kj=1 Π ( j) ε , [σ(n)(u)] = 0, [u] = 0, x ∈ S int, σ(n)(u) = 0, x ∈ ∂Ωε \ S ε, u = 0, x ∈ S ε, where L = µ∇2 + (λ + µ) grad div is the Lamé operator (λ and µ are the Lamé elastic moduli), u is the displacement vector, σ(n)(u) = σn is the stress vector, ρ is the mass density, ω is the corresponding eigenfrequency and x = (x1, x2, x3) are the Cartesian coordinates in R3. Here Ω is the 3D cap, Π ( j) ε , j = 1, 2, . . . , K are the thin cylinders and S ε is the union of their base regions. S int is used for all internal surfaces where the material parameters have discontinu- ities. These discontinuities are denoted by, for example, [u] for the displacements. We assume that λ, µ and ρ are constants in each subdomain. 2.2. BRIEF DESCRIPTION OF THE GENERAL ASYMPTOTICS We assume that the 3D body moves like a rigid body to the leading order, i.e. v = α + β × x in Ω, where α and β are constant vectors. We also assume that the displacements above are valid under the effect of body forces given by Ψ = c+d×x, where c and d are constant vectors. The dependency between the vectors v and Ψ can be established by obtaining the resultant force and the resultant moment (which appear in the equations of equilibrium) in terms of the vectors α, β, c and d (see, for example, Aslanyan et al., 2003). It was shown in Aslanyan et al., 2002 and 2003 that the spectral parameter associ- ated with the skeleton is found by t ≡ ρν2 = minv�0{ ∫ Ω (v,Ψ(v)) dx/ ∫ Ω (v, v) dx}, which can be reduced to the following characteristic equation: det(Φ − t Γ) = 0. Here, Φ and Γ are 6×6 matrices ; they are symmetric and positive definite, and all the roots of the characteristic equation above coincide with the positive definite matrix ΦΓ−1. Hence, they are all positive: 0 ≤ t1 ≤ t2 ≤ . . . ≤ t6. Therefore, it is possible to approximate the values of the first six eigenfrequencies of the multi-structure Ωε: fk = √ tk/ρ/(2π), k = 1, 2, . . . , 6. Here and in what follows ρ is associated with the 3D body. 3. A particular example: A 1D-3D multi-structure with six legs Here we study some particular configurations of a 1D-3D multi-structure. In ad- dition, we also consider different spacing between the neighbouring legs. To sim- plify the finite element modelling, the 3D body is chosen as a parallelepiped of dimensions a1, a2, a3. We assume that the legs are of length L and have b × b cross-sections. They are assumed to be homogeneous with density ρ, isotropic with Poisson’s ratio ν and Young’s modulus E, with the exception that the sixth leg is inhomogeneous, with Young’s modulus is given by: E(y(6)3 ) = {E1, 0 < y (6) 3 < L1, E2, L1 < y (6) 3 < L2, E3, L2 < y (6) 3 < L}. (1) We introduce a dimensionless normalised parameter ε = b/L (the characteristic ratio of the cross-section to the length of a leg) and use the following coordi- nates of the junction points: a(1) = (a1/2, 0,−a3/2 + p3), a(2) = (a1/2, 0,−a3/2 − VIBRATIONS OF DAMAGED 1D-3D MULTI-STRUCTURES 47 q3), a(3) = (p1, a2/2,−a3/2), a(4) = (−q1, a2/2,−a3/2), a(5) = (0, p2, 0), a(6) = (0,−q2, 0), where 0 < pk, qk < ak/2, k = 1, 2, 3. Using the definitions of the stiffness coefficients C( j)i , i = 1, 2, 3, 4; j = 1, 2, . . . , 6, and the fact that C(6)3 (see, for example, Kozlov et al., 1999) is much larger than the other stiffness coefficients, we define τ = 1 ε2EL { ∫ L 0 dz C(6)3 (z) }−1. Taking into account (1), to be able to model different types of damage of the leg, we consider the following cases: E2 = E3 = E, E1 = E∗; E1 = E2 = E, E3 = E∗; E1 = E3 = E, E2 = E∗, (2) where E∗ < E. Hence, we can write τ = [1 + ∆LL ( E E∗ − 1)]−1, where ∆L denotes the length of the damaged part of the rod. The first case in (2) corresponds to the configuration when the bottom part of the rod is damaged, the second to the case when the top part of the rod is damaged and the third to the case when the middle part of the rod is damaged. It is clear that in all cases the length of the damaged part plays an important role. We note that ε is a small parameter connected with the geometry of the 1D-3D multi-structure, whereas τ is a parameter indicating the strength of the damage. It is clear that 0 ≤ τ ≤ 1, where τ = 1 means no damage and τ = 0 means that the weak part of the sixth rod is completely damaged. In fact, the latter one corresponds to a 1D-3D multi-structure with five legs only. 3.1. ASYMPTOTIC FORMULAE Similar to the derivation in Aslanyan et al., 2002, 2003 and Aslanyan et al., 2005 analytic asymptotic formulae can be derived for certain six eigenfrequencies of the multi-structure. We note that four of these formulae are identical to the ones derived for a homogeneous multi-structure of the same geometry: f1,2 ∼ ε 2π √ 2EL ρΩ0 , f3 ∼ ε 2π √ 24ELp21 ρΩ0(a21 + a 2 2) , f4 ∼ ε 2π √ 24ELp23 ρΩ0(a21 + a 2 3) , f5,6 ∼ ε 2π √ ELz5,6(τ) ρΩ0 , (3) ε → 0, where Ω0 = a1a2a3. In the above formulae, z5,6 are the roots of the equation d2z2 − (1 + τ)d1z + d0τ = 0, where d2 = 112 (a 2 2 + a 2 3), d1 = p 2 2 + d2, d0 = 4p22. Note that asymptotic formulae (3) are valid for all the cases described in (2). We would also like to emphasise that the formulae (3) can be justified when τ ≥ τ0 > 0, where τ0 is some small constant. In the next section, we discuss the range of applicability of these formulae. 3.2. NUMERICAL RESULTS AND DISCUSSION For the numerical calculations, we consider the following fixed values of the parameters a1 = a2 = 108 cm, a3 = 60 cm, b = 4 cm , L = 200 cm, ∆L = 10 cm, E = 2.1 · 1012 g/(cm sec2), ρ = 7.8 g/cm3, ν = 0.3, and p1 = p2 = q1 = q2 = (2 + 2p) cm, p3 = q3 = (2 + p) cm. Referring to Aslanyan et al., 2005, we note that the skeleton approximation and 3D finite element computations 48 ASLANYAN, DRONKA, MISHURIS, SELSİL 0 0.05 0.1 0.15 0.2 0.25 0 5 10 15 20 25 30 35 40 τ f [Hz] f t (3) f r (1) Figure 1. Solid lines denote the eigenfrequencies f j, j = 1, . . . , 6 versus the parameter τ for the case when p = 11, obtained by the skeleton approximation, dotted lines denote the analytical as- ymptotic solution given in (3). The first eight eigenfrequencies obtained by direct 3D finite element calculations are denoted as follows: Star (*), right-sided triangle (�) and left-sided triangle (�) corresponds to damage localised in/at the middle, top and bottom of the leg, respectively. differ in the region where the neighbouring legs are close to the ends of the 3D body, i.e. the parameter p is large, whereas the difference between the skeleton approximation and its asymptotics is negligible. However, when the neighbouring legs are close to each other (p is small), the skeleton approximation and 3D finite element computations almost perfectly match, but the skeleton approximation and its asymptotics differ considerably. This is due to the existence of a second small parameter (the parameter p) appearing in addition to ε, which needs to be taken into account while constructing the asymptotics. Hence, in this short study, we omit the numerical calculations associated with the small values of the parameter p. In Fig. 1, we present numerical computations for the case when p = 11. Here, solid lines denote the eigenfrequencies f j, j = 1, . . . , 6 obtained by the skeleton approximation versus the parameter τ, and dotted lines denote the analytical as- ymptotic solution given in (3). We show these solid curves only for the damage localised on the top of the leg. This is because these approximations are practically indistinguishable for the damage localised in other parts of the leg, namely in the VIBRATIONS OF DAMAGED 1D-3D MULTI-STRUCTURES 49 middle and at the bottom. We underline that two of the eigenfrequencies obtained by using the skeleton approximation are very close to each other and their as- ymptotics are identical (see (3)). Hence, there are only five solid and five dotted curves visible in Fig. 1. It is clear from Fig. 3 that the skeleton approximation and its asymptotics are sufficiently close to each other. This is violated for f (1)r , the eigenfrequency corresponding to the rotation of the 3D body about the x1−axis. The only other eigenfrequency affected by the damage appears to be f (3)t , the eigenfrequency corresponding to the translation of the 3D body along the x3−axis. In addition to these, we denote the eigenfrequencies obtained by 3D finite el- ement calculations for three different values of the damage parameter τ, i.e. τ = 0.01, 0.05, 0.25, by stars (*), right (�) and left (�) sided triangles. They correspond to damage localised in/at the middle, top and bottom of the leg, respectively. The shift between the symbols on the figure along the horizontal axis is only for pre- sentational purposes. Here, the first eight eigenfrequencies are included since all these are within the range of six eigenfrequencies obtained by using the skeleton approximation and its asymptotics. Clearly, there is almost no difference between the cases when the damage is lo- calised at the top or at the bottom of the leg. In fact, for six eigenfrequencies associated with rigid body motion of the 3D body, the computations indicate that the values of the eigenfrequencies associated with the damage localised in the middle of the leg are also close to the eigenfrequencies obtained for the other two damage configurations discussed above. At this point, we would like to clarify that the remaining two eigenfrequencies plotted in Fig. 1 (indistinguishable on the figure) are not due to rigid body translations and rotations of the 3D body but due to bending type vibrations of the damaged leg. The ends of this leg can be considered as fixed for these vibration modes, since the 3D body remains motion- less. Therefore, it is also possible to find these eigenfrequencies analytically. It is evident from a physical point of view that these two eigenfrequencies computed for damage localised at the top or bottom will differ significantly with eigenfre- quencies computed for the damage localised in the middle, when the damage parameter τ is small. This is clearly shown in the figure by a double arrow for τ = 0.01. We present numerical computations for the case when p = 19 in Fig. 2. We use the same style of curves and same symbols for finite element computations as in Fig. 1. As expected, the skeleton approximation and its asymptotics are closer to each other, compared with the case when p = 11. On the other hand, the finite element computations for certain eigenfrequencies differ more (see ellipses on the figure). The other effects are as discussed in the case when p = 11. 50 ASLANYAN, DRONKA, MISHURIS, SELSİL 0 0.05 0.1 0.15 0.2 0.25 0 5 10 15 20 25 30 35 40 45 τ f [Hz] Figure 2. Solid lines denote the eigenfrequencies f j, j = 1, . . . , 6 versus the parameter τ for the case when p = 19, obtained by the skeleton approximation, dotted lines denote the analytical as- ymptotic solution given in (3). The first eight eigenfrequencies obtained by direct 3D finite element calculations are denoted as follows: Star (*), right-sided triangle (�) and left-sided triangle (�) corresponds to damage localised in/at the middle, top and bottom of the leg, respectively. References Aslanyan, A. G., Movchan, A. B., Selsil, O. (2002) Vibrations of multi-structures including elastic shells: Asymptotic analysis. Euro. Jnl of Applied Mathematics 11, 109-128. Aslanyan, A. G., Movchan, A. B., Selsil, O. (2003) Estimates for the low eigenfrequencies of a multi-structure including an elastic shell. Euro. Jnl of Applied Mathematics14, 313-342. Aslanyan, A.G., Movchan, A.B., Selsil, O. (2003) Asymptotics for the first six eigenfrequencies of a 1D-3D multi-structure. Proceedings of IUTAM international symposium on Asymptotics, Singularities and Homogenisation in Problems of Mechanics, Liverpool, UK, 221-228. Aslanyan, A.G., Dronka, J., Mishuris, G.S., Selsil, O. (2005) Vibrations of a weakly nondegenerate 1D-3D multi-structure. Submitted for publication to Archives of Mechanics. Aslanyan, A.G., Dronka, J., Mishuris, G.S., Selsil, O. (2005) The effect of structural damage on the vibrations of 1D-3D multi-structures. Proceedings of IIIrd Syposium on Damage in Materials and Structures, Augustow, Poland, 11-14. Kozlov, V.A., Maz’ya, V.G., Movchan, A.B. (1999) Asymptotic analysis of fields in multi- structures, Oxford University Press. 51 SIZING OF A SPHERICAL SHELL AS VARIABLE THICKNESS UNDER DYNAMIC LOADS Nurcan Aşçı1, Habib Uysal2 and Ümit Uzman1 1 Gümüşhane Engineering Faculty, KT , 29000, Gümüşhane, Turkey 2 Engineering Faculty, Atat rk University, 25240, Erzurum, Turkey Abstract. In this paper, optimization of a spherical shell under various dynamic loads is investi- gated. The aim of this optimization problem is to minimize the volume of the shell. Design variables are corner thicknesses of each finite element. Constraints are stresses obtained from von Mises stress criterion not to exceed the yield stress in corner nodal points of each finite element at the top and bottom surfaces of shell and thicknesses are restricted not to be less than 2.5 mm. In addition to shell’s own weight, the vertical loads with equal intensity are applied at the nodal points on the upper edge of spherical shell, varying with respect to time function P(t). Time varying load vector is considered three different cases such as step, step after ramp and impulse functions.A program is coded with MapleV for optimization of spherical shell and finite element package program ANSYS is used for structural analysis. Obtained results are presented in graphical and tabular form. Finally, concluded remarks are given. Key words: size optimization, design variables, constraints, dynamic loads 1. Introduction Shell structures are used engineering areas such as water towers, tanks and other liquid containing structures, domes, silos and pressure vessels where optimiza- tion procedure have been successfully used. The first research on optimization of structure was done by Zienkiewicz and Campell (1973). They have discussed the problem of finding the optimum shape of two-dimensional structure known as shape optimization. Since then several researches have been contributed in this area (Haftka and Prasad, 1979), (Ding, 1986), (Gates and Accorsi, 193), (Soares et al., 1994), (Pourazady and Fu, 1996) .The growing interest on this subject reflects the importance of the effect of structural shape on structural performance. Vibration Problems ICOVP 2005, c© 2007 Springer. .– 51 56 Ü ü (eds.), . şE. Inan and A. Kırı 52 AŞÇI, UYSAL, UZMAN 2. Optimization 2.1. MATHEMATICAL STATEMENT OF AN OPTIMIZATION PROBLEM Optimization problem is generally defined as the determination of the design vari- ables to minimize F objective function under gi constraints. This problem is stated mathematically as: Find → minF(S ) Constraints → gi(S ) ≥ 0 i=1,2,...,nv S Lk ≤ S k ≤ S U k k=1,2,...,n (1) In here F is the objective function, gi; i=1,2,...,nv are the constraint functions,nv is the number of constraints, S Lk and S U k are the lower and the upper limits of the design variable S k, respectively and n is the number of design variables. 2.2. PROBLEM DEFINITION A spherical shell of radius 10 m and uniform thickness of 25 mm with an apex cut of radius 2.5 m is considered (Asci 2004). Spherical shell made of steel material is used for analysis. It was assumed that the material is isotropic, linear elastic and Mindlin-Reissner shell theory is considered. The Modulus of Elasticity, Poisson ratio, density are taken 2×1011N/m2, 0.3 and 7860 kg/m3 , respectively. The finite element mesh is generated using SHELL93 element. Geometry, loading condition and SHELL93 element are given in Fig. 1. r, , ) y z x M J N O L K 6 2 5 1 P 8 4 3 7 SHELL93 Figure 1. Geometry of spherical shell, loading condition and SHELL93 element The region (radius 10 m, 0 ≤ θ ≤ 900 and 0 ≤ ϕ ≤ 75.50) consisting of intersection of xy, yz and zx planes and a plane with 14.50 angle to z axis is considered. Only, it is supported along lower edge points of the shell at the xy plane as fixed. SIZING OF A SPHERICAL SHELL 53 2.3. LOADING CONDITION In this study, in addition to shell’s own weight, the vertical loads with equal inten- sity are applied at the nodal points on the upper edge of spherical shell, varying with respect to time function P(t) . Time varying load vector is considered three different cases such as step, step after ramp and impulse functions (Fig. 2). Step Loading Step After Ramp Loading Impulse Loading fttP t tP 0 00 )( fv v v tttP tt t t P t tP 00 )( fv v ttt ttP t tP 0 0 00 )( 0.00 0.50 1.00 Time (s) P (t ) P 0.00 0.50 1.00 Time (s) P ( t) P 0.00 0.50 1.00 Time (s) P (t ) P Figure 2. The varying P(t) load vector for three load cases Vertical loads with equal intensity re applied at the nodal points on the upper edge of spherical shell. For P(t) function, concentrated loads are taken to be 2 × 103N for step and step after ramp loadings and 10 × 103N for impulse loading. 2.4. OBJECTIVE FUNCTION, DESIGN VARIABLES AND CONSTRAINT FUNCTIONS In this study, the objective function is to minimize weight (minimizing volume) of spherical shell. Totally 81 design variables which are thicknesses of corner nodal points of each finite element are considered independently from each other (Fig. 3). S4S1S2 S7 S8S6 S9 S3 S5 S16 S22 S25 S26 S24 S44 S69 S71 S72 S64 S65 S67 S66 S68 S70 S51 S61 S62 S58 S60 S55 S59 S57S56 S79 S81 S80 S75S74S73 S77 S78 S76 S31 S29 S34 S46 S47 S48 S50 S52 S49 S53 S63 S35 S33 S28 S30 S32 S36 S41 S42 S37 S40 S38 S39 S43 S45 S14 S17 S10S11 S12S13 S15 S19 S23 S21S20 S27 S18 S54 Figure 3. Considered design variables for spherical shell The constraint functions are stresses (σe) obtained from von Mises stress cri- terion not to exceed the yield stress(σe ≤ σ0 = 220 MPa) in corner nodal points 54 AŞÇI, UYSAL, UZMAN of each finite element at the top and bottom surfaces of shell and the thicknesses are restricted not to be less than 2.5 mm. σe = [σ 2 x + σ 2 y + σ 2 z − σxσy − σxσz − σyσz + 3(τ2xy + τ2xz + τ2yz)](1/2) (2) In the optimization process consist of such as description of geometric model, dynamic analyze, mesh generation, sensitivity analysis and mathematical pro- gramming method, respectively. The above steps are repeated until convergence is achieved using iterative convergence strategies. 3. Numerical examples At the beginning of the optimization process design variables are chosen as 0.025 m. In this case, spherical shell’s volume is obtained 3.802m3. In Fig. 4, displace- ment time curves are illustrated for three different load functions. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Time (s)-0.50 0.00 0.50 1.00 1.50 2.00 D is p la ce m e n t (m ) Step Step After Ramp Impulse (Uz)max(Uz)max (Uz)max Figure 4. Displacement time curves for three load cases at the first iteration According to Fig. 4, maximum displacements are occurred 1.161 m at 0.525 s for step loading, 1.129 m at 0.62 s for step after ramp loading and 0.931 m at 0.37 s for impulse loading at 1st and 9th design variables at the first iteration. In Fig. 5, displacement time curves for step loading are given. 0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.54 0.56 0.58 0.60 Time (s) 0.60 0.70 0.80 0.90 1.00 1.10 1.20 D is p la ce m e n t (m ) I III IV II Figure 5. Displacement-time curves for several sequential optimization processes for step loading It can be seen from Fig. 5 the time values for maximum displacement and opti- mum volumes approach each other after several sequential optimization processes. For step loading, optimization time is considered as 0.525ths. Also, time is ob- tained 0.44 s, 0.47 s, 0.465 s and maximum displacements are obtained 0.875 m, SIZING OF A SPHERICAL SHELL 55 0.795 m, and 0.838 m for II, III and IV curves, respectively. For step after ramp loading, optimization time is considered as 0.62nds . In Fig. 6 time values for maximum displacements are given for step after ramp loading. 0.50 0.52 0.54 0.56 0.58 0.60 0.62 0.64 0.66 0.68 0.70 Time (s) 0.60 0.70 0.80 0.90 1.00 1.10 1.20 D is p la ce m e n t (m ) I III II Figure 6. Displacement time curves for several sequential optimizations for step after ramp loading In Table 1, optimum volumes and design variables are given for three load cases at the end of the optimization process. TABLE I. Obtained optimum volumes and design variables for three load cases at the end of the optimization process Design Iteration Number Variables S tep L. S tep A f ter Ramp L. Impulse L. (m) 30 45 Opt. 30 45 Opt. 30 45 Opt. S 3, S 7 .0467 .0468 .0468 .0444 .0450 .0450 .0779 .0767 .0774 S 4, S 6 .0481 .0478 .0478 .0467 .0469 .0469 .0789 .0749 .0768 S 5 .0518 .0514 .0513 .0503 .0505 .0505 .0802 .0785 .0806 S 11, S 17 .0597 .0592 .0592 .0584 .0581 .0580 .0884 .0858 .0854 S 20, S 26 .0711 .0699 .0698 .0712 .0704 .0704 .1125 .1039 .1040 S 28, S 36 .0636 .0642 .0642 .0646 .0648 .0649 .0519 .0560 .0565 S 37, S 45 .0664 .0663 .0663 .0686 .0685 .0685 .0614 .0696 .0681 S 46, S 54 .0577 .0575 .0575 .0606 .0605 .0605 .0369 .0478 .0471 S 64, S 72 .0745 .0743 .0743 .0739 .0736 .0736 .0438 .0418 .0416 S 74, S 80 .0529 .0569 .0570 .0352 .0386 .0388 .0135 .0088 .0084 Volume(m3) 1.788 1.781 1.781 1.790 1.787 1.786 1.983 2.004 2.007 Because of loading condition and geometry are symmetric, design thicknesses are also obtained as symmetric. The objective function became stable at the se- quential optimization steps and finally changes in the objective function approach zero. In Fig. 7, time values for maximum displacements are given for impulse loading and optimization time is considered as 0.37ths for this loading. 56 AŞÇI, UYSAL, UZMAN 0.30 0.32 0.34 0.36 0.38 0.40 0.42 0.44 Time (s) 0.60 0.70 0.80 0.90 1.00 D is p la ce m e n t (m ) I IIIII Figure 7. Displacement-time curves for several sequential optimization processes for impulse loading 4. Conclusions In this study, optimization of spherical shell is investigated under dynamic load- ing. The stresses are evaluated at the top and bottom surfaces of the spherical shell at each optimization process. Then, von Mises criterion is compared with yield stress in corner nodal points of finite elements. In some points, the thicknesses are increased in case criterion is higher than the yield stress; otherwise they are decreased in these points by program. The changes in the thicknesses of shell affect the rigidity of shell, the natural frequency of shell and the time values for maximum displacement. In other points, the minimum thickness criterion is active. At the end of the optimization process, it is concluded that the stress cri- terion is active in the near points of the shell edges, and the thickness criterion is active in the far points from the shell edges. So, the optimum shell is resulted as a shell strengthened at the edges. Hence, it can be considered as a spherical shell supported by beams at the edges. References Asci, N. (2004) Optimization of a Spherical Shell with Variable Thickness Under the Various Dy- namic Loads, Ph.D. Thesis, Department of Civil Engineering, Karadeniz Technical University, Turkey. Ding Y. (1986) Shape Optimization of Structures: A Literature Survey, Computers and Structures 24 985-1004. Gates, A. A., Accorsi, M. L. (1993) Automatic Shape Optimization of Three Dimensional Shell Structures with Large Shape Changes, Computers and Structures 49 167-178. Haftka, R. T., Prasad, B. (1979) Programs for Analysis and Resizing of Complex Structures, Computers Structures 10 323-330. Mota Soares, C. M., Mota Soares, C. A., Barbosa, J. I. (1994) Sensitivity Analysis and Optimal Design of Thin Shells of Revolution, AIAA Journal 32 1034-1042. Pourazady, M. Fu, Z. (1996) An integrated Approach to Structural Shape Optimization, Computers and Structures 60 279-289. Zienkiewicz, O. C. and Campbell, J. S. (1973) Shape Optimization and Sequential Linear Programming. In Optimum Structural Design, John Wiley, New York. 57 VIBRATION ANALYSIS OF SIMPLY SUPPORTED FUNCTIONALLY GRADED BEAMS Metin Aydoğdu and Vedat Taşkın Department of Mechanical Engineering, Trakya University, 22030, Edirne,Turkey Abstract. In this study, free vibration of simply supported functionally graded (FG) beam was investigated. Young modulus of beam is varying in the thickness direction according to power law and exponential law. Governing equations are found by applying Hamilton’s principle. Navier type solution method was used to obtain frequencies. Different higher order shear deformation theories and classical beam theories were used in the analysis. Results were given for different material properties and different slenderness ratios. Key words: functionally graded beam, vibration, shear deformation theories 1. Introduction Use of structures like beams, plates and shells, which are made from FG materials, is increasing because of smooth variation of material properties. Compared with FG plates and shells, studies for FG beams are relatively less (Sankar, 2001; Vel and Batra, 2002). Different theories can be used to analyse structural members. For laminated composite plates, some higher order shear deformation theories (HSDT) are developed by Reddy, 1984; Soldatos, 1992 and Karama et al., 2003. Generalization of these theories is made by Soldatos and Timarci, 1993. This new theory called Unified Shear Deformation Theory (USDT) has been applied to vibration of composite cylindrical shells by Timarci and Soldatos, 1995. In the present study, free vibration of simply supported FG beams investigated by using Classical Beam Theory (CBT) and USDT. Firstly, governing equation of FG beam for free vibration problem are found by applying Hamilton principle and Navier type solution method is used to solve vibration problem. It is assumed that elasticity modulus is changing in the thickness direction and all other ma- terial properties are taken to be constant. Variation of elasticity modulus in the thickness direction is taken in two different forms like power product form and exponential form. Different shear deformation theories are used in USDT when finding frequencies. Frequency results are given for FG beams for different mate- also given for different material properties. Vibration Problems ICOVP 2005, c© 2007 Springer. .– 57 rial properties and different length to thickness ratios and some mode shapes are 62(eds.), . şE. Inan and A. Kırı 58 AYDOĞDU, TAŞKIN 2. Analysis A functionally graded beam of constant thickness of h with cross-section dimen- sions L and b is considered. It is assumed that the Young modulus of FG beam vary through the thickness of the beam according to exponential and power law form. The exponential law, which is generally used in fracture studies (Kim and Paulino, 2002) is given by E(z) = Ete (−δ(1−2z/h)) δ = 1 2 ln ( Et Eb ) (1) and the power law introduced by Wakashima et al., 1990 is given by E(z) = (Et − Eb) ( z h + 1 2 )k + Eb (2) Et and Eb denote values of the elasticity modulus at top and bottom of the beam, respectively, and k is a variable parameter. According to this distribution, bottom surface (z=-h/2) of functionally graded beam is pure metal, the top surface (z=h/2) is pure ceramics, and for different values of k one can obtain different volume fractions of metal. The state of stress in the beam is given by the generalised Hooke’s law as follows: σx = Q11εx, τxz = Q55γxz, (3) where Qi j are the transformed stiffness constants in the beam co-ordinate system and defined as: Q11 = E(z) 1 − v2 Q55 = E(z) 2(1 + v2) (4) Assuming that the deformations of the beam are in the x-z plane and denoting the displacement components along the x, y and z directions by U, V and W respectively, the following displacement field for the beam is assumed on the basis of the general shear deformable shell theory presented by Soldatos and Timarci, 1993: U(x, z; t) = u(x; t) − zw,x + Φ(z)u1(x; t), V(x, z; t) = 0, W(x, z; t) = w(x; t), (5) Here u and w represent middle surface displacement components along the x and z directions respectively, while u1is unknown function that represents the effect of transverse shear strain on the beam middle surface, and Φ represents VIBRATION ANALYSIS OF FUNCTIONALLY GRADED BEAMS 59 the shape function determining the distribution of the transverse shear strain and stress through the thickness. CBT is obtained as a particular case by taking the shape function as zero. Although different shape functions are applicable, only the ones which convert the present theory to the corresponding Parabolic Shear Deformation Beam Theory (PSDBT), First Order Shear Deformation Beam The- ory (FSDBT) and Exponential Shear Deformation Beam Theory (ESDBT) are employed in the present study. This is achieved by choosing the shape functions as follows: FS DBT : Φ(z) = z, PS DBT : Φ(z) = z(1 − 4z2/3h2), ES DBT : Φ(z) = z exp [ −2(z/h)2 ] (6) By substituting the stress-strain relations into the definitions of the force and moment resultants of the theory given in Soldatos and Timarci, 1993 and using the notation given by Soldatos and Sophocleous, 2001, the following constitutive equations are obtained: Ncx Mcx Max = Ac11 B c 11 B a 11 Bc11 D c 11 D a 11 Ba11 D a 11 D aa 11 u,x −w,xx u1,x (7) Qax = A a 55u1 (8) The extensional, coupling and bending rigidities appearing in Eq. (7) are, respectively, defined as follows: Ac11 = ∫ h/2 −h/2 Q11 dz, ( Bc11, B a 11 ) = ∫ h/2 −h/2 Q11 [ z, φ(z) ] dz, ( Dc11, D a 11, D aa 11 ) = ∫ h/2 −h/2 Q11 [ z2, zφ(z), φ2(z) ] dz, (9) Moreover, the transverse shear rigidity appearing in Eq.(8) is defined accord- ing to Aa55 = ∫ h/2 −h/2 Q55 [ φ′(z) ]2 dz. (10) It should be pointed out that the extensional Ac11, coupling B c 11 and bending Dc11 rigidities are the ones usually appearing even in the CBTs. Among the addi- tional rigidities in Eq.(7), the one denoted as Ba11 is considered as additional cou- pling rigidity while the ones denoted as Da11 and D aa 11 are considered as additional bending rigidities. 60 AYDOĞDU, TAŞKIN Upon employing the Hamilton’s principle, the three variationally consistent equilibrium equations of the beam are obtained as: Ncx,x = (ρ0u + ρ01u1 − ρ1w,x),tt, Mcx,xx = (ρ1u,x + ρ11u1,x + ρ0w − ρ2w,xx),tt, Max,x − Qax = (ρ01u + ρ02u1 − ρ11w,x),tt, (11) Here ,tt denotes time derivatives and the ρ’s are defined as: ρi = h 2∫ −h 2 ρzidz, (i = 0, 1, 2), ρ jm = h 2∫ −h 2 ρz jΦmj dz, ( j = 0, 1; m = 1, 2). (12) and the force and moment components: (Ncx, M c x) = h/2∫ −h/2 σx(1, z)dz, Q a x = h/2∫ −h/2 τxzφ ′(z)dz, Max = h/2∫ −h/2 σxφ(z)dz. (13) The simply supported boundary conditions are given as follows: Ncx = w = M c x = M a x = 0 (14) Navier-type solutions to Eq.(11) that satisfy the boundary conditions Eq.(14) can be expressed in the form of u = Am cos mπx L sinωt, L u1 = Bm cos mπx L sinωt,w = Cm sin mπx L sinωt (15) where m is the half wave number in the x direction and Am, Bm and Cm are un- determined constant coefficients. Inserting Eq.(15) in Eq.(11) leads to following eigen-value problem: (K − λ2M) {∆} = 0 (16) here K and M are stiffness matrix, inertia matrix, respectively and ∆ is column vector of unknown coefficients and λ is free vibration frequencies. 3. Results and discussions The constituent material properties of the FG beam were chosen as follows: Al: Em=70GPa, νm = 0.3, Ceramic: Ec=380GPa, νc = 0.3 VIBRATION ANALYSIS OF FUNCTIONALLY GRADED BEAMS 61 and constant density is assumed for FG beam. Frequency parameter is non - dimensionalized as λ = ωL2 h √ ρm Em . (17) For given material properties δ = 0.846 was found and different k values can be chosen. Non-dimensional frequency parameters were given for fundamental frequency and fifth frequency in Table 1 and Table 2 for L/h=5 respectively for different theories and for different material distributions. It is seen from tables that since increasing k leads to approach to metal behaviour, frequency parameter is decreasing with increasing k and increasing with increasing L/h ratios. Since deformations are increasing with increasing mode number, difference between CBT and shear deformation theories for fifth mode is greater than the difference for the first mode. TABLE I. Comparison of frequency parameter with different theories for different material distribution (L/h=5, fundamental mode). Theory k=0 k=0.1 k=1 Exp k=2 k=10 Metal PSDBT 6.5742 6.2483 4.6595 4.2667 4.1037 3.5482 2.8216 ESDBT 6.5849 6.2581 4.6659 4.2725 4.1091 3.5539 2.8262 FSDBT 6.5633 6.2371 4.6528 4.2639 4.1019 3.5639 2.8169 CBT 6.8470 6.4999 4.8216 4.4235 4.2517 3.7372 2.9387 TABLE II. Comparison of frequency parameter with different theories for different material distribution (L/h=5, fifth mode). Theory k=0 k=0.1 k=1 Exp k=2 k=10 Metal PSDBT 92.7810 88.593 67.088 60.5028 58.2308 46.2902 39.821 ESDBT 94.0910 89.810 67.916 61.2693 58.9629 46.9512 40.383 FSDBT 91.1638 87.019 65.946 59.6582 57.4233 46.7166 39.127 CBT 128.8669 122.12 86.409 77.9335 74.0286 65.9794 55.309 62 AYDOĞDU, TAŞKIN 4. Conclusions This study dealt with free vibration of simply supported FG beams. Young modu- lus of beam assumed varies in the thickness direction according to power law and exponential law. Governing equations are found by applying Hamilton’s principle. Navier type solution method was used to obtain frequencies. Different Higher Order Shear Deformation Theories and Classical Beam Theories were used in the analysis. It is found that CBT gives higher frequencies especially for higher modes. This study can be extended to study different boundary conditions and temperature effects can be included in the analyses. References Karama, M., Afaq, K. S., Mistou, S. (2003) Mechanical Behaviour of Laminated Composite Beam by the New Multi-Layered Laminated Composite Structures Model with Transverse Shear Stress Continuity, Int. J. of Solids and Struct 40 1525-1546. Kim, J., Paulino, G. H. (2002) Finite Element Evaluation of Mixed Mode Stress Intensity Factors in Functionally Graded Materials, International Journal for Numerical Methods in Engineering 53 1903-1935. Reddy J. N. (1984) A Simple Higher-Order Theory for Laminated Composite Plates, Journal of Applied Mechanics 51 745-752. Sankar B. V. (2001) An Elasticity Solution for Functionally Graded Beams, Composites Science and Technology 61 689-696. Soldatos K. P. (1992) A Transverse Shear Deformation Theory for Homogeneous Monoclinic Plates, Acta Mechanica 94 195-220. Soldatos, K. P., Timarci T. (1993) A Unified Formulation of Laminated Composite, Shear De- formable, Five Degrees of Freedom Cylindrical Shell Theories, Composite Structures 25 165-171. Soldatos, K. P., Sophocles C. (2001) On Shear Deformable Beam Theories: The Frequency and Nor- mal Mode Equations of the Homogeneous Orthotropic Bickford Beam, J. of S. and Vibration 242 215-245. Timarci T., Soldatos, K. P. (1995) Comparative Dynamic Studies for Symmetric Cross-Ply Circular Cylindrical Shells on the Basis of a Unified Shear Deformable Shell Theory, Journal of S. and Vibration 187 609-624. Vel S. S., Batra R. C. (2002) Exact Solution for the Cylindrical Bending Vibration of Function- ally Graded Plates, Proceedings of the American Society of Composites, Seventh Technical Conference, Purdue University West Lafayette, Indiana. Wakashima K., Hirano T., Niino M. (1990) Space Applications of Advanced Structral Materials, ESA, SP303:97. 63 DAMPING EFFECTS OF RUBBER LAYER IN LAMINATED COMPOSITE CIRCULAR PLATE DURING FORCED VIBRATION Aysun Baltacı, Mehmet Sarıkanat and Hasan Yıldız Mechanical Engineering Department, Ege University, İzmir, Turkey Abstract. In this study, the effects of the location of the rubber between composite layers which is located to supply damping on the vibration frequency of laminated composite circular plates having different fiber directions are investigated by using the finite element method. The laminated composite circular plate having free outer edge is manufatured as having a central hole. Hybrid composite circular plate is clamped at the inner edge and subjected to transverse impuls load on a location at the outer edge. The best location of the rubber layer in terms of the damping capability are determined by using finite element analysis. Key words: damping, forced vibration, composite circular plate 1. Introduction Damping is a usefull for controlling vibration and movement in the design of the structure. The plate with the constrained damping layer treatment is well known to have high damping capasity and high resistance to resonant noside and vibration. Also, numerous researchers have developed the vibration and damping properties of basic structures, such as beams, rectangular and circular plates with constrained damping layer treatment in laminated composites. Haddad and Feng (2003) de- veloped Quasi-static models by using a forced balance approach, to define the effects of selected microstructural parameters, on the damping and stiffness of a class of poliymeric, discontinuous fibre composite systems. Chen and Levy (1999) discussed the temperature effects on frequency, loss factor and control of a flexible beam with a constrained viscoelastic layer and shape memory alloy layer. In that study, they also discussed the effects of damping layer shear modulus and damp- ing layer height as affected by the temperature. The axisymmetric vibration and damping of rotating annular plates with constrained damping layer treatments are analyzed by Wang and Chen (2004). In the present work, the effects of the location vibration frequency of laminated composite circular plates having different fiber directions are investigated by using the finite element method. Vibration Problems ICOVP 2005, c© 2007 Springer. .– 63 of the rubber between composite layers which is located to supply damping on the 68(eds.), . şE. Inan and A. Kırı 64 BALTACI, SARIKANAT, YILDIZ 2. Transient analysis Transient dynamic analysis (sometimes called time-history analysis) is a tech- nique used to determine the dynamic response of a structure under the action of any general time-dependent loads. You can use this type of analysis to determine the time-varying displacements, strains, stresses, and forces in a structure as it responds to any combination of static, transient, and harmonic loads. The time scale of the loading is such that the inertia or damping effects are considered to be important. If the inertia and damping effects are not important, you might be able to use a static analysis instead. The basic equation of motion solved by a transient dynamic analysis is Mü + Cu̇ + Ku = F(t) (1) Where M = mass matrix, C = damping matrix, K = stiffness matrix, ü = nodal acceleration vector, u̇ = nodal velocity vector, u = nodal displacement vector, F(t) = load vector. At any given time, t, these equations can be thought of as a set of ”static“ equilibrium equations that also take into account inertia forces (Mü) and damping forces (Cu̇). 3. Results and discussion Figure 1-3 show the effects of the position of the rubber on the time response of the composite circular plate. As seen from figures, when the rubber is located at the middle layer, the shortest damping time is obtained. Figure 4 shows the effect of the position of the rubber on the maximum amplitude of vibration of the composite circular plate having [0/0]2s fiber orientation for different rubber thickness. Figure 5 shows the effect of the position of the rubber on the maximum amplitude of vibration of the composite circular plate for different fiber orientations. Figure 6 and 7 show the effect of the rubber thickness on the maximum amplitude and damping time of composite circular plate, respectively. As seen from figures, the maximum amplitude and damping time desrease as the thickness of the rubber increases. 4. Conclusions The maximum amplitude of the vibration decreases as the thickness of the rubber increases. The damping time decreases as the thickness of the rubber increases. There is no effect of the fiber orientation on the damping time. The maximum amplitude of the vibration decreases as the location of the rubber gets closer to the middle axis of the laminated composite. Amplitude response is symmetric DAMPING EFFECTS OF RUBBER LAYER 65 Figure 1. The effects of the position of the rubber on the time response of the composite circular plate (rubber is pleased top of the composite plate) Figure 2. The effects of the position of the rubber on the time response of the composite circular plate (rubber is pleased middle of the composite plate) 66 BALTACI, SARIKANAT, YILDIZ Figure 3. The effects of the position of the rubber on the time response of the composite circular plate (rubber is pleased botton of the composite plate) Figure 4. The effect of the position of the rubber on the maximum amplitude of vibration of the composite circular plate having [0/0]2s fiber orientation for different rubber thickness DAMPING EFFECTS OF RUBBER LAYER 67 Figure 5. The effect of the position of the rubber on the maximum amplitude of vibration of the composite circular plate for different fiber orientations Figure 6. The effect of the rubber thickness on the maximum amplitude of composite circular plate 68 BALTACI, SARIKANAT, YILDIZ Figure 7. The effect of the rubber thickness on the damping time of composite circular plate with respect to the middle axis. When the rubber is located as the top or the bottom layer of the composite structure: the structure produces different vibration pattern in response to impulse input, damping time reaches to a maximum value. Difference between damping times decreases when the rubber is located at the other layers. The shortest damping time is obtained when the rubber is located at the middle layer. References Chen Q. and Levy C. (1999) Vibration analysis and control of flexible beam by using smart damping structures, Composites: Part B 30, 395-406. Haddad Y. M. and Feng J. (2003) On the trade-off between damping and stiffness in the design of discontinuous fbre-reinforced composites, Composites: Part B 34, 11-20. Wang, H. J. and Chen L. W. (2002) Vibration and damping analysis of a three-layered composite annular plate with a viscoelastic mid-layer, Composite Structures 58, 563-570. Wang H. J. and Chen L. W. (2004) Axisymmetric vibration and damping analysis of rotating annular plates with constrained damping layer treatments, Journal of Sound and Vibration 271, 25-45. 69 69–78. A CRITICAL STUDY ON THE APPLICATION OF CONSTANT DEFLECTION CONTOUR METHOD TO NONLINEAR VIBRATION OF PLATES OF ARBITRARY SHAPES M. M. Banerjee Retired Reader, Department of Mathematics, A.C.College, Jalpaiguri-735101, W. B., India [202 Nandan Apartment, Hillview (N), S B Gorai Road, Asansol-713 304, W. B., India] Abstract. The present work attempts to utilize the Constant Deflection Contour Method in the investigation of nonlinear vibration of thin elastic plates. An attempt has also been made to use con- formal mapping technique for plates having uncommon or complex boundary. The usual boundary conditions have also been transformed accordingly. The equations for some iso-deflection curves of practical interest have been presented here for further investigations in conjunction with the present theory. Key words: constant deflection method, conformal mapping, vibration of plates 1. Introduction (Leissa, 1969) in his book provides a detailed study of linear problems and nu- merous references of significant contributions, within the linear framework, may be found. The paucity of literature concerning nonlinear (large amplitude) vibration analy- sis is, probably, due to the fact that the two basic Von Kármán field equations extended to the dynamic case, involved the deflection and the stress functions in a coupled form. For solution of plate problems having uncommon boundaries is a difficult task and a more complicated one when geometrical nonlinearities are involved. (Nowinski and Ohanabe, 1972) made some reservations on the free hand use of this method. Recently a new idea has been put forward by (Banerjee, 1997) to study the dynamic response of structures of arbitrary shapes based on “Constant Deflection Contour” (CDC) method. (Mazumdar and others, 1975, 1978, 1979, 1997) have previously developed the method. However, the application of this method has been restricted to linear cases only. The present paper aims at extending this method to the nonlinear analysis of plates of arbitrary shapes vibrating at large amplitude. The analysis carried Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 70 BANERJEE out in this paper may readily be applied to other geometrical structures and as a byproduct the static deflection is also obtainable. A combination of the constant deflection contour method and the Galerkin procedure is employed. 2. Some preliminary remarks on the CDC method When the plate vibrates in a normal mode then at any instant tθ , the intersections between the deflected surface and the parallels z = constant, will yield contours which after projection onto z = 0 surface are a set of level curves, u(x, y) = constant, called the “Lines of Equal Deflections” (Banerjee, 1976), which are iso- amplitude contours. The boundary of the plate, irrespective of any combination of support, is also a simple curve Cu belonging to the family of lines of equal deflections As defined by (Mazumdar, 1970) the family of nonintersecting curves may be denoted by Cu, for Cu, 0 ≤ u ≤ u�, so that Cu(u = 0) is the boundary and Cu coincides with the point(s) at which the maximum u = u� is obtained. 2.1. GOVERNING EQUATIONS - A NEW APPROACH Consider a thin elastic plate, which vibrates with moderately large amplitude in the transverse direction, under the action of a uniform load q. The usual procedure is to consider Karman type field equations extended to the dynamic case are D ∇4 w = h S (φ, w) + q − ρ h wtt (1) ∇4 φ = −(E/2) S (w, w), (2) with usual notations and a suffix is taken as an indication of partial differentiation with respect to the implied variable and the operator S is defined by S (I, J) = Ixx Jyy − 2 Ixy Jxy + Iyy Jxx. As an alternative to deriving subsequent equations followed earlier (Banerjee and Rogerson, 2002), has already been treated where the deflection function and the stress function in the separable form (Yamaki, 1961) w(x, y, t)=h W(x, y) F(t), φ(x, y, t)=h Φ(x, y) F2 (t) (3) where F(t) is an unknown function of time to be determined. Let us now make the following transformation ∂w ∂x =wx= dw du ux,wxx= d2w du2 u2x+ dw du uxx,wy= d w d u uy,wxy= d2 w du2 uxuy+ dw du uxy etc. (4) CONSTANT DEFLECTION CONTOUR METHOD 71 With transformations exemplified by those shown above, Eqns.(1) and (2) are transformed to D 4∑ i=1 λi d5−iW du5−i hF(t) = h3 [ λ5 d2W du2 dΦ du + λ6 d2Φ du2 dW du + λ7 dW du dΦ du ] F3(t) + q − ρh2WF̈, F̈ = d2F dt2 (5) 4∑ i=1 λi d5−iW du5−i = Eh [ λ8 d2W du2 dW du + λ9 ( dW du )] (6) in which λi (i = 1 − 9) are functions of partial derivatives of u only. Eqns. (5) and (6) can then be transformed into a set of equations (Banerjee and Rogerson, 2002) as hF(t) 4∑ i=1 f1i d4−iW du4−i + h3 2D f15 dW du dΦ du F3(t)+ f16 D q+ ρh2 D f17F̈(t) u∫ u� Wdu0 = 0, (9) and Λ2 [ Φ, dW du , dΦ du , . . . ] = 0 or, 4∑ i=1 g1i d4−iΦ du4−i + Eh 2 g15 ( dW du )2 = 0 (10) where f1i and g1i are functions of u and u� only. To avoid the integral appearing in Eqn.(9), (Jones and Mazumdar, 1997) have taken the derivative of it making the equation of order four again, viz., hF(t) 5∑ i=1 f2i d5−iW du5−i + h3 f25 2D d du ( dW du dΦ du ) F3(t) + f26 D q + ρh2 D f27WF̈(t) = 0 (11) Eqns. (9) and (10) or Eqns. (10) and (11) may be utilized to form the basic equations governing the motion of the structure. 2.2. SUPPORTING CONDITIONS Different types of boundary conditions have also been considered in our previous paper (Banerjee and Rogerson, 2002). Their transformations to u−variables have already been established therein and hence for brevity the readers may be referred to (Chakraborty et al., 1997). 3. The line of equal deflection -u(x, y) It is very important to know the equation to the line of equal deflection but it is not always possible to know the exact eqution of the line of equal deflection. 72 BANERJEE 3.1. SOME KNOWN EQUATIONS FOR LINE OF CONTOUR LINES Shapes of structures u(x, y) (a) Elliptic plates u(x, y) = 1 − (x/a)2 − (y/b)2 (b) Circular plates u(x, y) = a2 − x2 − y2 (a, b) length of the semi axes Annular plates a, b are the outher and inner radii u = a2 − b2 on the inner boundary u = 0 on the outher boundary Equilateral Triangular plates (a) Clamped : u(x, y) = 12a ( x3 − 3xy2 − ax2 − ay2 + 427 a 3 ) (b) Simply-supported : u(x, y) = ( x3 − 3xy2 − ax2 − ay2 + 427 a 3 ) × ( 4 9 a 2 − x2 − y2 ) Clamped right-angled isosceles triangular plate u(x, y) = a2 (x + y) − xy − 1 2 (x 2 + y2) 4 a 2 π2 ∞∑ n=0 (−1)n{sinh(Kny) cos(Kny)+sinh(Kxy) cos(Kyy)} (2n+1)3 sinh(Kna/2) Doubly-curved shallow shells Elliptical in plan ; a, b are the semi-major and semi-minor axes of the elliptic base u(x, y) = 1 − (x/a)2 − (y/b)2 Shallow spherical shell on an equilateral triangular base u(x, y) = ( x3 − 3xy2 − ax2 − ay2 + 427 a 3 ) × ( 4 9 a 2 − x2 − y2 ) Shallow spherical shell upon a square base a = the dimension of the base u(x, y) = (x2 − a2/4)(y2 − a2/4) 4. Application of conformal mapping technique (Pombo et al., 1977) presented a very convenient tool to conformably transform the given domain into a simpler one, i.e., onto a unit circle for such complicated boundaries. Yet their problems are confined to linear in character only. The present author has made some honest attempt to extend this technique to plates exhibiting large deflections (Banerjee, 1976) and vibrating at large amplitudes (Laura and Chi, 1965). Introducing the complex coordinates z = x + iy, z = x − iy in conjunction with the concept of the present CDC-Method the basic governing equations (1-2) will take (on transformation to u−variables as performed earlier) further reduce to 16D 4∑ i=1 li d5−iw du5−i = −4h [ l5 d du ( dw du dφ du ) + 2l6 dw du dφ du ] + q − ρhwtt (12) CONSTANT DEFLECTION CONTOUR METHOD 73 16 4∑ i=1 li d5−iφ du5−i hF(t) = 4E [ l5 d2w du2 dw du + l6( dw du )2 ] (13) where l1 = u2z u 2 z̄ , l2 = (u 2 z uz̄z̄ + 4uzuz̄uzz̄ + uzzu 2 z̄ ), l3 = (2u2zz̄ + 2uzuzz̄z̄ + 2uz̄uzzz̄ + uzzuz̄z̄), l4 = uzzz̄z̄, l5 = (u2z uzz − 2uzuz̄uzz̄ + uz̄z̄u2z ), l6 = (uzzuz̄z̄ − u2zz̄). (14) Now, if there be a functional relation z = f (ς) which maps the given shape of the plate in the z− plane into a unit circle in the ς− plane, one obtains equations (12-13) more or less in the same form. The only difference is that the li’s are substantially changed to, say l/i dependent on ς, ς̄, namely, 16D 4∑ i=1 l′i d5−iw du5−i = −4h [ l′5 d du ( dw du dφ du ) + 2l′6 dw du dφ du ] + q − ρhwtt (15) 16 4∑ i=1 l′i d5−iφ du5−i hF(t) = 4E [ l′5 d2w du2 dw du + l′6( dw du )2 ] (16) where l/i are functions of derivatives of ς, ς̄ and and z = f (ς), z̄ = f (ς̄), z z̄ and ς(= reiθ), ς̄ are complex conjugate in their complex planes, respectively. 5. Method of solution a) For the real plane Considering (9) and (10) as the basic equations with appropriate boundary conditions, it starts with finding the exact or approximate solution for Φ from Eqn. (10). For nonlinear analysis one may have to seek an approximate solution for which the form of the deflected function must be first assumed compatible with the boundary conditions. Next we solve for Φ from equation (10) in conjunction with a Galerkin procedure. With this expression for Φ and previously assumed form of W yield an ordinary time differential equation when the Galerkin procedure is applied again. Let u(x, y) = u be the representative of one of the family of the iso-deflection curves, then for any prescribed boundary conditions the deflection function w(u, t) can be assumed to take the form W = h n∑ i=1 AiΓ(u)F(t) (17) and φ = hΦ(u)F2(t) (18) 74 BANERJEE Equation (10), in combination with (15) and (16), will yield Φ = Φ(u). Substi- tuting this value of Φ with W as in (17), Eqn. (7) or Eqn.(9) will yield the error function ε = Λ1 [ u, F(t), F̈(t) ] (19) Since equation (17) is not an exact expression for W, rather an approximation, the associated error function may be minimized using Galerkin method. The appro- priate orthogonality condition applied to Eqn. (19) will yield the following “Time Differential Equation” with known constants C1 and C2, e.g., F̈(t) + C1F(t) + C2F 3(t) = C3q (20) the solution of which is well known from which the required investigations can be carried out. A detailed outline and illustrations are cited in (Banerjee and Roger- son, 2002), (Chakraborty and Banerjee, 1999). For brevity interested readers may go through those papers. b) For the ς− plane (complex plane) It is not possible to obtain an exact solution for the transformed equations (12,15) and we have to search for a suitable expression for the mapping function z = f (ς) and the deflection function in the form of W(ζ, − ζ) compatible with the prescribed boundary conditions. About the mapping function While employing a conformal technique one must be careful that a proper map- ping function for the contour is known. Sometimes the mapping function may be expressed in a power series. For example, in the case of a regular polygon, the power series may be expressed in the form (Laura and Luisoni, 1976) z = f (ζ) = apAs ∞∑ n=0 anζ 1+ns (21) where As are known as the mapping function coefficients. (Pombo et al., 1977) has prescribed the values of z = f (ς) and As for some structures of arbitrary shapes which may be given here for ready reference. For the deflection function a general expression has also been reported in (Pombo et al., 1977) in the form z = f (ς) = apλ ( ζ + µζm+1 ) (22) W(ζ, − ζ) = N∑ n=1 bn [ 1 − ( ζ − ζ )n] (23) CONSTANT DEFLECTION CONTOUR METHOD 75 Shapes s As m λ µ Triangular plate 3 1.1352 Square plate 4 1.0788 Pentagon 5 1.0516 Hexagon 6 1.0376 Heptagon 7 1.0279 Octagon 8 1.0220 Circular plate 1.00 0.00 Circular plate with Two flat sides 2 0.9224 0.00 Elliptic plate x2 4/3 + y2 4/5 = 1 0.99 0.101 Square plate with rounded corner 4 25/45 -0.4 6. Some results carried out in this investigation (a) Using CDC-Method TABLE I. Frequency parameter Ω� = ωa2 √ ρh D for a clamped annular plate ν = 0.3, (b/a = 0.5) Finite Strip (Banerjee 1976) FEM (Laura et al., 1973) (Leissa 1969) Present Ω =17.747 17.85 17.70 17693 76 BANERJEE The static deflections may be obtained as a byproduct as qa4 Eh4 = 5.8489ξ + 2.754ξ3..................(Yamaki, 1961) 5.861ξ + 2.762ξ3‘..................(RCP-rigid-circular-plate, present) 17.655ξ + 44.3237ξ3..............(AP-annular plate, present) (b) Using conformal mapping technique Considered here vibration of a square plate of side 2a for which z = f (ς) = a(1.08ς − 0.108ς5 + 0.045ς9) has been considered with W ( ζ, − ζ, t ) = k∑ k=1 bk ψk(ζ, − ζ)F(t) = k∑ k=1 bk[1 − (ζ, − ζ)2F(t) and following the method described earlier, the time differential equation is ob- tained as 0.1599 .. F + 6.4851F + ( Eh / ρa4 ) F = 1.7033 q / ρh2, ρ = h f (t) which is in excellent agreement with that of (Yamaki, 1961). Moreover, the con- formal approach has been successfully made in author’s note (Laura and Chi, 1965) when the large deflection problem was investigated in close agreement with known available results. Also the results are in excellent agreement with those given in (Banerjee and Rogerson, 2002). The results for static deflection of annular plates cannot be compared for non-availability of such studies. However, for the linear case the result can be compared with those given in (Timoshenko and Woinowisky-Krieger, 1959), e.g., Wmax/h = 0.0575 qa4 Eh4 (Ref.[22]), Wmax/h = 0.0566 qa4 Eh4 (Present) Moreover, in the nonlinear static case the results for RCP are even better than those of (Yamaki, 1961), in the sense that the present results are closer to those of (Way, 1934). In conclusion it may be accepted that the application of constant deflection contour method is justified and appears to be simpler than existing methods. The application of polynomial expressions for the deflection and the stress functions CONSTANT DEFLECTION CONTOUR METHOD 77 in conjunction with the Galerkin procedure appears to produce excellent results. In the case of nonlinear vibration, the results compare very well with those previ- ously obtained (Nowinski and Ohanabe, 1972). Additionally, the load deflection relations coincide very closely with that of (Way, 1934). The present analysis offers potential for analyzing the nonlinear response of complex structures. The negative aspect of this approach is that the use of Conformal mapping technique restricts one in finding the proper mapping function. References Banerjee M. M. (1976) Note on the large deflections of irregular shaped plates by the method of conformal mapping, Trans. ASME, 356-357. Banerjee M. M., Das J. N. (1990) Use of conformal mapping technique on dynamic response of plate structures, Proc. of ICOVP-1, A. C. College, Jalpaiguri, 129-131. Banerjee M. M. (1997) A new approach to the nonlinear vibration analysis of plates and shells, trans. 14th Intl. Conf. On Struc. Mech. In Reactor Tech., (SMIRT-13), Divn. B, Lyon, France 247. Banerjee M. M., Rogerson G. A. (2002) An application of the constant contour deflection method to non-linear vibration, Archive of Applied Mechanics 72, 279-292. Bucco D., Mazumdar J. (1979) vibration analysis of plates of arbitrary shape- A new approach, J. Sound and Vibration 67, 253-262. Chakraborty S., Roychoudhury B., Banerjee M. M. (1997) Nonlinear vibration of elliptic plates by the application of constant deflection contour method, Maths. and stats. In Eng. & Tech. Education, Ed. A. Chattopadhyay, Narosa Publishing House, New-Delhi., National Seminar on recent trends and advances of Maths. and Stats. In Eng. & Tech. Chakraborty S., Banerjee M. M. (1999) Large amplitude vibration of plates of arbitrary shape with uniform thickness, Proc. 4th ICOVP, Jadavpur University (INDIA) 85-89. Jones, R., Mazumdar J., Chiang F. P. (1975) Further studies in the application of the method of constant deflection lines to plate bending problems, Intl. J. Eng. Sci. 13, 423-443. Jones R., Mazumdar, J. (1997) Transverse vibration of shallow shells by the method of constant deflection contours, J. Accoust. Soc. Am. 56, 1487-1492. Laura P. A. A., Chi M., (1965) An application of conformal mapping to a three dimensional unsteady heat conduction problem, Aeronautical Quarterly 16, 221-230. Laura P. A. A., Rumanelli E. (1973) Determination of eige values in a class of wave-gudes of doubly connected cross section, JI. Sound and Vibration 26, 395-400. Laura P. A. A., Luisoni L. E. (1976) Discussion on t paper “An analytical and experimental study of vibrating equilateral triangular plates”, Proc. of the Soc. of Experimental stress analysis 33, 279-280. Leissa A. W., Laura P. A. A., Gutierrez R. H. (1967) Transverse vibrations of circular plates having non-uniform edge constraints, JI. Acous. Soc. Am. 66, 180-184. Leissa A. W. (1969) Vibration of plates, NASA SP-160. Mazumdar J. (1970) A method for solving problems of elastic plates of arbitrary shapes, J. Aust. Math. Soc. 11, 95-112. Mazumdar J., Bucco D. (1978) Transverse vibrations of visco-elastic shallow shells, J. Sound and Vibration 57, 323-331. Nowinski J., Ohanabe H. (1972) On certain inconsistencies in Berger equations for large deflections of plastic plates, Intl. J. Mech. Sc. 14, 165-170. 78 BANERJEE Pombo J. L., Laura P. A. A., Gutierrez R. H., Steinberg D. S. (1977) Analytical and experimen- tal investigaiton of the free vibrations of clamped plates of regular polygonal shape carrying concentrated masses, J. Sound and Vibration 55, 521-532. Timoshenko S. P., Woinowisky-Krieger, (1959) Theory of plates and shells, 2nd Edn. McGraw-Hiil, Newyork. Yamaki N. (1961) Influence of large amplitude on flexural vibrations of elastic plates, ZAMM 41, 501-510. Volmir A. S. (1956) flexible plates and shells, Moscow, 214. Way S. (1934) Bending of circular plates with large deflection, Trans. ASME, 56, 627-636. 79 79–84. NONLINEAR WAVE EQUATIONS AND BOUNDARY CONTROL USING VISCO ELASTIC DAMPERS Stephen P. Banks1, Metin U. Salamcı2 1The University of Sheffield, Automatic Control and Systems Eng., Sheffield, UK 2Gazi University, Mechanical Engineering Department, Ankara, Turkey Abstract. In this paper we shall study the boundary control and stabilization of nonlinear wave systems which are controlled by viscoelastic dampers on the boundary. This requires the use of fractional derivatives and a recent ‘iteration’ technique for converting nonlinear problems into a sequence of time-varying linear ones. We shall use the theory of evolution operators on Hilbert space and obtain an abstract setting for the problem. Key words: nonlinear systems, wave equations, time varying approximations 1. Introduction In this paper we generalize results of (Mbodje), (Krall, 1989) to the case of the boundary control of nonlinear wave-type systems with a boundary viscoelastic damper of fractional derivative type (used, for example, in high performance air- craft, etc.). Using a recent iteration technique which converts the system into a sequence of linear, time-varying systems we show that the nonlinear controlled system is asymptotically stable. The full proofs of the results, which are fairly technical, will be given in an extended version of the paper. We shall require the exponentially modified versions of fractional derivative and integration of order α, 0 < α < 1, as follows (the basic properties of fractional derivatives and their calculus can be found in (Podlubny, 1999)): (Dα,η f )(t) = ∫ t 0 (t − τ)−αe−η(t−τ) Γ(1 − α) d f (τ) dτ dτ (1) and (Iα,η f )(t) = ∫ t 0 (t − τ)α−1e−η(t−τ) Γ(α) f (τ)dτ, (2) which have the relationship Dα,η f = I1−α,ηD f . The systems we consider are of the form ∂2t u(x, t) = ∂ 2 xu(x, t) − ∂tu(x, t) · f (u, ∂tu, ∂xu) Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 80 BANKS, SALAMCI u(0, t) = 0 ∂xu(1, t) = −k∂α,ηt u(1, t), 0 < α < 1, η ≥ 0, k > 0, (the boundary control) u(x, 0) = u0(x), ∂tu(x, 0) = v0(x) (3) for some nonlinear, bounded function f (not necessarily positive). This system will be studied as the limit of a sequence of linear, time-varying systems of the form ∂2t u [i](x, t) = ∂2xu [i](x, t) − ∂tu[i](x, t) · f (u[i−1], ∂tu[i−1], ∂xu[i−1]) u[i](0, t) = 0 ∂xu [i](1, t) = −k∂α,ηt u[i](1, t), 0 < α < 1, η ≥ 0, k > 0, u[i](x, 0) = u0(x), ∂tu [i](x, 0) = v0(x) (4) with the starting system ∂2t u [1](x, t) = ∂2xu [1](x, t) − ∂tu[1](x, t) · f (u0, v0, u0x) u[1](0, t) = 0 ∂xu [1](1, t) = −k∂α,ηt u[1](1, t), 0 < α < 1, η ≥ 0, k > 0, u[1](x, 0) = u0(x), ∂tu [1](x, 0) = v0(x). (5) As shown in (Mbodje), the boundary control term is equivalent to the system ∂tϕ(ξ, t) + ξ 2ϕ(ξ, t) + ηϕ(ξ, t) − ∂tu(1, t)µ(ξ) = 0, ϕ(ξ, 0) = 0 (6) ∂xu(1, t) = − k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ(ξ, t)dξ (7) where, µ(ξ) = |ξ|(2α−1)/2. Hence we have a sequence of systems of the form ∂2t u [i](x, t) = ∂2xu [i](x, t) − ∂tu[i](x, t) · f (u[i−1], ∂tu[i−1], ∂xu[i−1]) u[i](0, t) = 0 (8) ∂tϕ [i](ξ, t) + ξ2ϕ[i](ξ, t) + ηϕ[i](ξ, t) − ∂tu[i](1, t)µ(ξ) = 0, −∞ < ξ < ∞ (9) ∂xu [i](1, t) = − k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ[i](ξ, t)dξ u[i](x, 0) = u0(x), ∂tu [i](x, 0) = v0(x), ϕ [i](ξ, 0) = 0. (10) In order to apply the general theory of evolution operators, we write these systems in abstract form as follows: first introduce the Hilbert space H = W1(0, 1) × L2(0, 1) × L2(−∞,∞) (11) NONLINEAR WAVE EQUATIONS 81 with the inner product 〈(u, v, ϕ), ( f , g, h)〉H = 1 2 ∫ 1 0 (∂xu)(∂x f )dx + ∫ 1 0 v · gdx + k π sin(απ) ∫ ∞ −∞ ϕhdx (12) and define the unbounded operator A by A u v ϕ = v ∂2xu −(ξ2 + η)ϕ + v(1)µ(ξ) The domain of A is D(A) = (u, v, ϕ) ∈ H : u ∈ W2(0, 1), v ∈ W1(0, 1) −(ξ2 + η)ϕ + v(1)µ(ξ) ∈ L2(−∞,∞) ∂xu(1) + kπ sin(απ) ∫ ∞ −∞ µ(ξ)ϕ(ξ)dξ = 0 |ξ|ϕ ∈ L2(−∞,∞). (13) We can now write the original system in the form d dt X(t) = AX(t) + F (X(t)) · X(t) (14) where, X(t) = (u(t), ∂tu(t), ϕ(t))T , X(0) = (u0, v0, ϕ0)T ∈ H and F (X) = 0 0 0 f (u, ∂tu, ∂xu) 0 0 0 0 0 In the abstract form, the iteration scheme now takes the form d dt X[i](t) = AX[i](t) + F (X[i−1](t)) · X[i](t), i ≥ 2 (15) with X[i](0) = (u0, v0, ϕ0)T ∈ H and d dt X[1](t) = AX[1](t) + F (X(0)) · X[1](t) (16) with X[1](0) = (u0, v0, ϕ0)T . As in (Mbodje), the Lumer-Phillips theorem shows that A generates a strongly continuous semigroup, i.e. a set of operators {T (t) : t ≥ 0} such that AX = lim t→0+ 1 t (T (t) − I)X (17) for X ∈ D(A). Thus, for X ∈ D(A), we have d dt (T (t)X) = AT (t)X = T (t)AX, t > 0. (18) 82 BANKS, SALAMCI Note that ∩nD(An) is dense in H. Moreover, we shall see that X[i−1](t) → X(t) (in H), the solution of the nonlinear problem, and each X[i−1](t) is stable, so that A + F (X[i−1](t)) generates an evolution operator. 2. Stability of the controlled system Consider the energy measure E1(t) = 1 2 ∫ 1 0 { |∂tu(t, x)|2 + |∂xu(t, x)|2 } dx. (19) We have dE1(t) dt = ∫ 1 0 ( ∂tu∂ 2 t u + ∂xu∂xtu ) dx = ∫ 1 0 ∂tu∂ 2 t udx − ∫ 1 0 ∂2xu∂tudx +∂tu(1, t) k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ(ξ, t)dξ = ∫ 1 0 ∂tu(−∂tu f )dx +∂tu(1, t) k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ(ξ, t)dξ. (20) The energy term connected with the diffusion is E2(t) = k 2π sin(απ) ∫ ∞ −∞ |ϕ(ξ, t)|2dξ (21) and we have dE2(t) dt = − k π sin(απ) ∫ ∞ −∞ (ξ2 + η)ϕ2(ξ, t)dξ + k π sin(απ)∂tu(1, t) ∫ ∞ −∞ µ(ξ)ϕ(ξ, t)dξ. (22) Hence, for the total energy E = E1 + E2, we have dE dt = − ∫ 1 0 (∂tu) 2 f dx − k π sin(απ) ∫ ∞ −∞ (ξ2 + η)ϕ2(ξ, t)dξ ≤ ∫ 1 0 (∂tu) 2 f−dx − k π sin(απ) ∫ ∞ −∞ (ξ2 + η)ϕ2(ξ, t)dξ (23) NONLINEAR WAVE EQUATIONS 83 where f− = { 0 i f f ≥ 0 f i f f < 0 . Hence dEdt ≤ 0 if ∫ 1 0 (∂tu) 2 f−dx ≤ k π sin(απ) ∫ ∞ −∞ (ξ2 + η)ϕ2(ξ, t)dξ (24) i.e. if the positive part of the system ‘damping’ is dominated by the boundary damping, which will be the case if | f−| is sufficiently small. The following results, for each typical time-varying approximation above are now simple extensions of the theory given in (Rodriguez and Banks, 2003; Banks, 2001; Cimen and Banks, 2004): Theorem 1 Although not exponentially stable, in general, the system ∂2t u [i](x, t) = ∂2xu [i](x, t) − ∂tu[i](x, t) · f (u[i−1], ∂tu[i−1], ∂xu[i−1]) u[i](0, t) = 0 (25) ∂tϕ [i](ξ, t) + ξ2ϕ[i](ξ, t) + ηϕ[i](ξ, t) − ∂tu[i](1, t)µ(ξ) = 0, −∞ < ξ < ∞ (26) ∂xu [i](1, t) = − k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ[i](ξ, t)dξ u[i](x, 0) = u0(x), ∂tu [i](x, 0) = v0(x), ϕ [i](ξ, 0) = 0 (27) is asymptotically stable, for sufficiently high gain k, in the sense that it converges to the maximal invariant subset of I = {X ∈ H : dE(t) dt = 0, ∀ t ≥ 0}.� (28) (This follows from LaSalle’s invariance principle.) Theorem 2 If η > 0 then the solution of each time-varying approximation decays as 1/t, so that ‖X(t)‖ ≤ K t (29) where K depends on sup ‖ f ‖ and X0, but not on the iteration argument i, provided the fedback gain k is large enough. � The stability of the nonlinear system now follows from the general theory of linear approximation developed in (Rodriguez and Banks, 2003; Banks, 2001; Cimen and Banks, 2004): Theorem 3 The system ∂2t u(x, t) = ∂ 2 xu(x, t) − ∂tu(x, t) · f (u, ∂tu, ∂xu) u(0, t) = 0 (30) 84 BANKS, SALAMCI ∂tϕ(ξ, t) + ξ 2ϕ(ξ, t) + ηϕ(ξ, t) − ∂tu(1, t)µ(ξ) = 0, −∞ < ξ < ∞ (31) ∂xu(1, t) = − k π sin(απ) ∫ ∞ −∞ µ(ξ)ϕ(ξ, t)dξ u(x, 0) = u0(x), ∂tu(x, 0) = v0(x), ϕ(ξ, 0) = 0 (32) is asymptotically stable, for sufficiently high gain k, in the sense that it converges to the maximal invariant subset of I = {X ∈ H : dE(t)dt = 0, ∀ t ≥ 0}. Moreover we have ‖X(t)‖ ≤ Kt if η > 0. � An estimate for the required gain is given by k > sup π ∫ 1 0 (∂tu)2 f−dx sin(απ) ∫ ∞ −∞(ξ 2 + η)ϕ2(ξ, t)dξ . (33) 3. Conclusions In this paper we have proved stabilizability results for nonlinear wave systems with boundary viscoelastic damping. Using an iteration sequence which converts the system into a sequence of linear, time-varying approximations which can be stabilized by a simple extension of the existing theory. The method applies to other types of nonlinearities in the dynamics, such as ∂2t u(x, t) = (1 + ν(u, ∂tu, ∂xu))∂ 2 xu(x, t) − ∂tu(x, t) · f (u, ∂tu, ∂xu) (34) by using perturbation theory of semigroups. Acknowledgements This work was partially supported by The British Council, Turkey under the science partnership programme. References Banks S. P. (2001) Exact boundary controllability and optimal control for a generalised Korteweg-de Vries equation, J. Nonlinear Anal.- Apps and Methods 47, 5537–5546. Cimen T., Banks S. (2004) Global Optimal Feedback Control for General Nonlinear Systems with Nonquadratic Performance Criteria, Sys. Cont. Letts. pp. 327–346. Krall A. (1989) Asymptotic Stability of the Euler-Bernoulli Beam with Boundary Control, J. Math. Anal. App. 137, 288–295. Mbodje B. (to appear in IMA) Wave Energy Decay under Fractional Derivative Control, Journal of Math. Cont. and Inf. Podlubny I. (1999) Fractional Differential Equations, Maths. in Sci. and Eng., Vol 198, Academic Press. Rodriguez M. T., Banks S. (2003) Linear Approximations to Nonlinear Dynamical Systems with Applications to Stability and Spectral Theory, IMA J. Math. Cont 20, 89–103. 85 85–90. EFFECT OF VIBRATIONS ON TRANSPORTATION SYSTEM Gülin Birlik1 and Önder Cem Sezgin2 1Dept. of Engineering Sciences Middle East Technical University, Ankara, Turkey 2Health and Counseling Center Middle East Technical University, Ankara, Turkey Abstract. In overly populated cities people living in suburban areas have to endure long journeys in order to reach their job sites. Whether they go by train, bus or by car they are inevitably exposed to vibrations, of considerable magnitude, in vertical (z) and lateral (x, y) directions. The immediate effect of vibration exposure is the fatigue of ones’ muscles. This is verified by the blood and saliva analysis of the volunteers travelling in a train. Their lactic acid levels were increased by 34% at the end of a 5̃ hr journey. The most affected people by vibration were, without doubt, the train operators and bus drivers. 42% of the suburban train operators had pain complaints at their waists. az( f loor) in the machinist cabin of a suburban train was measured to be, on the average, 0.23 m/s2. Max peak was 1.34 m/s2. The bus and car drivers were exposed to lower vibrations but they were exposed to multiple shocks originating from the non-standardized humps placed on the roads. Peak az(seat) = 0.054 m/s2 (f = 5.25 Hz) (vcar = 30 km/hr) on an asphalt road increased considerably while crossing over a hump. This value was 1.27 m/s2 (f = 4.5 Hz) in case of bus drivers (vbus = 20 km/hr). Studies have been done to provide practical measures for the reduction of the vibrations transmitted to the drivers. The waist belts filled with fluids of different viscosities prepared for this purpose seemed to be promising. The cushions filled with glycerin and gel were observed to be the best alternatives. Key words: vibration, transportation system 1. Introduction It is a well known fact that people who are exposed to vibration may have serious problems not only in their musculo-skeletal systems but also in their cardiovascu- lar and gastrointestinal systems. In this study two groups of people exposed to vibrations are observed both from the point of view of the severity of the vibration exposure and the after- effects of the vibration. First group was the drivers of the vehicles i.e., the group exposed to vibration in their working environment. Passengers comprised the sec- ond group, i.e., the group either exposed to vibrations regularly (but at intermittent intervals) or from time to time. For this purpose acceleration measurements have been done in the suburban train, intercity train, service bus and a car during their routine cruises. Among the drivers of the vehicles train operators were observed to be the ones most severely affected by vibrations transmitted to their bodies. In case of passengers the adverse effects of the humps, crossed over frequently, E. Inan and A. (eds.), Vibration Problems ICOVP 2005, c . © 2007 Springer. K r şıı 86 BİRLİK, SEZGİN during the travel was found to be a serious problem which deserves a detailed follow-up study. 2. Acceleration measurements in trains 2.1. MACHINIST CABIN The measurements were done in machinist cabins of suburban train (ttravel=35 mins), (vtrain=100 km/hr) and Baskent Express (ttravel=5 hrs), (vtrain ≤100 km/hr). Since the speed of the train change continuously during the measurements, only the mean acceleration values are given in Table 1. AEQ and maxP define re- TABLE I. Acceleration Values (m/s2) (Suburban Train) (direction of travel: x direction) Measurement Point AEQ Max P SUM y Z x y z Machinist Cabin Floor (av. of 10 measurements) 0.195 0.226 0.28 0.92 1.34 0.358 Machinist Seat (av. of 10 measurements) 0.197 0.234 0.46 0.90 1.40 0.372 spectively the equivalent continuous vibration (acceleration) level and maximum vibration level, which was reached during the measurement period. SUM values (SUM= √ (1.4ax)2 + (1.4ay)2 + a2z ) are calculated using measured ax, ay and az, acceleration values in the x (fore and aft), y (lateral) and z (vertical) directions. Values given in the tables show that the machinists were exposed to peak acceler- TABLE II. Acceleration Values (m/s2) (Baskent Express Train) (direction of travel: x ) Measurement Point AEQ Max P SUM y z x y z Machinist Cabin Floor (av. of 10 measurements) 0.345 0.675 0.46 1.23 2.81 0.843 Machinist Seat (av. of 10 measurements) 0.279 0.652 2.66 1.61 3.71 0.952 EFFECT OF VIBRATIONS ON TRANSPORTATION SYSTEM 87 ations higher than 1 m/s2 both in standing and sitting postures in the lateral and vertical directions. The mean age of the machinists was 40 (youngest: 27, oldest: 50); and they work in this environment on the average 64 hours a week (Minimum: 40 hrs/week, maximum: 75 hrs/week). 75% of the machinists also go to their homes by train. (Home - work: minimum 20 minutes; maximum 120 minutes). 42% of the operators had waist and 17% had back complaints. 42% of them had neck pain. Open windows of the locomotives can be held responsible for the neck pain of the operators. The high accelerations measured on the operator seat displayed clearly the inefficiency of the seat in isolating the vibrations transmitted from the train floor to the operator. Compared to suburban train machinists of the Baskent express were exposed to high levels of vibration (It has to be noted that duration of the exposure is generally > 6 hrs) which is defined in the ISO2631/1 as “undesired” for the health. 2.2. PASSENGER COACHES Acceleration values (floor SUM = 0.318 m/s2; seat SUM = 0.324 m/s2) in the passenger coach of the suburban train were less than the accelerations measured in the operator cabin. The situation did not change in case of intercity train. Even though, maximum acceleration transmitted, in z direction, to the seat (seat No: 25) of the passenger of Baskent Express (vtrain=98 km/hr) was 0.13 m/s2 (at f = 1.5 Hz) (ay = 0.0508 m/s2 at f = 8.5 Hz). 34% increase in the lactic acid levels (in blood) showed of the 9 volunteers observed at the end of an 5̃ hours of travel indicates the fatigue of the muscles of the passengers during the travel. The decrease of both in hemoglobin and WBC levels, and the increase in ESR levels could be however thought to be as a result of vasodilatation of microcirculation. Saliva analysis did not exhibit meaningful changes in Na+ and K+ concentrations. Na+ concentrations increased on the average 4%, and K+ decreased 1%, and the rate of K + /Na+ decreased form 1.50 to 1.43. 3. Acceleration measurements in buses Acceleration measurements were done on the floor and at the interfaces between the buttocks and the seat and between the waist (and back) of the driver (or the passenger) and the back rest in x, y and z directions simultaneously, while crossing over a hump (3.1 m wide, 12 cm high) 4. Some measures To reduce the effects of vibrations transmitted to the waist of the driver waist cushion (28.5 cm x 21.5 cm) belts were prepared. The cushions were tied to the waist (cushion part placed at the back of the driver at L3 level) of the bus driver. 88 BİRLİK, SEZGİN Figure 1. Vibrations transmitted to the bus driver (vbus = 20 km/hr) while crossing over the hump Compared to the cushion filled with gel the cushion filled with glycerin seemed to be more effective in reducing the low frequency vibrations. 5. Acceleration measurements in cars Peak acceleration values measured in an old car (production year: 1973) are tab- ulated in Table 3. In all the measurements velocity of the car was 30 km/hr. TABLE III. Peak az values (m/s2) Seat Location Floor Seat Waist Driver (f=3.25 Hz) (Birlik and Sezgin, 2001) 0.237 0.503 0.264 Front Seat Passanger (f=3.25 Hz) 0.355 0.660 0.432 EFFECT OF VIBRATIONS ON TRANSPORTATION SYSTEM 89 Figure 2. Vibrations, transmitted to the bus passenger seated at the first row, while crossing over the hump (vbus = 20 km/hr) Figure 3. Isolation efficiency of waist cushion belts while crossing over the hump 90 BİRLİK, SEZGİN 6. Conclusions The analysis of the blood samples taken from the volunteers had clearly displayed the tiredness of the passengers. From the point of view of vehicle drivers this result has to be regarded not as a problem of just tiredness but rather as a risk for health in the long run. Until more secure traffic calming measures are introduced the waist cushion belts may be used in alleviating the adverse effects of vibrations transmitted to the drivers. References Birlik G., Sezgin O. (2001) Speed reducers on the roads, Inter-noise, The Hague, the Netherlands 1073-1077. 91 91–96. IDENTIFICATION OF BONE MICROSTRUCTURE FROM EFFECTIVE COMPLEX MODULUS ∗ Carlos Bonifasi-Lista and Elena Cherkaev University of Utah, Department of Mathematics, Salt Lake City, UT 84112, USA Abstract. This work deals with the problem of reconstruction of bone structure from measurements of its effective mechanical properties. We propose a novel method of calculation of bone porosity from measured effective complex modulus. Bone is modelled as a medium with a microstructure composed of trabecular bone (elastic component) and bone marrow (viscoelastic component). We model bone as a cylinder subjected to torsion and assume that the effective complex modulus can be measured as a result of experiment. The analytical representation of the effective complex mod- ulus of the two-component composite material is exploited to recover information about porosity of the bone. The microstructural information is contained in the spectral measure in the Stieltjes representation of the effective complex modulus and can be recovered from the measurements over a range of frequencies. The problem of reconstruction of the spectral measure is very ill-posed and requires regularization. To verify the approach we apply it to analytically and numerically simulated response of a cylinder subjected to torsion assuming that it is filled with a composite material with known (laminated) microstructure. The values of porosity calculated from the effective shear modulus are in good agreement with the model values. Key words: inverse homogenization, bone microstructure, torsion, viscoelasticity, porosity, effec- tive properties 1. Introduction Trabecular bone has a porous structure formed by trabeculae and filled by bone marrow. Various models of trabecular bone have been developed based on homog- enization of materials with microstructure. They are used in computations of effec- tive mechanical properties of bone, its static and dynamic responses (Tokarzewski et al., 2001). On the other hand, for practical applications, it is important to be able to estimate parameters of the bone structure from measured effective mechanical properties. Bone porosity is one of such parameters important for osteoporosis monitoring. We propose a novel method of evaluation of bone porosity and struc- ture from measured effective complex modulus based on inverse homogenization approach. Inverse homogenization approach was developed in (Cherkaev, 2001) ∗ Supported by NSF grant DMS-0508901 and The University of Utah Research Foundation seed grant. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 92 BONIFASI-LISTA, CHERKAEV for recovering microstructural information from the effective complex permittivity of a two-component composite material. In the present paper we consider bone as a beam or a cylinder filled with two component composite consisting of an elastic matrix (cancellous bone) and a viscoelastic solid (bone marrow). The cylinder is subjected to torsion; we assume that the effective complex modulus can be measured as a result of torsion experiment. We derive homogenized equations for the torsion of a cylinder filled with material with microstructure. We show that the effective complex modulus admits a Stieltjes analytic representation. The information about the structure of the medium is contained in the spectral mea- sure in the analytic representation, the spectral function can be recovered from the measurements over a range of frequencies. From the computational point of view, the problem of reconstruction of the spectral function is extremely ill-posed and requires regularization. We show computational results of reconstruction of the spectral function and estimation of bone porosity from numerically simu- lated effective complex modulus of a third-rank composite using a regularized algorithm. 2. Mathematical model We consider bone as a cylinder of finely scaled porous elastic matrix (cancellous bone) filled with a viscoelastic solid (bone marrow) (Tokarzewski et al., 2001). Both materials are assumed to be isotropic. The bottom of the cylinder is fixed and the top is subjected to a force couple. Let ui be the displacement vector field, σ and � be the stress and strain respectively. The tensor ani jkl, n = 1, 2, represents the material properties which depend on the microstructure, the tensor can take different values corresponding to the elastic and viscoelastic components of the bone. The governing equations and boundary conditions are: σni j, j = 0 on Ωn, n = 1, 2 (1) where Ωn, n = 1, 2 is the domain occupied by the first or second material. The constitutive relations correspond to isotropic materials with complex Lame moduli λ and µ: σni j = a n i jkl� n kl = λnu n k,kδi j + µn ( uni, j + u n j,i ) in Ωn, n = 1, 2 (2) The interface boundary conditions are given by: u1i = u 2 i , σ 1 i jn j = σ 2 i jn j on ∂Ω1 ∩ ∂Ω2 (3) On the side surface of the cylinder: σ1i jn j = 0 on ∂Ω (4) IDENTIFICATION OF BONE STRUCTURE 93 3. Homogenization of heterogeneous material in torsion problem We assume that the microstructure is periodic and using two-scale asymptotic expansion method, we follow the homogenization procedure (Sanchez-Palencia, 1980). We introduce a small parameter � characterizing the fine scale of variations of microstructure and reformulate the problem: σ�i j, j = 0 on Ω � , σ�i j(x) = a � i jkl(x) � � kl(x) (5) The material properties tensor a is �-Y periodic in the x scale and is defined as: a�i jkl(x) = ai jkl(x/�) = ai jkl(y) = χ(y) a 1 i jkl + (1 − χ(y)) a 2 i jkl (6) where χ(y) is the characteristic function of the domain occupied by the first mate- rial: χ(y) = { 1, if y ∈ Ω1 0, otherwise (7) Here � is the characteristic length of the cell of periodicity Y . We introduce the following two-scale asymptotic expansions for the vector displacement field: u� = u(x, y) = u0(x) + �u1(x, y) + �2u2 + ... (8) The terms u0 and u1 of the expansion of the displacement field u� have the form (Muskhelishvili, 1963): u0(x) = (−βx2x3, βx1x3, βw0(x)); u1(x, y) = (const, const, βw1(x, y)) where β is the torsional angle, and only the u3 component depends on the micro- scopic scale y. Notice that u3 only varies in the (x1, x2) (and (y1, y2)) plane. For the given displacement field, the only nonzero stresses are σ031(x, y) and σ 0 32(x, y). Substituting the asymptotic expansions in (5) we obtain the following local prob- lem: �−1 → ∂σ0i j ∂y j (x, y) = 0 (9) Here σ0i j = ai jlm [ �lmx(u 0) + �lmy(u 1) ] = ai jlm 2 ∂u0m ∂xl + ∂u0l ∂xm + ∂u1m ∂yl + ∂u1l ∂ym From the local problem (9), we obtain for i=3: ∇y · ( a(y)∇yw1 ) + ∇y · a q = 0 (10) where the vector q is constant over the cell Y: q1 = −x2 + ∂w0(x) ∂x1 , q2 = x1 + ∂w0(x) ∂x2 94 BONIFASI-LISTA, CHERKAEV Solution of (10) can be represented as w1(x, y) = 2∑ k=1 φk(y) ∂ ∂xk ( w0(x) + (−1)kx j xk ) (11) where φk is a Y-periodic solution to the problem ∇y · a(y)∇yφk + ∇y · a(y)∇yyk = 0 (12) Since we consider the model of bone formed by isotropic materials, the com- ponents a3131, a3232 of the material properties tensor are the same: a3131(y) = a3232(y), so that a(y) = µ(y), and µ(y) = χ(y) µ1 + (1 − χ(y)) µ2 (13) where µi, i = 1, 2 is the complex shear modulus of the bone or the bone marrow. Homogenized complex shear modulus µ∗ is defined now by averaging the solution of the problem over the Y-cell: µ∗jk = ∫ Y µ(y) ( δ jk + ∂φk(y) ∂y j ) dy (14) 4. Stieltjes analytic representation for the effective shear modulus We derive the Stieltjes analytic representation of the effective shear modulus sim- ilar to the analytic integral representation of the effective complex permittivity which was developed for computing bounds for the effective permittivity of a two component composite (Bergman, 1978; Golden and Papanicolaou, 1983). The Stieltjes integral represents a function F(s) as an analytic function off [0, 1]- interval in the complex s−plane: F(s) = 1 − µ ∗ µ2 = ∫ 1 0 dη(z) s − z , s = 1 1 − µ1/µ2 (15) To derive this representation, we rewrite the local problem (12) on the cell Y as ∇ · µ(y)∇(φk + yk) = 0 (16) and notice that because of Y-periodicity of φk, the average of the corresponding strain field T = ∇(φk + yk) is given by the unit vector ek in the k−th direction, < T >= ek, where < · > stands for averaging over Y-cell. Using (13), we bring the last problem to the form: ∇ · χ (∇φk + ek) = s∆φk (17) IDENTIFICATION OF BONE STRUCTURE 95 Applying (−∆)−1 to both sides of this equation, then taking gradient, and intro- ducing the operator Γ as Γ = ∇(−∆)−1(∇·) , we can express T as a function of Γχ: T = s(sI + Γχ)−1ek. Using (13), we express F(s) as F(s) = 1 − 〈 µT, ek〉/µ2 = 〈 s−1 χ T, ek 〉 . The spectral resolution of Γχ with the projection valued measure Q results in the representation, F(s) = 〈 χ ( sI + Γ χ )−1ek, ek〉 = ∫ 1 0 〈 χ dQ(z) ek, ek〉 s − z (18) Introducing a function η corresponding to the spectral measure Q, dη(z) = 〈 χ dQ(z)ek, ek〉, we end up with integral representation (15), where the positive measure η is the spectral measure of a self-adjoint operator Γχ . The representation (15) separates information about the microstructure of the composite material, which is contained in the spectral function η from information about the properties of the materials. It is shown in (Cherkaev, 2001) that the spectral function η can be uniquely reconstructed if measurements of the effective property are available on an arc in a complex plane. Because the shear modulus of the viscoelastic com- ponent is frequency dependent, measurements of µ∗ in an interval of frequency should provide the required set of data. After the function η is reconstructed, the volume fraction of bone marrow can be calculated as zero moment of the function η (Bergman, 1978; Golden and Papanicolaou, 1983). 5. Computational method and simulation results The problem of reconstruction of the function η is equivalent to inverse potential problem, which is ill-posed problem. Practically this means that small variations in the data or computational noise in a numerical algorithm can lead to arbitrary large variations of the solution. With s = x + iy, real and imaginary parts of F(s) are Re(F(s)) = ∫ 1 0 (x − t) dη(t) (x − t)2 + y2 , Im(F(s)) = − ∫ 1 0 y dη(t) (x − t)2 + y2 (19) To construct a regularized solution we discretize the problem, and formulate min- imization problem having introduced a stabilization functional which constrains the set of minimizers. We reformulate the problem as an unconstrained mini- mization using the Lagrange multiplier. In case of a quadratic stabilization func- tional, the minimization problem and its Euler equation take the following form equivalent to Tikhonov regularization with regularization parameter α: min m∈Rn { ‖Km − f ‖22 + α ‖Lm‖ 2 } , mα = ( K∗K + αL∗L )−1 K∗ f . (20) We model bone as a cylinder filled with 3rd rank laminated composite made of cancellous bone and bone marrow with bone and marrow material properties 96 BONIFASI-LISTA, CHERKAEV 0 10 20 30 40 50 60 70 80 90 100 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 −0.01 −0.005 0 0.005 0.01 0.015 0.02 0.025 (a) (b) Figure 1. (a) The effective complex shear modulus over a range of frequency. (b) Reconstruction of η using Tikhonov regularization TABLE I. Computed estimates for bone porosity Technique / Porosity 70 % 75% 80 % 85% 90 % 95% Tikhonov regularization, Im[F(s)] 69.99 74.97 79.97 84.98 90.02 94.99 Tikhonov regularization, Re[F(s)] 69.99 74.98 79.97 84.99 90.02 95.00 (Cowin, 2001). The effective shear modulus of these structures is calculated ana- lytically and is used as a data set in reconstruction algorithm. Figure 1(a) shows data used in computations. Figure 1(b) shows function η for 3d rank laminated composite computed with Tikhonov regularization. The volume fraction of bone marrow is estimated for several values of bone porosity, the results are shown in Table 1. References Bergman D. J. (1978) The dielectric constant of a composite material - A problem in classical physics, Phys. Rep. C 43 377-407. Cherkaev E. (2001) Inverse homogenization for evaluation of effective properties of a mixture, Inverse Problems 17 1203-1218. Cowin S. C. (2001) Bone Mechanics Handbook, CRCPress. Golden K., Papanicolaou G. (1983) Bounds on effective parameters of heterogeneous media by analytic continuation, Comm. Math. Phys. 90 473-491. Muskhelishvili N. I. (1963) Some basic problems of the mathematical theory of elasticity, 22, Noordhoff ltd Groningen. Sanchez-Palencia E. (1980) Non-homogeneous Media and Vibration Theory, Springer-Verlag Berlin Heidelberg New York. Tokarzewski S., Telega J. J., Galka A. (2001) Torsional Rigidities of Cancellous Bone Filled with Marrow: The Application of Multipoint Pade Approximants, Engineering Transactions 49 135- 153. 97 97–102. BEAM VIBRATIONS WITH NON-IDEAL BOUNDARY CONDITIONS Hakan Boyacı Department of Mechanical Engineering, Celal Bayar University, 45140 Manisa, Turkey Abstract. A simply supported damped Euler-Bernoulli beam with immovable end conditions is considered. The concept of non-ideal boundary conditions is applied to the beam problem. In accordance, the boundaries are assumed to allow small deflections and moments. Approximate analytical solution of the problem is found using the method of multiple scales, a perturbation technique. Key words: stretched beam vibrations, non-ideal boundary conditions, method of multiple time scales 1. Introduction Beams are frequently used as design models for vibration analysis. In such analy- sis, types of support conditions are important and have direct effect on the so- lutions and natural frequencies. The real systems are considered to satisfy those ideal boundary conditions. However, small deviations from ideal conditions in real systems indeed occur. For example, a beam connected at ends to rigid supports by pins is modeled using simply supported boundary conditions which require deflections and moments to be zero. However the hole and pin assembly may have small gaps and/or friction which may introduce small deflections and/or moments at the ends. Similarly a real built-in beam may have very small vari- ations in deflection and/or slope. To represent such behavior, non-ideal boundary condition concept has been proposed recently.Non-ideal boundary conditions are modeled using perturbations. Different beam vibration problems having non-ideal boundaries have been treated (Pakdemirli&Boyaci, 2001, 2002, and 2003). Here in this work, the idea is widened to a damped forced nonlinear simple-simple beam vibration problem in which the non-linearity is due to stretching effects. As non-ideality for a simple-simple support case, small variations at deflections and moments are allowed at both ends. Ideal and non-ideal frequencies as well as frequency response curves are contrasted. Combined effects of non-linearity and non-ideal boundary conditions on the natural frequencies and mode shapes are Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 98 BOYACI examined using the method of multiple scales, a perturbation technique. While the stretching effect increases the frequencies, non-ideal boundary conditions may increase or decrease the frequencies. 2. Problem formulation and solution The model considered is an Euler-Bernoulli beam with immovable end conditions causing nonlinear stretching effects. The dimensionless equation is ẅ + wiv = 1 2 εw′′ 1∫ 0 w′2dx − 2εµẇ + F̂ cosΩt (1) where w is the deflection, µ is the damping, t is the time, x is the spatial vari- able and F and Ω are the magnitude and frequency of the external excitation respectively. Dot denotes differentiation with respect to time t and prime denotes differentiation with respect to the spatial variable x. Here it is assumed that the beam is simply supported at both ends. However, the boundary conditions are not ideal and some slight variations occur in deflections and moments, hence w(0, t) = εα1(t),w′′(0, t) = εβ1(t), w(1, t) = εα2(t),w′′(1, t) = εβ2(t) (2) ε is a small perturbation parameter denoting that the variations in deflections and moments are small. An approximate solution of the below form is assumed w(x, t) = w0(x, T0, T1) + εw1(x, T0, T1) + . . . (3) where T0 is the usual fast time scale and T1=εt is the slow time scale in the method of multiple scales. Only primary resonance case is considered and hence, the forcing term is ordered as F̂ = εF. The time derivatives are defines as Substituting equation (3) into equations (1) and (2), separating at each order of ε, one has Order 1 D20w0 + w iv 0 = 0 w0(0, T0, T1) = 0,w′′0 (0, T0, T1) = 0, w0(1, T0, T1) = 0,w′′0 (1, T0, T1) = 0 (4) Order ε D20w1 + w IV = −2µD0w0 − 2D0D1w0+ 1 2 w ′′ 0 1∫ 0 w20dx + F cosΩT0w1(0,T0,T1) = α1(T0,T1) w′′1 (0,T0,T1) = β1(T0,T1) w1(1,T0,T1) = α2(T0,T1),w′′1 (1,T0,T1) = β2(T0,T1) (5) BEAM VIBRATIONS WITH NON-IDEAL BOUNDARY CONDITIONS 99 The solution at the first order is w0 = (A(T1)e iωT0 + cc)Y(x) (6) where cc stands for complex conjugates of the preceding terms. Substituting equa- tion (6) into equation (4) yields the normalized solution Y(x) = √ 2 sin nπx, ω = n2π2 n = 1,2,3, . . . (7) At order ε , one substitutes equation (6) into the right hand side of equation (5). The result is D20w1 + w iv 1 = −2iωD1A Y − 2iµωAY + 32 A2Ā Y ′′ 1∫ 0 Y ′2dx + 12 Fe iσT1 eiωT0 + N.S .T. + cc (8) where N.S .T. stands for non-secular terms. It is also assumed that the external excitation frequency is close to one of the natural frequencies of the system; such that Ω = ω + εσ (9) Here σ is a detuning parameter of order 1. A solution of the below form is assumed w1 = ϕ(x,T1)e iωT0 + W1(x,T0,T1) + cc (10) The first part of the solution is the one corresponding to secular terms and the second is the one corresponding to non-secular terms. Substituting equation (9) into equation (8) with boundary conditions yields ϕiv − ω2ϕ = −2iωD1A Y − 2iµωAY + 3 2 A2Ā Y ′′ 1∫ 0 Y ′2dx + 1 2 FeiσT1 (11) ϕ(0, T1) = α1A(T1), ϕ′′(0, T1) = β1A(T1), ϕ(1, T1) = α2A(T1), ϕ′′(1, T1) = β2A(T1) (12) In writing (11), the variations of deflections at the boundaries are considered to be of the same form as the time variations of the solutions and αi and βi are now constants. Since the homogenous problem has a non-trivial solution, the non- homogenous problem (10) and (11) has a solution only if a solvability condition is satisfied (Nayfeh, 1981). The solvability condition requires 2in2π2 (D1A + µA) + 3 2 n4π4A2Ā + √ 2nπKA − 1 2 f eiσT1 = 0 (13) 100 BOYACI where K = (β1 − β2 cos nπ − n2π2α1 + n2π2α2 cos nπ)and f = 1∫ 0 FYdx (14) Substituting the polar form A(T1) = 12 a(T1) e iθ(T1) into equation (12), separat- ing real and imaginary parts, one obtains n2π2 dadT1 = −n 2π2µa + 12 f sin γ n2π2a dγdT1 = n 2π2σa − 316 n 4π4a3 − √ 2 2 nπK + 1 2 f cos γ (15) where γ = σT1 − θ. In the steady state case, a′ = γ′ = 0 and solving for the detuning parameter yields σ = 3 16 n2π2a2 + √ 2 2nπ K ± √ f 2 4n4π4a2 − µ2 (16) For free undamped vibrations, non-ideal nonlinear natural frequencies are ωni = ω + ε 3 16 ωa2 + √ 2 2 √ ω K (17) In the above relation, the first term in O(ε) is due to the nonlinearity and the second term is due to the non-ideal boundary conditions. If only the non-ideality term of order ε is to be considered, it may increase or decrease the frequencies depending on the mode number n and amplitudes of end variations that are α1 due to deflections and β1due to moments. Also it is apparent that deflection effects become dominant compared to the moment effects as the mode number increases. Different cases are summarized in Table 1. In Figure 1a and 1b fundamental non- linear frequencies versus amplitudes are contrasted for the ideal and non-ideal cases using the same parameter values except α1 and α2. TABLE I. Effect of non-ideal boundary conditions on the frequencies n Frequency odd frequencies increase if β1+ β2 > n2π2(α1+ α2) frequencies decrease if β1+ β2 < n2π2(α1+ α2) no change if β1+ β2 = n2π2(α1+α2) even frequencies increase if β1 + n2π2α2¿ β2 + n2π2α1 frequencies decrease if β1 + n2π2α2¡ β2 + n2π2α1 no change if β1 + n2π2α2= β2 + n2π2α1 BEAM VIBRATIONS WITH NON-IDEAL BOUNDARY CONDITIONS 101 Figure 1a.,1b. Nonlinear frequencies versus amplitudes for ideal (dashed) and non-ideal (solid) cases for the first mode a)α1 = α2=0.1, β1 = β2=10, ε=0.1, b)α1 = α2=2.0, β1 = β2=10, ε=0.1 One can also obtain amplitude-excitation frequency relation from equations (9) and (14) as Ω = ω + ε 3 16 ωa2 + ε √ 2 2 √ ω K ± √ ε2 f 2 4ω2a2 − ε2µ2 (18) In Figure 2a and 2b frequency response graphs for the fundamental nonlinear frequencies are compared for the ideal and non-ideal cases using the same para- meter values except α1 and α2 . The approximate beam deflection to the first order are as follows w = a cos(Ωt − γ)Y(x) + O(ε) (19) where Y(x) is given in equation (9). Figure 2a.,2b. Frequency response curves for ideal (dashed) and non-ideal (solid) cases for the first mode 102 BOYACI b) α1 = α2=2.0, β1 = β2=10, ε=0.1, µ=0.5,f=10 3. Concluding remarks Non-ideal boundary conditions are defined and formulated using perturbation theory. A sample problem of stretched damped beam with external excitation is treated. Approximate analytical solution of the problem is presented using the method of multiple scales. It is shown that small variations of deflections and moments at the ends may affect the frequencies of the response. Depending on the mode numbers and amplitudes of variations, the frequencies may increase or decrease. Deviations from the ideal conditions lead to a drift in frequency- response curves which may be positive, negative or zero depending on the mode number and amplitudes of variations. References Nayfeh A. H. (1981) Introduction to Perturbation Techniques, New York. Pakdemirli M., Boyacı H. (2001) Vibrations of a stretched beam with non-ideal boundary conditions, Mathematical and Computational Applications 6 217-220. Pakdemirli M., Boyacı H. (2002) Effect of non-ideal boundary conditions on the vibrations of continuous systems, Journal of Sound and Vibration 249 815-823. Pakdemirli M., Boyacı H. (2003) Non-linear vibrations of a simple-simple beam with a non-ideal support in between, Journal of sound and Vibration 268 331-341. a) α1 = α2=0.1, β1 = β2=10, ε=0.1, µ=0.5,f=10, 103 103–109. A TWO-FRACTAL OVERLAP MODEL OF EARTHQUAKES Bikas K. Chakrabarti and Arnab Chatterjee Theoretical Condensed Matter Physics Division and Centre for Applied Mathe- matics and Computational Science, Saha Institute of Nuclear Physics, 1/AF Bid- hannagar, Kolkata 700064, India Abstract. We introduce here the two-fractal model of earthquake dynamics. As the fractured surfaces have self-affine properties, we consider the solid-solid interface of the earth’s crust and the tectonic plate below as fractal surfaces. The overlap or contact area between the two surfaces give a measure of the stored elastic energy released during a slip. The overlap between two fractals change with time as one moves over the other and we show that the time average of the overlap distribution follows a Gutenberg-Richter like power-law, with similar exponent value. Key words: earthquake, fractals, Cantor sets, Gutenberg-Richter law 1. Introduction The earth’s solid outer crust, about 20 kilometers in average thickness, rests on the tectonic shells. Due to the high temperature-pressure phase changes and the consequent powerful convective flow in the earth’s mantle (a fluid of very high density), at several hundreds of kilometers of depth, the tectonic shell, divided into a small number (about ten) of mobile plates, has relative velocities of the order of a few centimeters per year (Gutenberg and Richter, 1954; Kostrov and Das, 1989; Scholz, 1990). Over several tens of years, enormous elastic strains develop on the earth’s crust when sticking (due to the solid-solid friction) to the moving tectonic plate. When sudden slips occur between the crust and the tectonic plate, these stored elastic energies are released in ‘bursts’, causing the damages during the earthquakes. Earthquakes occur due to fault dynamics in the lithosphere. A geological fault is created by a fracture in the rock layers, and is comprised of the rock surfaces in contact. The two parts of the fault are in very slow relative motion which causes the surfaces to slide. Because of the uniform motion of the tectonic plates, the elastic strain energy stored in a portion of the crust (block), moving with the plate relative to a ‘stationary’ neighboring part of the crust, can vary only due to the random strength of the solid-solid friction between the crust and the plate. A slip occurs when the accumulated stress exceeds the resistance due to the frictional Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 104 CHAKRABARTI, CHATTERJEE force. The potential energy of the strain is thereby released, causing an earthquake. As mentioned before, the observed distribution of the elastic energy release in various earthquakes seems to follow a power law. The slip is eventually stopped by friction and stress starts developing again. Strain continues to develop till the fault surfaces again slip. This intermittent stick- slip process is the essential characteristic feature of fault dynamics. The overall distribution of earthquakes, including main shocks, foreshocks and aftershocks, is given by the Gutenberg-Richter law (1944, 1954): log10 Nr(M > M) = a − b M (1) where, Nr(M > M) denotes the number (or, the frequency) of earthquakes of magnitudes M that are greater than a certain value M. The constant a represents the total number of earthquakes of all magnitudes: a = log10 Nr(M > 0) and the value of the coefficient b is presumed to be universal. In an alternative form, the Gutenberg-Richter law is expressed as a relation for the number (or, the frequency) of earthquakes in which the energy released E is greater than a certain value E: Nr(E > E) ∼ E−b/β, suggesting Nr(E) ∼ E−γ, (2) for the number density of earthquakes, where γ = 1 + b/β. The value of the exponent γ is generally observed to be around unity (Knopoff, 2000); see also http://web.cz3.nus.edu.sg/˜chenk/gem2503 3/notes7 1.htm. One class of models for simulating earthquakes is based on the collective motion of an assembly of connected elements that are driven slowly, of which the block-spring model due to Burridge and Knopoff (1967) is the prototype. The Burridge-Knopoff model and its variants (Carlson et al., 1994; Olami et al., 1992) have the stick-slip dynamics necessary to produce earthquakes. The underlying principle in this class of models is self-organized criticality (Bak, 1997). Another class of models for simulating earthquakes is based on overlapping fractals, which will be discusses in details in the next sections. 2. Fractals A fractal is a geometrical object that displays self-similarity on all scales. For random fractals, the object need not exhibit exactly the same structure at all scales, but the same ‘type’ of structures appear on all scales. For example, the Black sea coastline measured with different length rulers will show differences: the shorter the ruler, the longer the ‘length’ measured but not exactly following the ratio of the inverse of the lengths of the rulers; greater than that! This is because more structures come into play at lower length scales; or in other words, the coastline is not really a ‘line’ but rather a fractal having dimension greater than unity. Look- ing at smaller and smaller length-scales, one can find self-similar structures (see Fig. 1). TWO-FRACTAL OVERLAP MODELS OF EARTHQUAKES 105 Figure 1. Fractal structure of Black sea coastline at Işık (Istanbul), represented actually by a time series of stock price. This shows the remarkable self-similarity (examplified by the blow up of the segment in the box) involved in many such natural processes Figure 2. Sierpinski gasket (at generation number n = 5), having fractal dimension log 3/ log 2 (when n → ∞) 106 CHAKRABARTI, CHATTERJEE n = 1 n = 2 n = 3 Figure 3. Construction process of Cantor set is shown (upto n = 3). The set becomes a fractal when n → ∞ One can easily construct such regular fractals as carpets or gaskets (see Fig. 2). In Fig. 2, a basic unit of equilateral triangle with an inner triangle obtained by joining the mid-points of each side and keeping the inner space void, one constructs a gasket. At each step, as the length of each side changes by a factor L = 2, the mass of the fractal changes by a factor M(= Ld f ) = 3, giving therefore the fractal dimension of the object to be d f = log 3/ log 2. This object in Fig. 2 of course represents a non-random fractal. However, a random fractal (as in Fig. 1) can be easily constructed if the void is not always at the center but at any of the 4 triangles at random for each generation; the (mass) dimension d f remain the same. A similar fractal of dimension log 2/ log 3 can be constructed as the Cantor set shown in Fig. 3. Starting from a set of all real numbers from 0 to 1 one removes the subset from the middle third, so that for each l = 3, the mass (size) of the set m = 2 and hence the dimension as n → ∞. This void subset can again be randomly chosen, giving a random Cantor set, with same d f . 3. Fractal overlap model of earthquake Overlapping fractals form a whole class of models to simulate earthquake dy- namics. These models are motivated by the observation that a fault surface, like a fractured surface (Chakrabarti and Benguigui, 1997), is a fractal object (Okubo and Aki, 1987; Scholz and Mandelbrot, 1989; Sahimi, 1993). Consequently a fault may be viewed as a pair of overlapping fractals. Fractional Brownian pro- files have been commonly used as models of fault surfaces (Brown and Scholz, 1985; Sahimi, 1993). In that case the dynamics of a fault is represented by one Brownian profile drifting on another and each intersection of the two profiles cor- responds to an earthquake (De Rubeis et al., 1996). However the simplest possible model of a fault − from the fractal point of view − was proposed by Chakrabarti and Stinchcombe (1999). This model is a schematic representation of a fault by a pair of dynamically overlapping Cantor sets. It is not realistic but, as a system of overlapping fractals, it has the essential feature. Since the Cantor set is a fractal with a simple construction procedure, it allows us to study in detail the statistics of the overlap of one fractal object on another. The two fractal overlap magnitude changes in time as one fractal moves over the other. The overlap (magnitude) time TWO-FRACTAL OVERLAP MODELS OF EARTHQUAKES 107 Figure 4. (a) Schematic representations of a portion of the rough surfaces of the earth’s crust and the supporting (moving) tectonic plate. (b) The one dimensional projection of the surfaces form Cantor sets of varying contacts or overlaps (s) as one surface slides over the other series can therefore be studied as a model time series of earthquake avalanche dynamics (Carlson et al., 1994). The statistics of overlaps between two fractals is not studied much yet, though their knowledge is often required in various physical contexts. It has been es- tablished recently that since the fractured surfaces have got well-characterized self-affine properties, the distribution of the elastic energies released during the slips between two fractal surfaces (earthquake events) may follow the overlap distribution of two self-similar fractal surfaces (Chakrabarti and Stinchcombe, 1999; Pradhan et al., 2003). Chakrabarti and Stinchcombe (1999) had shown analytically by renormalization group calculations that for regular fractal overlap (Cantor sets and carpets) the contact area distribution ρ(s) follows a simple power law decay: ρ(s) ∼ s−γ̃; γ̃ = 1. (3) Study of the time (t) variation of contact area (overlap) s(t) between two well- characterized fractals having the same fractal dimension as one fractal moves over the other with constant velocity, has revealed some features which can be utilized to predict the ‘large events’ (Pradhan et al., 2004). Bhattacharyya (2005) has recently studied this overlap distribution for two Cantor sets with periodic boundary conditions and each having dimension log 2/ log 3. It was shown, using exact counting, that if s ≡ 2n−k (n is the genera- tion number) then the probability ρ̃(s) to get an overlap s is given by a binomial distribution (Bhattacharyya, 2005) ρ̃(2n−k) = ( n n − k ) ( 1 3 )n−k (2 3 )k ∼ exp(−r2/n); r → 0, (4) 108 CHAKRABARTI, CHATTERJEE 24 25 26 27 28 0 100 200 300 400 500 600 s time Figure 5. For two Cantor sets (as shown in Fig. 4(b); n = 2), one moving uniformly over the other (periodic boundary conditions), the total measure of the shaded region contribute to the overlap s; the time variation of which is shown (for n = 4) where r2 = [ 3 2 ( 2 3 n − k )]2 . Expressing therefore r by log s near the maxima of ρ̃(s), one can again rewrite (4) as ˜̃ρ(s) ∼ exp ( − (log s) 2 n ) ; n → ∞. (5) Noting that ρ̃(s)d(log s) ∼ ρ(s)ds, we find ρ(s) ∼ s−γ̃, γ̃ = 1, as in (3) as the binomial or Gaussian part becomes a very weak function of s as n → ∞ (Bhattacharyya et al.,, 2005). It may be noted that this exponent value γ̃ = 1 is independent of the dimension of the Cantor sets considered (here log 2/ log 3) or for that matter, independent of the fractals employed. It also denotes the general validity of (3) even for disordered fractals, as observed numerically (Pradhan et al., 2003; Pradhan et al., 2004). Identifying the contact area or overlap s between the self-similar (fractal) crust and tectonic plate surfaces as the stored elastic energy E released during the slip, the distribution (3), of which a derivation is partly indicated here, reduces to the Gutenberg-Richter law (2) observed. TWO-FRACTAL OVERLAP MODELS OF EARTHQUAKES 109 4. Summary We introduce here the two-fractal overlap model of earthquake where the average distribution of the overlaps between the surfaces, as one fractal (here, Cantor set) moves over the other gives the Gutenberg-Richter like distribution (3), with γ̃ = 1 exactly in the model. We note that this is an exactly solvable model of earthquake dynamics and the result for the distributions compare favorably with the observations. Acknowledgements We are grateful to P. Bhattacharyya, P. Chaudhuri, M. K. Dey, S. Pradhan, P. Ray and R. B. Stinchcombe for collaborations in various stages of developing this model. References Bak P. (1997) How Nature Works, Oxford Univ. Press, Oxford. Bhattacharyya P. (2005) Physica A 348, 199. Bhattacharyya P., Chatterjee A., Chakrabarti B. K. (2005) arXiv:physics/0510038. Brown S. R., Scholz C. H. (1985) J. Geophys. Res. 90, 12575. Burridge R., Knopoff L. (1967) Bull. Seismol. Soc. Am. 57, 341. Carlson J. M., Langer J. S., Shaw B. E. (1994) Rev. Mod. Phys. 66, 657. Chakrabarti B. K., Benguigui L. G. (1997) Statistical Physics of Fracture and Breakdown in Disorder Systems Oxford Univ. Press, Oxford. Chakrabarti B. K., Stinchcombe R. B. (1999) Physica A 270, 27. De Rubeis V., Hallgass R., Loreto V., Paladin G., Pietronero L., Tosi P. (1996) Phys. Rev. Lett. 76, 2599. Gutenberg B., Richter C. F. (1944) Bull. Seismol. Soc. Am. 34, 185. Gutenberg B., Richter C. F. (1954) Seismicity of the Earth, Princeton Univ. Press, Princeton. Knopoff L. (2000) Proc. Natl. Acad. Sci. USA. 97, 11880. Kostrov B. V., Das S. (1990) Principles of Earthquake Source Mechanics, Cambridge Univ. Press, Cambridge. Okubo P. G., Aki K. (1987) J. Geophys. Res. 92, 345. Olami Z., Feder H. J. S., Christensen K. (1992) Phys. Rev. Lett. 68, 1244. Pradhan S., Chakrabarti B. K., Ray P., Dey M. K. (2003) Physica Scripta T106, 77. Pradhan S., Chaudhuri P., Chakrabarti B. K. (2004) Continuum Models and Discrete Systems, Bergman D. J., and Inan E. Eds, Nato Sc. Series, Kluwer Academic Publishers, Dordrecht, pp.245-250; cond-mat/0307735. Sahimi M. (1993) Rev. Mod. Phys. 65, 1393. Scholz C. H., Mandelbrot B. B. Eds. (1989) Fractals in Geophysics, Birkhaüser, Basel. Scholz C. H. (1990), The Mechanics of Earthquake and Faulting, Cambridge Univ. Press, Cambridge. 111 111–122. DYNAMICS OF STRUCTURES WITH BISTABLE LINKS ∗ Andrej Cherkaev1, Elena Cherkaev1 and Leonid Slepyan2 1University of Utah, Department of Mathematics, Salt Lake City, UT 84112, USA 2Tel Aviv Univ., Dept. of Solid Mechanics, Materials and Systems, 69978, Israel Abstract. The paper considers nonlinear structures with bistable links described by irreversible, piecewise linear constitutive relation: the force in the link is a nonmonotonic bistable function of elongation; the corresponding elastic energy is nonconvex. The transition from one stable state to the other is initiated when the force exceeds the threshold; the transition propagates along the chain and excites a complex system of waves. Mechanically, the bistable link may consist of two rods joined at the ends, the longer rod initially being inactive. This rod starts to resist when large enough strain damages the shorter rod. The transition wave is a sequence of breakages, it absorbs the energy of the loading force, transforms it into high frequency vibrations, and distributes the partial damage throughout the structure. In the same time, the partial damage does not lead to failure of the whole structure, because the initially inactive links are activated. The developed model of dynamics of cellular chains allow us to explicitly calculate the speed of the transition wave, conditions for its initiating, and estimate the energy of dissipation. The dissipation or absorption of the energy can be significantly increased in a structure characterized by a nonlinear discontinuous constitutive relation. The considered chain model reveals some phenomena typical for waves of failure or crushing in constructions and materials under collision, waves in a structure specially designed as a dynamic energy absorber and waves of phase transitions in artificial and natural passive and active systems. Key words: phase transition, local instabilities, bistable links, nonlinear dynamics, protective struc- tures 1. Introduction The paper is concerned with phase transition and dynamics of structures with non- monotonic bistable stress-strain constitutive relations described by a two-branched diagram. These systems have multiple equilibria that correspond to local energy minima. Such inertial systems describe various natural phenomena, bistable-link structures are found in many engineering applications, specially designed, these structures are capable of absorption and dissipation of large amounts of energy. The bistable model describes an irreversible transition such as the damage prop- agation in foams, structural materials, and constructions. The simplest example ∗ Supported by ARO grant 41363-MA, NSF grant DMS-0072717, and The Israel Science Foundation grant 1155/04. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 112 CHERKAEV, CHERKAEV, SLEPYAN is the falling domino. Another example is metallic open-cell foams. Even larger scales are used to describe demolition of a multilevel construction. Here, the tran- sition means crashing of some links and compactification. The transition wave is a result of multiple sequential breaks of its elements (cells); after the transition, the cells regain their strength in a new state. Once initiated, the wave propagates through the structure which would be stable under an equal static load. The novel area of application of the bi-stable structures is design of materials absorbing large amounts of energy. While the theoretical limit for the energy absorbed in a material is the energy corresponding to the melting point, real structures are destroyed by a tiny fraction of this energy due to material’s in- stabilities and an uneven distribution of the stresses throughout the structure. The goal of design is to find a structure that absorbs maximal energy without rupture or breakage. The effect is achieved by using bistable breakable links [(Cherkaev et al., 2005; Slepyan et al., 2005)]. The key of the approach is a special structure or a geometric principle of assembling solid elements in a structure. The superior properties of these energy absorbing structures are due to their special morphol- ogy. The structures possess several length scales and their averaged constitutive relations are non-monotonic due to special spatial configuration of their elements. These “waiting element structures” allow propagation of waves of “partial dam- age” which delocalize and eventually disperse the energy of an impact. Bistable-bond chain model was considered in a number of works (see [(Slepyan and Troyankina, 1984, 1988); (Slepyan, 2000, 2001, 2002), (Ngan and Truski- novsky, 1999), (Puglisi and Truskinovsky, 2000), (Balk et al., 2001a, 2001b)], [(Charlotte and Truskinovsky, 2002)], [(Cherkaev et al., 2005; Slepyan et al., 2005)]. In the bistable chain, the transition wave is accompanied by a pronounced dissipation due to the excitation of structure-associated high-frequency waves. The more complex problem of breakable two-dimensional structure with wait- ing links was numerically investigated in [(Cherkaev and Zhornitskaya, 2004)]. The bistable chains from elastic-plastic material were studied in [(Cherkaev et al., 2005a)]. Transition waves in two-dimensional lattices was analyzed in [(Slepyan and Ayzenberg-Stepanenko, 2004)]. 2. Analytic description of the phase transition The dynamics of the chain is described by the equation Müm(t) = T (um+1 − um) − T (um − um−1), m = −∞, . . . ,∞, (1) where um is the displacement of the m-th mass, t is time, M is the mass of the particle. A non-monotonic (and discontinuous) function T is the tensile force in the locally unstable link. One of two-branch constitutive diagrams that permit analytical solution, is shown in Fig. 1(a). DYNAMICS OF BISTABLE STRUCTURES 113 Equivalent problem Consider an infinite bistable chain with an irreversible force– elongation relation shown in Fig. 1(a). The transition to the second branch of the diagram occurs at the moment t = t∗ when the elongation q first time reaches the critical value, q = q∗. This nonlinear dependence can be modelled using an equivalent problem which considers the intact linear chain with the bonds corresponding to the first branch of the diagram. In the equivalent problem, at the moment t = t∗ (different for different bonds) a pair of external forces ∓P∗ is instantly applied to the left and right masses connected by this bond. These forces compensate for the difference between the branches P = P∗+(1−γ)µ(q−q∗), (t > t∗) . Here µ is the stiffness of the bond. Thus, the parallel-branch problem can be considered as a linear one with these additionally applied external forces which model the nonlinearity. In a special case γ = 1, see Fig. 1(a), the value of the external forces is a constant. In this case, the problem is simplified, since the right-hand side of the first-branch equation is known in advance. In this case, if all the bonds to the left of a mass correspond to the second branch, while all the bonds to the right correspond to the first branch, the resulting external force is only applied to this mass. The other external forces compensate each other. In a general case, γ � 1, the external force depends on the current elongation of the bond, and we have different equations from different sides of the transition wave front. Below we outline main steps in solving the equivalent problem. Steady-state formulation Assume that the transition wave propagates to the right with a constant speed, v > 0. The nondimensional time interval between the tran- sition of neighboring bonds is equal to a/v, where a is the distance between two neighboring masses at rest. We assume that u̇m+1(t + a/v) = u̇m(t). The motion of a mass is a function of a continuous variable, η = vt− x, where x is the continuous coordinate, and it is steady for any value of η. In the ‘steady-state’ regime we have: u̇m(t) = −v du(η) dη , η = vt − am . (2) The transition wave speed v is unknown. Its value and conditions required for initiating the transition are determined by the solution of the problem. We are looking for v as a function of the intensity of an external source of energy which causes the transition wave. However, it is more convenient to consider an implicit problem and find the source intensity as a function of the speed v. The resulting piece-wise linear problem is solved analytically and/or numerically. Analytical solution Analytical solution is derived using the Fourier transform. This solution is used to estimate the total dissipation rate. The steps are as follows. We consider nondimensional equation for the bistable, parallel branch chain with 114 CHERKAEV, CHERKAEV, SLEPYAN an introduced relation qm(t) = um(t) − um−1(t) and the external forces in the right- hand side: üm(t) − qm+1(t) + qm(t) = 0 if (1, 1) or (2, 2) −P̂∗ if (1, 2) P̂∗ if (2, 1) (3) where the pairs (1, 1) . . . , (2, 2) denote the state (the first branch (1) or the second one (2)) of the bonds connecting the mth mass with the left and the right neighbors, respectively. The steady-state version of (1) is the following v2 d2q(η) dη2 +2q(η)−q(η−1)−q(η+1) = P∗[2H(−η)−H(−η−1)−H(−η+1)] , (4) where H is the Heaviside function, and we assume M = 1. To complete the formulation we introduce conditions at infinity. We consider a semi-infinite chain, m = −N,−N + 1, ..., assuming number N to be large. We consider two types of excitation conditions which lead to the same solution. (1) Assume that a constant tension force P directed to the left, is applied to the end mass, m = −N, at the moment of time t = −N/v. (2) Alternatively, this mass is forced to move with a constant speed, u̇−N = −w. We also assume that there is no energy flux from infinity on the right, as m → ∞. So that, only the waves with group velocities exceeding the transition front velocity v, can exist ahead of the transition front. Because of linearity of the equivalent problem, the strain q(η) can be rep- resented as a sum of a homogeneous solution q0 of (4) (this corresponds to a zero wavenumber incident wave caused by an external action at η = −∞) and an inhomogeneous solution q(η) (that corresponds to waves excited by the right-hand side forces). The homogeneous solution corresponds to the energy brought from infinity; the inhomogeneous solution constructed under the causality principle, describes the energy carried from the transition front to ±∞. We consider the homogeneous part of the total solution, and calculate the source at infinity as a force acting on the end mass of the chain or the velocity of the the end mass. This is the sought relation between the source intensity and the transition wave speed expressed in terms of parameters of the bistable links. The incident wave moves to the right with unit speed which is larger than the speed of the transition front. This wave is characterized by q = q0, u̇ = −q0. To distinguish the solution of the inhomogeneous equation (4) we denote the corresponding elongations by q(η). Using the Fourier transform qF(k) = ∫ ∞ −∞ q(η) exp(ikη) dη (5) DYNAMICS OF BISTABLE STRUCTURES 115 on η as a continuous variable we obtain qF(k) = 2P∗(1 − cos k) ikh(k) , h(k) = (0 + ikv)2 + 2(1 − cos k) (6) Here 0 + ikv = lims→+0(s + ikv). The integrant contains poles as zeros of h(k) (the finite number of zeros and their values depend on the speed, v). According to the causality principle for a steady-state solution (see book of Slepyan [(Slepyan, 2002)])), these poles are placed outside the real axis if s > 0 approaching the axis in the limit, s → 0. The rule defines the half-space - positive or negative, - containing the pole. In turn, this defines sign of the half-residual at the pole. The integral is thus represented as a sum of the half-residuals and a Cauchy principal value of the integral q(η) = − P∗v 2(1 − v2) + P∗ n∑ ν=0 (Q2ν+1 cos(h2ν+1η) − Q2ν cos(h2νη)) − 2P∗ π V.p. ∫ ∞ 0 1 − cos k k[2(1 − cos k) − k2v2] sin(kη) dk , (7) where Q2ν+1 = v2h2ν+1 2(sin h2ν+1 − v2h2ν+1) , Q2ν = v2h2ν 2(sin h2ν − v2h2ν) , Q0 = 0 , (8) and h2ν, h2ν+1 are real zeros of the function h(k). The obtained solution depends on the speed v which is still unknown. It can be determined using the condition at η → −∞ (m = −N) that the applied force P equal to the uniform part of the tensile force T , or that a given uniform part of the particle velocity −w. The corresponding relations depend only on the ratios P0 = P/q∗, P0∗ = P∗/q∗, w0 = w/q∗ , but not on the parameters P, P∗ and w. In these normalized terms, the relation between the speed v and the applied force P become 1 − 2v2 2(1 − v2) P0∗ + P 0 ∗ n∑ ν=0 (Q2ν+1 − Q2ν) = 1 − P0 . (9) The relation between the speed v and the initial velocity −w is similar. Relation (9) gives the equation for finding the transition wave speed v. It contains two nondimensional parameters; parameter P0∗ defines the material properties, and P0 (or w0) describes the level of external excitation. Analytically calculated speed v of the transition wave for a bi-stable model was compared with numerical estimates in [(Slepyan et al., 2005)]. This compar- ison is shown in Figure 1(b). 116 CHERKAEV, CHERKAEV, SLEPYAN q * P * 1 2 q T 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 V x x x x x x x P 0 (a) Stress-strain relation (b) Speed of transition wave Figure 1. (a) The force–elongation dependence for the parallel-branch piecewise linear diagram: 1. T = µq (the first branch) and 2. T = µq − P∗ (the second branch). (b) Analytically calculated speed of the transition wave compared with numerical estimates shown by stars The analytical technique is presented in details in [(Slepyan, 2002; Slepyan et al., 2005)]. The last work also deals with a more complicated case of non-parallel branches when γ � 1. In this case, the Wiener-Hopf technique is exploited after taking the Fourier transform of q(η). An additional analysis is required to account for the variable elastic moduli, and for linkage of elements in a two-dimensional triangular lattice. It is possible to describe plane and crack-like waves of transi- tion in the lattices using similar formalism [(Slepyan and Ayzenberg-Stepanenko, 2004)]. 3. Dynamic homogenization Description of the effective behavior of a bistable chain cannot be derived from Gibbs principle of minimum of energy, it requires corresponding “dynamic ho- mogenization”. We are interested in the effective elongation of the chain due to a force applied to its end. The system is conservative, but a part of the work of the applied force is transformed to the kinetic energy of the phase transition. In open systems, a part of the energy is radiated by high frequency oscillations and waves. The adequate homogenization model should account for these energy transformations, which leads to the homogenized equation of the type: L(u) = ρü − ∇ · σ + Θ(u, u̇,∇u), σ = ∂ ∂q HdF(q) (10) Here the “pseudo-temperature” Θ(u, u̇,∇u) accounts for the energy stored in ex- cited high-frequency waves which become invisible in the average description of the lower frequency part, and HdF is an analog of the quasiconvex envelope of DYNAMICS OF BISTABLE STRUCTURES 117 the nonconvex energy F: the dynamical average. The description can be obtained as a Γ-limit of the atomistic model taking into account the excited high-frequency waves. This high frequency part cannot be neglected because the kinetic energy of the system is bounded below from zero due to dynamical instabilities. 4. Numerical simulations Model of bistable bonds The discussed irreversible tensile force can be mod- elled using a damage function c(t) which introduces dependence on prehistory [(Cherkaev et al., 2005)]. The damage parameter c instantly jumps from zero to one at the first time moment t′ when the elongation q exceeds the critical value q∗, c(t) = { 0 if maxt′∈[0,t) q(t′) < q∗ 1 otherwise (11) The non-monotonicity of the force-elongation dependence corresponds to a non- convex strain energy of the bond. The state of the bond under a time-dependent loading is described as T (q, c) = [1 − c(t)]T1(q) + c(t)T2(q) (12) where T1(q) is the dependence for the first stable region (the first branch) and T2(q) is that for the second branch which includes the unloading. The waiting-element structure A bistable link can be mechanically realized as a “waiting element” [(Cherkaev and Slepyan, 1995)]. A link, consists of two rods, one straight and one slightly curved, joined by their ends. We assume that the links are made of an elastic-brittle material. Using the introduced damage parameter c(t) we express the stress-strain relation in a rod as T (q, c) = (1 − c(t))H(q) µ q (13) where H(q) is the Heaviside function. The force-elongation dependence in the waiting rod is similar to the one in the basic link, but is shifted by τ. We assume that the waiting link starts to resist when the basic link is broken. The whole link is bistable, the tensile force T is the sum, T (q, cb, cw) = Tb(q, cb) + Tw(q, cw) where Tb, Tw are the tensile forces in basic and waiting links described by (13) with the stiffness moduli µb and µw and the damage parameters cb and cw corresponding to the basic and waiting rods. The design parameters of the bistable link include: the critical elongations of the basic and waiting rods, q∗ and q∗∗, the shift parameter τ (the gap between the branches of the stress-strain diagram is equal to τ− q∗) and the moduli µb and µw. In the following calculation, we use the representation µb = αµ, µw = (1 − α)µ 118 CHERKAEV, CHERKAEV, SLEPYAN where α is the relative thickness of the basic rod and µ is the stiffness modulus of the total cross-section of the element. In numerical simulations shown below, we assume that the waiting rod is stronger than the basic rod, or α < 0.5. Equations The dynamics of the chain is described by a system of difference– differential equations with respect to the displacements um of the masses and elongations qm = um − um−1. The system has the form Müm + γu̇m = T (qm+1, c m+1 b , c m+1 w ) − T (qm, cmb , c m w) . (14) Here m is the index of the mass, γ is the coefficient of viscosity introduced to stabilize numerical simulations, and cmb and c m w are the damage parameters for the basic and waiting rods in the mth link. We consider the following model: A rested chain of N masses M bonded by the waiting-element links is impacted by a large mass M0 moving with initial velocity v0. In simulations, we increase the mass M0 until the chain is broken. The goal is to optimize the parameters of the links to increase the amount of energy absorbed by the chain before breaking or to increase the value of the loading mass that can be withstood by the chain. The dynamics of the chain attached to a support, is described by the equations (14), condition for the mass in the root u0 = 0, the equation for the loading mass at m = N M0üN = −T (qN , cNb , c N w ), and the equations (11) for the damage indicators cNb and c N w where i = 1, . . . , N. The initial conditions are um = ma, m = 1, . . . , N (15) u̇m = 0, m = 1, . . . , N − 1, u̇N = v0 (16) In the numerical simulations, we use a chain of 32 elements. The initial speed v0 of the loading mass is 7% of the speed of the sound in material (the long wave speed in the undamaged chain c = a √ µb/M). The additional elongation τ of the waiting link is larger than the limiting elongation q∗, so that the basic link is completely broken before the waiting link is activated. Results of numerical simulations The trajectories of the nodes in a chain of regular and of bi-stable elements are shown in Figures 2(a) - 2(d). Figures 2(a) and 2(b) show trajectories of the knots of the chain without waiting elements. First of them shows the results of simulation of the chain impacted by a mass M0 = 100. The chain sustains the impact. However, being impacted by the mass M0 = 125, the chain is broken, see 2(b). Figures (c) and (d) show the bi-stable waiting element chain made of the same amount of material. The chain withstands DYNAMICS OF BISTABLE STRUCTURES 119 much higher impact. Figure 2(c) shows the trajectories of the knots in the chain with waiting elements, α = 0.25, τ = 0.5, impacted by the mass M0 = 375: The chain sustains the impact. One can observe the elastic wave and two waves of a partial damage originated at both ends of the chain where the magnitude of the elastic wave is maximal. One can notice intensive oscillations of the masses after all links are partially broken. Trajectories of the knots in the chain with the same waiting elements and with a small dissipation, γ = 0.2, impacted by the mass M0 = 600 are shown in Figure 2(d). The chain sustains the impact. The dissipation has little influence on the waves propagation, but suppresses subsequent chaotic oscillations of the masses in the relaxed chain. The computer experiments show that structures with waiting links sustain the loading masses 5 - 6 times greater than the conventional structures. We stress that this result is achieved using the same amount of the same material, the only difference is the morphology of the structure. Modification of the model: Dissipation The dynamic of the chain with partially broken links indicates that the masses are involved in a chaotic motion after the basic links are broken. It is exactly the chaotic motion phase that leads to the final damage of the chain due to excessive elongation of a spring. Observing this phenomenon, we may question the adequacy of the model that does not take into account a small dissipation that is presented in a mechanical system; this dissipation would reduce the chaotic motion, especially in a long time range. We introduce a small dissipation into the numerical scheme; this practically does not influence the initial wave of the damage but reduces the intensity of afterward chaotic oscillations. The model with small dissipation gives even larger increase in the energy absorption capacity of the chain that can now sustain the twice larger loading mass. Notice that a small dissipation would have practically no influence on the behavior of chains made of a stable material, if the excitation is slow. However, the chain of unstable materials excite intensive fast-propagating waves no matter how slow is the impact, see [( Balk et al., 2001b)]. The intensity of the transition wave is determined not by the external excitation but by the inner structure of bi-stable elements. The wave is excited when the elongation of a bi- stable element is larger than the critical value and even a slow impact creates intensive waves which lead to significant energy dissipation. Large dissipation The increase of the dissipation coefficient allows for more orderly transition since the propagating and reflected waves are suppressed at a distance. We observe the wave of phase transition originated at the point of impact that propagate to the root of the chain. However, a second to the root link partially breaks when the weaken viscous-elastic wave reach the far end and reflects. 120 CHERKAEV, CHERKAEV, SLEPYAN 10 20 30 40 50 60 70 80 90 100 −1 0 1 2 3 4 5 6 7 8 9 10 Conventional chain: M = 100 P os iti on o f t he m as se s Time 0 10 20 30 40 50 60 70 80 0 2 4 6 8 10 12 Conventional chain: Loading Mass M = 125 Time P os iti on o f t he m as se s (a) (b) 0 100 200 300 400 500 600 0 2 4 6 8 10 12 14 16 18 20 Chain with waiting elements: Loading Mass M =375, No Dissipation Time P os iti on o f t he m as se s 0 100 200 300 400 500 600 0 5 10 15 20 25 Chain with waiting elements: M=600 Time P os iti on o f t he m as se s (c) (d) Figure 2. (a) The motion of the knots of the chain without waiting elements impacted by a mass M0 = 100: The chain sustains the impact. (b) The motion of knots in the same chain impacted by a large mass M0 = 125: The chain is broken. (c) The motion of the knots in the chain with waiting elements, α = 0.25, τ = 0.5, impacted by the mass M0 = 375: The chain sustains the impact. Observe the elastic wave and two waves of a partial damage originated at both ends of the chain where the magnitude of the elastic wave is maximal. (d) Trajectories of the knots in the chain with the same waiting elements and with a small dissipation, γ = 0.2, impacted by the mass M0 = 600: The chain sustains the impact 5. Conclusion The results show a dramatic increase of the strength of the chain with waiting elements compared to the conventional design. The dissipation further increases the effect. DYNAMICS OF BISTABLE STRUCTURES 121 0 1000 2000 3000 4000 5000 6000 7000 8000 −5 0 5 10 15 20 25 30 35 40 45 Figure 3. Trajectories of the knots in the chain with waiting elements, α = 0.3, and with a larger dissipation, γ = 1, impacted by a slow moving mass M0 = 250, 000 with the initial velocity 0.01. The chain sustains the impact; the dissipation suppresses the oscillations. Observe the wave of par- tial breakage that propagate starting from the impact point. The closest to the root link experiences a partial damage because of the viscous-elastic wave References Balk A. M., Cherkaev A. V., Slepyan L. I. (2001a) Dynamics of chains with non-monotone stress- strain relations. I. Model and numerical experiments, J. Mech. Phys. Solids 49 131-148. Balk A. M., Cherkaev A. V., Slepyan L. I. (2001b) Dynamics of chains with non-monotone stress- strain relations. II. Nonlinear waves and waves of phase transition, J. Mech. Phys. Solids 49 149-171. Cherkaev A., Slepyan L. (1995) Waiting element structures and stability under extension, Int. J. Damage Mech. 4 58-82. Cherkaev A., Zhornitskaya L. (2004) Dynamics of damage in two-dimensional structures with waiting links, In: Asymptotics, Singularities and Homogenisation in Problems of Mechanics, A.B. Movchan editor 273-284, Kluwer. Cherkaev A., Cherkaev E., Slepyan, L. (2005) Transition waves in bistable structures I: Delocaliza- tion of damage, J. Mech. Phys. Solids 53 383-405. Cherkaev A., Vinogradov V., Leelavanichkul S. (2005a) The waves of damage in elastic-plastic lattices with waiting links: design and simulation, Mechanics of Materials accepted. Charlotte M., Truskinovsky L. (2002) Linear chains with a hyper-pre-stress, J. Mech. Phys. Solids 50 217-251. Ngan S. C., Truskinovsky L. (1999) Thermal trapping and kinetics of martensitic phase boundaries, J. Mech. Phys. Solids 47 141-172. Puglisi G., Truskinovsky L. (2000) Mechanics of a discrete chain with bi-stable elements, J. Mech. Phys. Solids 48 1-27. Slepyan L. I., Troyankina L. V. (1984) Fracture wave in a chain structure, J. Appl. Mech. Techn. Phys. 25 921-927. Slepyan L. I., Troyankina L. V. (1988) Impact waves in a nonlinear chain, In: ‘Strength and Visco- plasticity’ (in Russian), Journal Vol 301-305, Nauka. 122 CHERKAEV, CHERKAEV, SLEPYAN Slepyan L. I. (2000) Dynamic factor in impact, phase transition and fracture, J. Mech. Phys. Solids 48 931-964. Slepyan L. I. (2001) Feeding and dissipative waves in fracture and phase transition. II. Phase- transition Waves, J. Mech. Phys. Solids 49 513-550. Slepyan L. I. (2002) Models and Phenomena in Fracture Mechanics, Springer-Verlag. Slepyan L. I., Ayzenberg-Stepanenko M. V. (2004) Localized transition waves in bistable-bond lattices, J. Mech. Phys. Solids 52 1447-1479. Slepyan L., Cherkaev A., Cherkaev E. (2005) Transition waves in bistable structures II: Analytical solution, wave speed, and energy dissipation, J. Mech. Phys. Solids 53 407-436. 123 123–136. REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU (MO) BRIDGE AND RECORDED EARTHQUAKE RESPONSE Mehmet Çelebi USGS, Menlo Park, 94025, Ca, USA Abstract. This paper introduces the state of the art, real-time and broad-band seismic monitoring network implemented for the 1206 m [3956 ft] long, cable-stayed Bill Emerson Memorial Bridge in Cape Girardeau (MO), a new Mississippi River crossing, approximately 80 km from the epicentral region of the 1811-1812 New Madrid earthquakes. Design of the bridge accounted for the possi- bility of a strong earthquake (magnitude 7.5 or greater) during the design life of the bridge. The monitoring network consists of a superstructure and two free-field arrays and comprises a total of 84 channels of accelerometers deployed on the superstructure, pier foundations and free-field in the vicinity of the bridge. The paper also introduces the high quality response data obtained from the network. Such data is aimed to be used by the owner, researchers and engineers to (1) assess the performance of the bridge, (2) check design parameters, including the comparison of dynamic char- acteristics with actual response, and (3) better design future similar bridges. Preliminary analyses of low-amplitude ambient vibration data and that from a small earthquake reveal specific response characteristics of this new bridge and the free-field in its proximity. There is coherent tower-cable- deck interaction that sometimes results in amplified ambient motions. Also, while the motions at the lowest (tri-axial) downhole accelerometers on both MO and IL sides are practically free-from any feedback from the bridge, the motions at the middle downhole and surface accelerometers are significantly influenced by amplified ambient motions of the bridge. Key words: earthquake, monitoring, real-time, response, ambient, frequency 1. Introduction In seismically active regions, acquisition of structural response data during earth- quakes is essential to evaluate current design practices and develop new method- ologies for analysis, design, repair and retrofitting of earthquake resistant struc- tural systems, including lifelines such as bridges. This is particularly true for urban environments in seismically active regions. The New Madrid area, where the great earthquakes of 1811-1812 occurred, is a highly active seismic active region requiring earthquake hazard mitigation programs, including those related to investigation of strong shaking of structures and the potential for ground failures in the vicinity of structures (Nuttli, 1974; Woodward-Clyde Consultants, 1994). The Bill Emerson Bridge (here in after, Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 124 ÇELEBİ the Cape Girardeau Bridge), in service since December 2003, is located approx- imately 80 km due north of New Madrid, Missouri. Design of the bridge (a) accounted for the possibility of a strong earthquake (magnitude 7.5 or greater) dur- ing the design life of the bridge and, as a result (b) was based on design response spectrum anchored to a zero-period acceleration (ZPA) of 0.36 g with a 10% prob- ability of being exceeded in 250 years (Woodward-Clyde, 1994). Therefore, the Federal Highway Administration (FHWA), Missouri Department of Transporta- tion (MoDOT), the Multi-disciplinary Center for Earthquake Engineering Re- search (MCEER), and the United States Geological Survey (USGS) collaborated in developing the seismic monitoring system for the bridge. For seismic monitoring of a bridge, three main categories in recording motions are sought: a. instrumentation of the superstructure and pier foundations to capture and define (i) overall motion of the cable-stayed bridge, and motions of the (ii) two towers, to assess their translational and torsional behavior - relative to the caissons and deck levels and (iii) the deck, to assess the fundamental and higher mode translational (longitudinal, transverse and vertical) and tor- sional components, and (iv) extreme ends of the bridge and intermediate pier locations to provide data for the translational, torsional, and rocking soil- structure interaction (SSI) at the foundation levels as well as the horizontal and vertical spatial variation of ground motion. b. instrumentation of the free-field in the vicinity of the structure including those related to downhole measurements and horizontal spatial arrays to assess the differential motions at the piers of the long span structure. c. ground failure arrays in the vicinity of the structure (outside the scope of this network). The general schematic of the bridge, shown in Figure 1, illustrates (a) overall lon- gitudinal dimensions of the bridge and key elevations, and (b) key sensor locations alongside the bridge. The objective of this paper is to introduce details of the extensive, state-of- the-art, real-time and broad-band seismic monitoring network deployed on and in the vicinity of the new Cape Girardeau Bridge. The real-time capability provides three basic advantages; (i) in addition to recording strong-motion events, it is possible to selectively record continuous real-time low-amplitude response data, on demand, with relative ease, (ii) use of the near real-time information can help make informed decisions related to the response and performance of the bridge (e.g. “health monitoring”) and (iii) maintenance of the system will be readily and easily enhanced as any malfunction of the sensors and related hardware will be detected via the real-time streamed information. REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 125 Figure 1. The figure illustrates (a) overall longitudinal dimensions of the bridge and key elevations, and (b) key sensor locations alongside the bridge, (c) general schematic of the seismic monitoring system with antennas for wireless communication of data between different locations. to and from the Data Acquisition Blocks 1 and 2 housed in Pier 2 and Pier 3 towers, and to the off-structure Central Recording System and (d) free-field arrays on both Missouri and Illinois sides.(Note: Shear wave velocities shown in fps. [1fps=0.3048m/s]) The paper also presents preliminary sample analyses of low-amplitude data including motions (at ∼ 10−4 − 10−3 g levels) caused by a small earthquake that occurred 175 km away and ambient vibrations caused by tower-cable-deck interactions sources of which maybe traffic or wind. For example, site response effects or soil-structure interaction or tracing of motions from the free-field to the top of tower or reverse phenomenon caused by excitations at the free-field caused by ambient vibrations of the structure (tower-cable-deck interaction) can be clearly observed and a variety of computations can be made to provide insight into vibrational behavior of the bridge. However, in this introductory paper, while sample analyses using data from the network are presented, detailed analyses of each phenomena occurring at particular location of the bridge or the free-field sites (e.g. site response) are not included due to space constraints. Earlier publication (Çelebi and others, 2004) related to this bridge did not include data analyses since data streaming started in March 2005. 126 ÇELEBİ 2. Instrumentation The detailed schematic of the bridge in Figure 2 shows key locations and orien- tations of the 84 channels of accelerometers deployed on the bridge and the two free-field arrays. Identifier two-character location codes are important for data management and for referencing motions this paper. All horizontal sensors ori- ented transverse (to the bridge) are more closely aligned to NS, and longitudinal sensors are more closely aligned to EW. Therefore, in this paper, directions of components of horizontal motions are hereafter referred to as NS and EW instead of lateral and longitudinal. Figure 2. General schematic illustrating longitudinal dimensions of Cape Girardeau Bridge. Also shown are the locations and orientations of the accelerometers - both for the bridge and its piers and the free-field arrays on Missouri and Illinois sides. Location identifier codes (e.g. T1 for tower top location 1 at south end of tower top at Pier 2) are used for data management. Each arrow indicates one channel of accelerometer and its orientation In addition to 84 channels of Kinemetrics EpiSensor1 accelerometers, addi- tional monitoring hardware consists of Q3301 digitizers, and data concentrator 1 Citing commercial hardware throughout this manuscript does not imply endorsement of vendors or their products. REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 127 and mass storage devices (herein called Balers2 ) with wireless communication units (Figure 1). In each one of the hubs at Piers 2 and 3, combination of Q330 digitizers/recorder and Baler45 units constitute a multi-channel Data Acquisition Block (DAQ Block) that collects the analog signals from the accelerometers lo- cated throughout the bridge and then transmits the digitized data to the Central Recording System (CRS) using wireless radio transmittal via the antennas shown in Figure 1. The CRS merges the streamed data from the DAQs and records at location (on a pre-planned manner using a trigger algorithm to produce file events) and broadcasts the streamed data out using standard TCP/IP communications protocol. From CRS, all 84 channels of data are now being transmitted to Incor- porated Research Institutions for Seismology (IRIS) via the Antelope3 software at the Central Recording System of the Cape Girardeau Bridge Seismic Monitoring System. From IRIS, data are disseminated online to the user community. Figure 2 also shows the general accelerometer deployment scheme for deck locations at the centerline (CL) and at the locations L1, L2, R1 and R2 of the cable-stayed deck. The exact sensor locations at these four locations are based on results of mathematical modal analyses (S. Dyke, written communication, 2001). It is noted herein that deck instrumentation at Pier 2 and 3 is on deck level at elevation 124.5 m (408.4ft). At both Piers 2 and 3, the deck is supported by the cables and does not rest on the piers. There are pot bearings where the edge beams rest on the pier cap. Therefore, there is a separate set of sensors at pier elevation 121.3 m (398 ft) (Figure 2). One of the two permanent surface and downhole free-field arrays is deployed at the Missouri [MO] side and the second at the Illinois [IL] side of the Missis- sippi River. The MO free-field array is approximately 100 m (∼ 399 ft) south of Bent 1 and the IL array is approximately 300m (∼ 900 ft) south-east of Pier 15 (Figures 1 and 2). Geotechnical characteristics of the boreholes that house the tri- axial downhole accelerometers at defined depths from the surface are qualitatively and quantitatively shown in Figures 1 and 2. On the MO side, the two downhole accelerometers are at 9ft (∼3m) and 23 ft. (∼7.5m) from the surface. On the IL side, the two downhole accelerometers are at 47 ft (∼14.3 m) and 94 ft. (∼30m) from the surface. These two free-field arrays, intended to be without any feedback from the structure, are essential in providing the input ground motions that may be used as a surrogate for the various piers of the bridge and also for convolu- tion and deconvolution studies of the free-field ground motion. However, as dis- cussed later in this paper, the data acquired to date indicates that while the lowest 2 Baler is a data concentrator and mass storage unit. These units gather data and pass it to the next location as they are instructed to do so. In essence, a Baler serves as the brain and router of the data acquisition system. Baler-14 works with a single Q330 unit. Baler45 works with multiple number of Q330 units. In the case of Cape Girardeau instrumentation scheme, since there are multiple Q330 units, then only Baler45 units are used. 3 The Antelope software is licensed and installed at the site server by Kinemetrics. Citing commercial names does not imply endorsement of the vendor. 128 ÇELEBİ (tri-axial) downhole accelerometers on both MO and IL sides (locations F3 and F6 in Figures 1 and 2) depict that there are practically free-from any feedback of motions of the bridge, the middle downhole and surface accelerometers are not. 3. Preliminary sample data analyses Preliminary analyses are performed for a 267 second interval of data that includes the earthquake (M=4.1) that occurred at 12:37:32 (UTC) on May 1, 2005 but, due to space limitations, only sample results are discussed. More detailed results are provided elsewhere (Çelebi, 2005). The epicenter, 175 km from the bridge, was located at 6 km (4 miles) SSE (162◦) from Manila, AR (with epicental coordinates 35.830◦N, 90.150◦W) . The hypocentral depth was estimated to be 10 km (6 miles). Figure 3 shows full 267 seconds of only the NS component of acceleration data from MO and IL free-field arrays. The length of the record neither repre- sents neither the duration of the earthquake nor the strong shaking caused by the earthquake. Such long duration shaking at this site from a small earthquake that occurred 175 km away is caused by crustal reverberations and surface waves (Hanks, 2005) and the network of broad-band instruments deliver strong signals even at small amplitudes (∼10-4 of g levels). The figure depicts (a) earthquake record of May 1, 2005 is between 30-130 seconds of the 267, this length of the ”earthquake” response caused by surface waves, (b) the motions at locations of the middle downhole accelerometers (F2 and F5) and of surface accelerometers (F1 and F4) are significantly amplified and are affected by (ambient) motions of the bridge when compared to the motions at lowest downhole accelerometer locations (F3 and F6), (c) when compared with the low-amplitude earthquake motions, the feedback from the bridge ambient motions are significant and occurs several times (exemplified by spikes) in the 267 second long record and (d) using the approximately 78-seconds of the earthquake portion of the 267 length of the data (approximately between 30-110 seconds), it is possible, as done later in the paper, to infer qualitative and quantitative site amplification that occurs at the free- field locations (even from such small amplitude motions caused by the distant small earthquake). Figure 4 shows NS time-history plots of motions at center of the deck (C2) , tower tops (T2 and T4) mid-tower locations (M2 and M4) side span deck (L2 and R6) and other pier locations and the free-field arrays. Compared to the ambient motions at deck center and other key structural locations (C2, T2, T4, L2, R6), the free-field motions due to the May 1, 2005 earthquake are much smaller. The total 267-second record reveals some important behavior of the deck and tower tops which are connected by cables under tension. Figures 5 shows dis- placement time histories in the EW (longitudinal) direction of the bridge at the north end of tower top locations (T2 at Pier 2 and T4 at Pier 3) and the north REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 129 Figure 3. 267 seconds of free-field (NS component) acceleration data from surface and downhole deploy- ments at both Missouri and Illinois sides. The earthquake of May 1, 2005 is best identified by the lower downhole accelerometers as they are the least affected by site response effects clearly seen at surface, and also by ambient vibration effects from the bridge reflected back to the surface and mid-downhole accelerometer locations. Short-duration spikes or bursts of energy, such as those preceding the P-Wave from the earthquake are likely due to traffic on the bridge end deck center location, C2. At approximately 0.33 Hz, vertical motions at C2 and EW motions at T2 are highly coherent and approximately 180◦ out of phase while vertical motions at C2 and EW motions at T4 are highly coherent but are in phase (0◦ phase angle) indicating that while deck center is going up, T4 is in phase and therefore moving eastward while T2 is 180◦ out of phase and therefore moving westward. Figure 6 shows similar plots for the (a) east-west motions at T2 and T4 which are perfectly coherent and (180◦) out of phase at 0.33 HZ and (b) north and south ends of deck center that are not vibrating perfectly in phase even though there is perfect coherency between the two. Thus, this implies that the amplified ambient motions of key locations of the structure (deck center, top of towers) are possibly due to time-variant cable action - that is, stored tensile energy in the cables vary and increase/decrease according to the displacements of the deck center and tower tops (as well as other locations where cables connect the deck to the towers). However, it is uncertain whether all cables on either north end or south end, and in particular those that are symmetrical with respect to center of the deck, remain in proportionally compatible tension. Thus, the cables store and release energy continuously. The amplified motions of the deck and tower tops may be caused by the cable actions. While detailed modes are not identified from the analyses of the low-amplitude data recorded, it is noted that the fundamental vertical mode of the bridge is approximately 0.33 Hz - similar to the 0.29 Hz frequency computed by modal analysis performed by Dyke (2001). Amplified 130 ÇELEBİ Figure 4. The (NS) motions at free-field locations are shown relative to the piers and deck center and top of towers. The amplified motions of the structure between 100-150 seconds into the record are anomalies caused by cable vibrations of the structure and not by the much smaller earthquake between 30-100 seconds into the record Figure 5. Displacement time-histories, cross-spectrum, coherency and phase-angle plots. [left]: EW motions between Tower Top at Pier 2 (Location T2) and vertical motions at center of deck, and [right]: EW motions between Tower Top at Pier 3 (Location T4) and vertical motions at center of deck ambient motions of the bridge caused by cable vibrations or wind actions occur quite often as observed from the continuously streaming data but are not further presented herein due to space constraints. May 1, 2005 earthquake data of 78 second duration is extracted from the 267 second data. Figure 7 shows the time-histories (left) and corresponding amplitude REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 131 Figure 6. (left) Displacement time-histories, cross-spectrum, coherency and phase-angle plots of (left) EW motions between Tower Top at Pier 2 (Location T2) and Tower Top at Pier 3 (Location T4) and (right) vertical motions at the north end (C2) and south end (C1) of the center of the deck spectra (right) of the NS motions at west of deck center. It is seen that the frequen- cies associated with the tower have significant peaks around 1 and 2.2 Hz. Side span L2 has significant energy between 2-2.3 Hz and also at 1.3 Hz. Deck center displays several peaks, the lowest at 0.75 Hz. Figure 8 shows the time-histories and corresponding amplitude spectra of the EW motions at west of deck center. The Pier 2 tower locations, T2 and M2, exhibit significant peaks between 1.8 and 2.0 Hz and other significant peaks between 2.65-3.0 Hz. Figure 7. Time-history plots of the NS accelerations (left) and corresponding amplitude spectra (right), at key locations to the west of the deck center, recorded during the May 1, 2005 earthquake 132 ÇELEBİ Figure 8. Time-history plots (left) and corresponding amplitude spectra (right) of the EW accelerations at bent 1, mid-tower and tower top at Pier 2, and at pier 4 [location P6] and mid-tower and tower top of Pier 3 recorded during the May 1, 2005 earthquake Figure 9 shows vertical accelerations (left) and corresponding amplitude spec- tra (right) at various deck locations, downhole at Pier 2, at Bent 1 and surface and free-field array to the west of deck center. Figure 10 shows vertical accelera- tions (left) and corresponding amplitude spectra (right) at various deck locations, downhole at Pier 3, Pier 15 and surface and free-field array to the east of deck center. There is a much richer set of modes displayed in the amplitude spectra of the motions at different locations of the deck. The lowest frequency of the deck vertical motions is displayed at 0.33 Hz. The amplitude spectra and associated frequencies of the motions at the non-structural locations (free-field including downholes) and support locations of the structure (e.g. B1, L4, R4, D1) to the west of deck center are very similar but are considerably different than the spec- tra and frequencies of the motions at the non-structural locations (free-field and downholes) and support locations (Pier 15, Pier 3, Pier 4, Pier 5, downhole at Pier 3). This is attributed to different site conditions (different shear wave velocities depths to hard soil) at the (MO) west side versus the (IL) east side of the bridge. As previously stated, one of the advantages in having continuous monitoring and streaming of data with broad-band instruments is that recurring unusual re- sponses and anomalies can be observed, interpreted and if necessary cautionary steps can be initiated. In addition to earthquake response, due to observance of amplified ambient motions in data, similar occurrences were searched in hours REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 133 Figure 9. Time-history plots (left) and corresponding amplitude spectra (right) of the vertical accelerations at various deck locations, downhole at Pier 2, bent 1 and surface and free-field array to the west of deck center recorded during the May 1, 2005 earthquake of data immediately before the earthquake of May 1, 2005 and found within 300 second long interval approximately 6 hours before the earthquake. Figure 11 shows a plot of this data from key locations of the bridge and also of the free-field locations on both MO and IL side of the bridge. Other occurrences of such amplified vibrations of the deck and towers are frequently observed in the continuously streaming data. Such random and irregular fluctuations in duration and amplitudes of mo- tions of the deck and towers of the bridge are likely caused by cable vibrations - particularly because the vertical motion of the deck center and the EW (longitu- dinal) motion of the towers are coherent and are generally in-phase for significant structural frequencies. Also noted in Figure 11 is that some of the motions that originate at the bridge deck and tower are reflected to the piers and bents and then to the free-field sites - reverse to that of an earthquake (e.g. motions starting at tower and deck travel to Bent1, then to F1, and then to F2 and are attenuated and possibly diminishes by the time they reach the lower downhole, F3). In other words, the structure is forcing the ground into motion. The frequency at ∼ 2 Hz observed in the time-history plots of accelerations at deck center and tower tops are considerably different than the 0.33 Hz frequency previously noted as belonging to the fundamental vertical mode of the structure and is likely that of higher vertical mode. The 10 Hz frequency superimposed on the 2 Hz motion is likely caused by cable vibrations. 134 ÇELEBİ Figure 10. Time-history plots (left) and corresponding amplitude spectra (right) of the vertical accelerations at various deck locations, downhole at Pier 3, Pier 15 and surface and free-field array to the east of deck center recorded during the May 1, 2005 earthquake. Cause of ambient motions at E3 unknown The ambient motions of the bridge deck and towers exhibit several frequencies as depicted in the amplitude spectra in Figure 12 plotted for both 0-20 Hz and 0- 5 Hz frequency bands. It is seen in the figures that the NS spectra of tower top T2 location (at Pier 2) and T4 (at pier 3) are very similar. The same is true for the EW spectra. The 0.33 Hz frequency is observed in the EW spectra of the tower tops T2 and T4 as also in the vertical motions of the deck center, C2 but is practically absent in the spectra of NS motions. The figures show that there are several frequencies identifiable in the spectra. Different modes corresponding to the frequencies are not investigated herein. 4. Conclusions and recommendations In this paper, a new, integrated network of broad-band seismic instruments de- ployed on and in the vicinity of the new cable-stayed bridge in Cape Girardeau, MO is introduced. This is a significant accomplishment providing opportunities to acquire high quality data even at small amplitudes in the order of 10−4 − 10−3 g. Continuously recorded low-amplitude motions can facilitate assessment of dy- namic characteristics of the structure and provide a basis for estimating levels of shaking during much less frequent stronger events (as in the New Madrid Seis- mic Zone). Preliminary analyses of low-amplitude data including an earthquake REAL-TIME SEISMIC MONITORING OF THE NEW CAPE GIRARDEAU 135 Figure 11. (Left) Data from approximately 6 hours before the occurrence of the May 1, 2005 earthquake exhibits amplified motions of the deck and also feedback to the MO side free-field array. (Right) 90-100 second windows of the 300 second data exhibits motions from the bridge traveling to the free-field. The 0.5 second (2 Hz) period waveforms superimposed by approximately 0.1 s period waveforms (10 Hz) are clearly apparent and are likely caused by cable vibrations. Note the near absence of significant motions on the deepest borehole sensor (F3) demonstrates that (a) identification of response characteristics are successful and (b) amplified ambient vibrations caused by tower-cable-deck interactions are ob- served. However, their amplitudes are very small compared to amplitudes that may be experienced by stronger shaking. In the long-term, data from these small- amplitude vibrations provide opportunities to study for rupture scenarios, material fatigue and low-cycle fatigue. Signals of the data at low amplitudes are so good that, feedback from even ambient vibrations originating at the structure and traveling to the surface and downholes are observed in detail - indicating that the structure moves the ground during stronger ambient vibrations caused by towers-cable-deck interactions. It is hoped that the advance planning for instrumentation of the Cape Gi- rardeau Bridge will set an example for future large projects in seismically active regions. In the future, whenever feasible, sufficient additional sensors (e.g. for soil-structure interaction (SSI) or liquefaction or wind related) can be integrated into the new system. 136 ÇELEBİ Figure 12. Amplitude spectra of ambient motions show that the frequencies of the motions of the two towers are similar and the deck center vertical motions and EW components of towers exhibit several frequencies that are similar References Çelebi M., Purvis R., Hartnagel B., Gupta S., Clogston P., Yen P., O’Connor J., Franke M. (2004) Seismic Instrumentation of the Bill Emerson Memorial Mississippi River Bridge at Cape Girardeau (MO) : A Cooperative Effort, 4th Int’l Seismic Highway Conf., Memphis, TN. January. Çelebi M. (2005) Real-Time Seismic Monitoring of the Cape Girardeau, MO, Bridge and Preliminary Analyses of Recorded Data: An Overview, paper in preparation. Dyke S. (2001) written communication. Hanks T. (2005) personal communication. Nuttli O. W. (1974) Magnitude-recurrence relation for Central Missippi Valley earthquakes, Seismological Society of America Bulletin 64, 1189-1207. Woodward-Clyde Consultants (1994) Geotechnical Seismic Evaluation Proposed New Mississippi River Bridge, (A-5076) Cape Girardeau, Mo, , Woodward-Clyde Consultants Report 93C8036- 500. 137 137–142. EARTHQUAKE RESPONSE OF MASONRY INFILLED PRECAST CONCRETE STRUCTURES Fuat Demir, Mustafa Türkmen, Hamide Tekeli and İffet Çırak Engineering Faculty, Department of Civil Engineering, SD , Isparta, Turkey Abstract. In design of precast concrete structures, frames are generally assumed to be two dimen- sional and analysed accordingly without taking into account participation of infill walls. However, it is well known that, the infill walls increase lateral stiffness and load carrying capacity of the frame structures. The objective of this study is to investigate the effects of infill masonry on structural behavior, performance and collapse mechanism of precast concrete structures. The diagonal strut model is adopted for modeling masonry infill. Load deformation relationship and collapse mecha- nism and of bare and infilled precast concrete structures are evaluated comparatively and results are given in figures. Key words: earthquake, infill masonry, precast concrete structure 1. Introduction Hinge connected precast concrete structures are ideal solutions to build factories, which requires open spaces without columns. Numerious buildings in industrial regions are constructed by using this type of structures. Precast concrete structures are also prefered for their economy and rapid production. In Turkey, industrializa- tion is developing in the unprecedented speed. In respect of this development, the requirements for industrial structures are increased gradually. On the other hand a large part of the industrial regions are located on high risk earthquake zones. In recent earthquakes, large numbers of people have died and many of injured due to complete collapse of concrete buildings constructed with and without precast concrete structures. It is known that infill walls increase global stiffness and strength of the struc- ture. However, this features of infills are generally not taken into account in the evaluation of the lateral load capacity of precast concrete structures. In this study, the effect of infill walls on the overall structural behavior and on the lateral load capacity of hinge connected precast concrete structures is investigated. Vibration Problems ICOVP 2005, c© 2007 Springer. Ü (eds.), . şE. Inan and A. Kırı 138 DEMİR, TÜRKMEN, TEKELİ, ÇIRAK 2. Precast structures There is a great demand for industrial buildings in Turkey. In order to minimize the time span of the construction, use of precast elements is generally preferred. Turkey is located at around one of the most active fault zones in the world and is exposed frequently to destructive earthquakes .In Turkey, 92% of the territory, 95% of the overall population and 98% of the entire industrial facilities are located on seismically active zones (Tankut and Korkmaz, 2005). Most of the factories, under seismic risk, are constructed with precast elements. Since redundancy of precast structures is low, their lateral drifts under lateral loads are very large, compared to cast-in-place structures. Consequently the ductility and inelastic load capacity of precast structures are low. Many industrial facilities of precast concrete collapsed as a result of failures at beam to column connections. Particularly in Marmara Earthquake, a considerable amount of precast structures were damaged as a consequence of poor nature of the connection details. 3. Effects of masonry infills to the behavior of precast frame structures Masonry infills find widespread use as partition and external walls in precast buildings in Turkey. These masonry infills are generally constructed after complet- ing of precast frames and they are rarely are included in analysis of the structural system subjected to vertical and lateral loads. Sometime precast panels are used instead of the infills. Masonary infills and precast panels are generally considered as non-structural elements. Although they are designed to perform architectural functions, masonry infills do resist lateral forces with substantial structural action. In addition to this, infills have a considerable strength and stiffness and they have significant effect on the seismic response of the structural system. There is a general agreement among researchers that infilled frames have greater strength as compared to bare frames. On the other hand, the presence of the infill also increases the lateral stiffness considerably. Due to the change in stiffness and mass in the structural system, the dynamic characteristics of the structural system change as well. Recent earthquakes of Erzincan, Düzce and İzmit showed that infills have an important effect on the resistance and stiffness of buildings (Demir and Sivri, 2002) provided that they are constructed properly. It is not seldom that a number of structures have been strengthened by adding masonary infills in order to increase the lateral load carrying capacity of the structural system. Infill masonry effects not only the lateral load capacity and lateral displace- ment of the system, but also the damage pattern of the frames as well (Hong et al., 2002). Since precast structures generally have low rigidity, these buildings respond to seismic forces by swaying with them, rather than by attempting to resist them with rigid connections and low redundancy. Depending on the level of seismic forces, this can be an elastic response as well as plastic one. When the EARTHQUAKE RESPONSE OF MASONRY 139 Figure 1. Models for masonry. (a) Diagonal strut model, (b) continuum model structure is subjected to seismic forces, low-level cracking which is distributed throughout walls comes into being, even the forces is small. This type distributed damage indicates the behavior of the structure is working. This working increases the energy dissipation of the structure (Langenbach, 2002-2003). However, the effect of the infills on the building under seismic loading is very complex and complicated. Since the behavior of the structural systems is highly nonlinear it is very difficult to predict it by analytical methods unless the analytical models are supported by experimental data. The effects of the infills on the analysis must be considered together with high degree of uncertainty related to the behavior, namely (Penelis and Kappos, 1997); 1. The variability of their mechanical properties, and therefore the low relia- bility in their strength and stiffness; 2. Their wedging condition, that is how tightly they are connected to the surrounding frame; 3.The potential modification of their integrity during the use of the building; 4. The non-uniform degree of their damage during the earthquake. In general, the presence of masonry infills affects the seismic behavior of the building as follow (Dowrick, 1987; Tassios, 1984) 1.Stiffness of the building is increased, the fundamental period is decreased and therefore the base shear due to seismic action is increased. 2.Distribution of the lateral stiffness of the structure in plan and elevation is modified. 3. Part of seismic action is carried by the infills, thus relieving the structural system. 4. Ability of the building to dissipate energy is substantially increased. In the conventional analysis of infilled frame systems, the masonry infill is modeled using either an equivalent diagonal strut model as given in Figure 1(a) or a refined continuum model shown in Figure 1 (b). The former is simple and computation- ally attractive, and generally preferred for simplicity but is theoretically weak. 140 DEMİR, TÜRKMEN, TEKELİ, ÇIRAK Figure 2. Compression diagonal strut model for representation of the infill stiffness. First, identifying the equiva- lent nonlinear stiffness of the in- fill masonry using diagonal struts is not straightforward, especially when some openings, such as doors or windows, exist, in walls. Furthermore, it is also not possi- ble to predict the damaged area of the infill either. The latter method based on continuum model can provide an accurate computa- tional representation of both ma- terial and geometrical aspects of the problem, if the properties and the sources of nonlinearity of the infill carefully defined (Hong et al., 2002). As it is shown in Figure 1, the in-plane stiffness of infill wall is represented with an equivalent diagonal compression strut. The width of the strut Wef.depends on various parameters of the wall. However, for practical analysis it is generally assumed that (Negro and Colombo, 1996) We f = 0.175 (λhH) −0.4 √ H2 + L2 (1) λh = 4 √ Eit sin 2θ 4EcIcHi (2) H and L are the height and length of the frame, Ec, and Ei are the elastic modulus of the column and of the infill wall, respectively; t is the thickness of the infill wall, is the angle of the diagonal strut, Ic is the modulus of inertia of the column and Hi is the height of the infill wall. 4. Numerical example A structure consisting of hinge connected precast concrete frames located in Cam- pus of Suleyman Demirel University is investigated to study the effect of infill walls on the overall behavior and on the lateral load capacity. The infill walls exist along two longitudinal exterior edges. The structure is 64 m long, 21 m wide and 6.5 m high as shown in Figure 3. Quality of concrete is evaluated experimentally and found to be C30. The infill walls of thickness 110 mm exist within the frames. In the analysis, the infill walls are modelled by using the diagonal strut model. EARTHQUAKE RESPONSE OF MASONRY 141 Figure 3. The precast concrete structure Figure 4. Base shear-displacement relation- ship of the precast concrete structure In order to study seismic behaviour of the frame structure a horizontal load is applied in the longitudinal direction at the top of the structure and a non- linear pushover analysis is carried out. Flexural moment hinges are assigned to the bottom of the columns only. Default hinge properties based on ATC-40 and FEMA-273 criteria are provided (ATC-40 Report, 1996; FEMA-273, 1997). The base shear and the top displacement of the structure obtained from nonlinear pushover analyses in cases of the bare and uniformly infilled precast concrete structure are given in Figure 4. As it can be seen from the figure, the maximum base shear is about 500 kN for the bare frame and the maximum base shear is about 2200 kN for the infilled frame. Due to the existence of the infill walls, the base shear capacity of the infilled precast concrete frame increases more than four times and the displacement gets almost 60% smaller as given in Figure 4. The fundamental period of the bare and the infilled frames are 1.72 s and 1.63 s, respectively. As it is expected, the infill walls increase the base shear capacity, whereas they make the structures more stiffer. In the bare frame, the first plastic Figure 5. Collapse mechanism of bare and infilled precast concrete structure hinges occur on the bottom of the columns, when the longitudinal load is about 450 kN. When the load reaches about 550 kN, the structure is in the level of immediate occupancy. The structure is in the level of life safety, when the load is 573 kN. However, a small increase of the loads brings the structure to collapse. 142 DEMİR, TÜRKMEN, TEKELİ, ÇIRAK In masonry infilled frames, the first plastic hinges in the infill struts occur, when the load is 1224 kN. For 1993 kN the first plastic hinges on the bottom of the columns comes into being. 5. Conclusions The numerical results of the nonlinear pushover analysis show that the presence of nonstructural infill wall modifies the seismic response of the precast concrete structures to a large extend. The stability and integrity of precast concrete frames are enhanced with infill walls. Presence of infill walls also alters displacements and base shear of the precast concrete frame. If there is enough infill walls and infills cause a significant increase in lateral strength and stiffness. Although the numerical results reported base on a specific structural system, it is recommended that contributions of the infill walls should be included into structural analysis for obtaining more realistic results. References Demir F., Sivri M. (2002) Earthquake Response of Masonry Infilled Frames, ECAS International Symposium on Structural and Earthquake Engineering METU, Ankara. Dowrick D. J. (1987) Earthquake Resistant Design for Engineers and Architects, John Wiley & Sons, New York. Hong Hao, Guo-Wei M., Yong L. (2002) Damage RC Frames Subjected to Blasting Induced Ground Excitations, Engineering Structures 24 671-838. Langenbach, R. (2002-2003) Traditional Timber-laced Masonry Buildings that Survived the Great 1999 Earthquakes in Turkey and the 2001 Earthquake in India, While Modern Buildings Fell, Senior Disaster Recovery analyst, Federal Emergency Management Agency, USA. Negro P., Colombo A. (1996) Irregularities Induced by Nonstructural Masonry Panels in Framed Buildings, Engineering Structures 19 576-585. NEHRP Guidelines for the Seismic Rehabilitation of Buildings, FEMA 273, Developed by the Building Seismic Safety Council for the Federal Emergency Management Agency (Report No. FEMA 273), Washington, D. C., (1997). Penelis G. G., Kappos A. J. (1997) Earthquake-Resistant Concrete Structure, E&FN Spon, London. Seismic Evaluation and Retrofit of Concrete Buildings, 1, ATC-40 Report, Applied Technology Council, Redwood City, California, (1996). Tankut T., Korkmaz H. H. (2005) Performance of a precast concrete beam-to-beam connection subject to reversed cyclic loading, Engineering Structures 27 1392-1407. Tassios T. P. (1984) Masonry Infill and R/C Walls Under Cyclic Actions, CIB Symposium on Wall Structure, Invited State- of-the Art Report, Warsaw. 143 143–150. NONLINEAR WAVES IN FLUID-FILLED ELASTIC TUBES: A MODEL TO LARGE ARTERIES Hilmi Demiray Department of Mathematics, Faculty of Arts and Sciences, University, Maslak, Turkey Abstract. In the present work, by treating the arteries as a prestressed thin walled elastic tube of variable radius ,and the blood as an incompressible inviscid fluid, we have studied the propagation of weakly nonlinear waves in such a medium through the use of long wave approximation and the reductive perturbation method. The KdV equation with variable coefficient is obtained as the evolution equation. By seeking a progressive wave type of solution to this equation, we observed that the wave speed decreases with increasing inner radius while it increases with decreasing inner radius of the tube. Such a result is to be expected from physical considerations. Key words: solitary waves, variable coefficient KdV equation 1. Introduction Due to its applications in arterial mechanics, by employing various asymptotic methods, the propagation of small-but-finite amplitude waves has been investi- gated by (Johnson, 1970), (Hashizume, 1985), (Yomosa, 1987) and (Demiray, 1996, 1997) and various evolution equations are obtained depending on the bal- ance between the nonlinearity, dispersion and dissipation. In obtaining such evo- lution equations, they treated the arteries as circularly cylindrical long thin tubes with constant radius. In essence, the arteries have variable radius along the tube axis. In the present work, by treating the arteries as a prestressed and thin walled elastic tube, we have studied the propagation of weakly nonlinear waves through the use of the long wave approximation and the reductive perturbation method. The variable coefficient KdV equation is obtained as the evolution equation. By seeking a progressive wave type of solution to this equation, we observed that the wave speed decreases with increasing inner radius while it increases with decreasing inner radius of the tube. Such a result is to be expected from physical considerations. Vibration Problems ICOVP 2005, c© 2007 Springer. şI kı İstanbul, (eds.), . şE. Inan and A. Kırı 144 DEMİRAY 2. Basic equations and theoretical preliminaries We consider a circularly cylindrical long thin tube made of an isotropic elastic material, which is assumed to be a model for arteries, and subjected to an axial stretch ratio λz and to a variable inner static pressure P0∗(Z∗). In the course of blood flow, a dynamical radial displacement u∗(z∗, t∗) is superimposed on this initial static deformation. In reality, blood is an incompressible non-Newtonian fluid. However, for the sake of its simplicity in the analysis, we shall treat it as an incompressible invis- cid fluid. Following (Demiray, 2004), the nondimensional field equations may be given as follows: 2 ∂u ∂t + 2( f ′ + ∂u ∂z )w + (λθ + f + u) ∂w ∂z = 0, ∂w ∂t + w ∂w ∂z + ∂p ∂z = 0, p = m λz(λθ + f + u) ∂2u ∂t2 + 1 λz(λθ + f + u) ∂Σ ∂λ2 − 1 (λθ + f + u) ∂ ∂z { ( f ′ + ∂u/∂z) [1 + ( f ′ + ∂u/∂z)2]1/2 ∂Σ ∂λ1 }. (1) where p is the total fluid pressure, u is the radial displacement, Σ is the strain energy density function of the tube, w is the axial fluid velocity, ( λz, λθ ) are the stretch ratios in the axial and the circumferential directions after static deforma- tion, f (z) is a function charaacterizing the change of radius along the tube, m is a constant characterizing the inertia of the tube, λ1 and λ2 are the stretch ratios along the meridional and circumferential directions, respectively, and defined by λ1 = λz[1 + ( f ′ + ∂u ∂z )2]1/2, λ2 = (λθ + f + u). (2) In equations (1), the first equation characterizes the conservation of mass, the second one the balance of linear momentum for fluid and the third one the balance of linear momentum of the tube in the radial direction. For our future purposes we need the power series expansion of the pressure in terms of the radial displacement u as p = +p0+ m λθλz ∂2ū ∂t2 −α0 ∂2ū ∂z2 +β1ū− m λ2θλz ū ∂2ū ∂t2 −α1( ∂ū ∂z )2−(2α1− α0 λθ )ū ∂2ū ∂z2 +β2ū 2, (3) NONLINEAR WAVES IN FLUID-FILLED ELASTIC TUBES 145 where p0 is the constant initial pressure, ū = u+ f , and the coefficients α0, α1, β1 and β2 are defined by α0 = 1 λθ ∂Σ ∂λz , α1 = 1 2λθ ∂2Σ ∂λθλz , β0 = 1 λθλz ∂Σ ∂λθ , β1 = 1 λθλz ∂2Σ ∂λ2θ − β0 λθ , β2 = 1 2λθλz ∂3Σ ∂λ2θ − β1 λθ . (4) 3. Long-wave approximation In this section we shall examine the propagation of small-but-finite amplitude waves in a fluid-filled elastic tube with variable radius, whose dimensionless gov- erning equations are given in (1) and (3). For this, we employ the reductive per- turbation method (Jeffrey and Kawahara, 1981). The nature of the problem suggests that we should consider it as the boundary value problem. For such problems, the frequency is specified and the wave number is calculated through the dispersion relation k = k(ω). Thus, it is convenient to introduce the following type of stretched coordinates ξ = �1/2(z − gt), τ = �3/2z, (5) where � is a small parameter measuring the weakness of nonlinearity and disper- sion and g is the scale parameter to be determined from the solution. Solving z in terms of τ we get z = �−3/2τ. Introducing this expression for z into f (z) we have f (z) = h(�; τ). (6) In order to take the effect of the variation of the radius into account, the order of h(�, τ) should be h(�, τ) = �l(τ), where the function l(τ) is of order of 1. We shall further assume that the field quantities u, w and p may be expanded into asymptotic series as u = �u1(ξ, τ) + � 2u2(ξ, τ) + ... w = �w1(ξ, τ) + � 2w2(ξ, τ) + ... p = p0 + �p1(ξ, τ) + � 2 p2(ξ, τ) + ... . (7) Noting the coordinate transformation (5) and introducing the expansion (7) into the field equations (1) and (3), and setting the coefficients of like powers of � equal to zero, the following set of differential equations are obtained O (�) equations: −2g∂u1 ∂ξ + λθ ∂w1 ∂ξ = 0, −g∂w1 ∂ξ + ∂p1 ∂ξ = 0, p1 = β1[u1 + l(τ)]. (8) 146 DEMİRAY O (�2) equations: −2g∂u2 ∂ξ + λθ ∂w2 ∂ξ + λθ ∂w1 ∂τ + 2w1 ∂u1 ∂ξ + u1 ∂w1 ∂ξ + l(τ) ∂w1 ∂ξ = 0, −g∂w2 ∂ξ + ∂p2 ∂ξ + ∂p1 ∂τ + w1 ∂w1 ∂ξ = 0, p2 = ( mg2 λθλz − α0) ∂2u1 ∂ξ2 + β1u2 + β2[u1 + l(τ)] 2. (9) 4. Solution of the field equations From the solution of the set of equations (7) we obtain u1 = U(ξ, τ), w1 = 2g λθ [U + w̄1(τ)], p1 = β1[U + l(τ)], (10) provided that the following relation holds true g2 = λθβ1 2 , (11) where U(ξ, τ) is an unknown function whose governing equation will be obtained later, 2gw̄1(τ)/λθ corresponds to the steady axial flow resulting from the pressure β1l(τ), which also causes the radial displacement l(τ) for the tube. Introducing the solution given in (10) and (11) into equation (9), we have −2g∂u2 ∂ξ + λθ ∂w2 ∂ξ + 2g( ∂U ∂τ + dw̄1 dτ ) + 4g λθ (U + w̄1) ∂U ∂ξ + 2g λθ U ∂U ∂ξ − 2g λθ h1(τ) ∂U ∂ξ = 0 −g∂w2 ∂ξ + ∂p2 ∂ξ + 2g2 λθ ( ∂U ∂τ + l dτ ) + 4g2 λ2θ (U + w̄1)) ∂U ∂ξ = 0, p2 = ( mg2 λθλz − α0) ∂2U ∂ξ2 + β2[U + l(τ)] 2 + β1u2. (12) Eliminating w2 between equations (12), we have 4g2 λθ ∂U ∂τ + ( 10g2 λ2θ + 2β2)U ∂U ∂ξ + ∂U ∂ξ [ 8g2 λ2θ w̄1 + 2g2 λ2θ l(τ) + 2β2l(τ)] NONLINEAR WAVES IN FLUID-FILLED ELASTIC TUBES 147 +( mg2 λθλz − α0) ∂3U ∂ξ3 + 2g2 λθ d dτ [w̄1 + l(τ)] = 0. (13) The equation (13) should even be valid when U = 0 (steady flow), which results in d dτ [l(τ) + w̄1(τ)] = 0, or ¯w1(τ) = −l(τ). (14) Introducing (14) into equation (13) yields the following Korteweg-de Vries equa- tion with variable coefficient ∂U ∂τ + µ1U ∂U ∂ξ + µ2 ∂3U ∂ξ3 + µ3l(τ) ∂U ∂ξ = 0, (15) where the coefficients µ1, µ2 and µ3 are defined by µ1 = 5 2λθ + β2 β1 , µ2 = ( m 4λz − α0 2β1 ), µ3 = − 1 λθ + β2 β1 . (16) By introducing the following coordinate transformation τ′ = τ, ξ′ = ξ − µ3 ∫ τ 0 l(s)ds, (17) the variable coefficient KdV equation reduces to the conventional KdV equation ∂U ∂τ′ + µ1U ∂U ∂ξ′ + µ2 ∂3U ∂ξ ′3 = 0. (18) This nonlinear evolution equation admits the following type of localized travelling wave solution U = a sech2ζ, (19) wher a is the wave amplitude and the variable ζ is defined by ζ = ( µ1a 12µ2 )1/2[ξ − µ3 ∫ τ 0 l(s)ds − µ1a 3 τ]. (20) Here, it is seen from equation (20) that the trajectory of the solitary wave is not a straight line, it is rather a curve. In our subsequent analysis we need the explicit expressions of the coefficients α0, β1, , β3. For that purpose we shall utilize the constitutive equation proposed by (Demiray, 1972) for soft biological tissues as Σ = 1 2α {exp[α(λ2θ + λ 2 z + 1 λ2θλ 2 z − 3] − 1}, (21) 148 DEMİRAY where α is a material constant. Introducing (21) into (4), the explicit expressions of the coefficients α0, α1, β1, β2, may be given as follows α0 = 1 λθ (λz − 1 λ2θλ 3 z ) F, β1 = [ 4 λ5θλ 3 z + 2 α λθλz (λθ − 1 λ3θλ 2 z )2] F, β2 = [− 10 λ6θλ 3 z + α λθλz (λθ − 1 λ3θλ 2 z )(1 + 11 λ4θλ 2 z ) + 2 α2 λθλz (λθ − 1 λ3θλ 2 z )3] F, (22) where the function F is defined by F = exp[α(λ2θ + λ 2 z + 1 λ2θλ 2 z − 3)]. (23) So far we have not said anything about the form of the function l(τ), which characterizes the variation of the radius. In what follows we shall study some special cases. (i) Linear Tapering: For this case the function l(τ) can be approximated by l(τ) = Φτ, (24) where Φ is the tapering angle. Introducing (24) into (20) we obtain ζ = ( µ1a 12µ2 )1/2[ξ − 1 2 (− 1 λθ + β2 β1 )Φτ2 − µ1a 3 τ]. (25) As can be seen from equation (25), the trajectory of the wave is a parabola, which is the result of variable radius of the tube. In order to study the numerical illustration we need the value of the material constant α. The mathematical model that we have presented here was compared before by us (Demiray, 1976) with some experimental measurements by Simon et al (Simon et al., 1972) on canine abdominal artery with characteristics Ri = 0.31cm Ro = 0.35cm and λz = 1.6 and the value of material constant α was found to be α = 1.948. Using this value of material constant α and the initial deformation λz = 1.6 and λθ = 1.6 the values of (β2/β1) and µ1 are calculated as β2/β1 = 3.348, µ1 = 4.89. For this case the variable ζ becomes ζ = ( µ1a 12µ2 )1/2(ξ − 1.35Φτ2 − 1.63aτ). (26) Choosing the values of Φ and a as unity, it becomes ζ = ( µ1 12µ2 )1/2(ξ − 1.35τ2 − 1.63τ). (27) NONLINEAR WAVES IN FLUID-FILLED ELASTIC TUBES 149 Here we note that ξ characterizes the time variable, whereas τ characterizes the space variable. The speed of propagation may be defined by vp = dτ dξ = 1 1.63a + 2.7Φτ . (28) Here 1/1.63a corresponds to the wave speed for a tube of constant radius. As is seen from this expression, the wave speed decreases with distance parameter τ for Φ > 0 (expanding tube ) but increases for Φ < 0 (narrowing tube). This is to be expected from physical consideration. (ii) Stenosised Artery : For this case the function l(τ) can be approximated as l(τ) = −c sech2τ, (29) where c is the magnitude of the stenosis. Introducing this expression (29) into equation (20) we obtain ζ = ( µ1a 12µ2 )1/2[ξ − (β2 β1 − 1 λθ )c tanh τ − 1.63aτ]. (30) Again, the profile of the travelling wave is bell-shaped and the speed of the solitary wave may be defined by vp = 1 [µ1a3 + ( β2 β1 − 1λθ )csech 2τ] . (31) Here, again, it is seen that the wave speed increases with distance τ from the center of the stenosis. 5. Conclusions The results of linearly tapered and stenosed arteries show that the mathemati- cal model presented here yields the solitary wave solution, as observed by (Mc- Donald, 1974) in his experimental measurements. Moreover, as it is to be ex- pected from physical considerations, the wave speed decreases for expanding tube whereas it increases for narrowing tubes. References Demiray H. (1972) A not on the elasticity of soft biological tissues, J. Biomechanics 5 309-311. Demiray H. (1976) Large deformation analysis of some basic problems in biophysics, Bull. Math. Biology 38 701-711. Demiray H. (1996) Solitary waves in a prestressed elastic tube, Bull. Math. Biology 58 939-955. Demiray H., Antar N. (1997) Nonlinear waves in an inviscid fluid contained in a prestressed viscoelastic thin tube, Z. Angew. Math. Phys. 48 325-340. 150 DEMİRAY Demiray H. (2004) Modulation of nonlinear waves in a fluid-filled elastic tube with stenosis, Int. J. Mathematics and Mathematical Sciences 60 3205-3218. Jeffrey A., Kawahara T. (1981) Asymptotic Methods in Nonlinear Wave Theory, Pitman, Boston. Johnson R. S. (1970) A nonlinear equation incorporating damping and dispersion, J. Fluid Mechanics 42 49-60. Hashizume Y. (1985) Nonlinear pressure waves in a fluid-filled elastic tube, J. Phys. Soc. Japan 54 3305-3312. McDonald D. A. (1974) Blood Flow in Arteries, The Williams and Wilkins Co., Baltimore. Simon B. R., Kobayashi A. S., Stradness D. E., Wiederhielm C. A. (1972) Re-evaluation of arterial constitutive laws, Circulation Research 30 491-500. Yomosa S. (1987) Solitary waves in large blood vessels, J. Phys. Soc. Japan 56 506-520. 151 151–156. THE NONLINEAR AXISYMMETRIC VIBRATIONS OF CIRCULAR PLATES WITH LINEARLY VARYING THICKNESS UNDER RANDOM EXCITATION Vedat Doğan Department of Aeronautical Engineering, İstanbul Technical University, İstanbul, Turkey Abstract. In this study, the nonlinear axisymmetric behavior of circular isotropic plates with lin- early varying thickness under random excitation is investigated. It is assumed that plate response is axisymmetric when plate is subjected to axisymmetric random loading. The Berger type nonlinear- ity is used to obtain the governing equations of motion for clamped circular plates. A Monte Carlo simulation of stationary random processes, single-mode Galerkin-like approach, and numerical in- tegration procedures are used to develop nonlinear response solutions. Response time histories, root mean squares and spectral densities are presented for different random pressure levels. Parametric results are also presented. Linear responses are included to investigate the nonlinear effects. Key words: circular plates, nonlinear, random vibration 1. Structural formulation The nonlinear vibrations of plates and shells were extensively studied (Sathyamooorthy 1997, Nayfeh 2004). Circular plates are commonly used en- gineering structural components and their vibrations have been investigated by many researchers (Chang 1994, Dumir et al., 1984, Efstathiades 1971, Ghosh 1986, Hadian and Nayfeh 1990, Liu and Chen 1996, Ramachandran 1975, Sathyamooorthy 1996, Smaill 1990, Sridhar et al., 1975). The Berger type nonlinear equation of motion for axisymmetric vibrations in terms of normal deflection is given as (Ramachandran 1975) D∇4w + ∂D ∂r { 2 ∂3w ∂r3 + 2 + ν r ∂2w ∂r2 − 1 r2 ∂w ∂r } + + ∂2D ∂r2 { ∂2w ∂r2 + ν r ∂w ∂r } − N∇2w = ρh∂ 2w ∂t2 − ch∂w ∂t + pr (r, t) (1) Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 152 DOĞAN and Nh2 12D = ∂u ∂r + 1 2 ( ∂u ∂r )2 + u r (2) where ρ , c and h are the material density, viscous damping coefficient and plate thickness, respectively. The pr (r, t) is random pressure acting on the circular plate. It is assumed that radial displacement vanishes at the boundary (i.e. u(a, t) = 0 and also u(0, t) = 0 ), after Eqn. (2) is multiplied by rdr and integrated over the plate area (Ramachandran 1975), then Nh21 12D1 a2 α2 { α + log(1 − α)} = −1 2 ∫ a 0 ( ∂w ∂r )2 rdr (3) where α is the taper ratio. Linearly varying thickness can be taken as h = h1 ( 1 − α r a ) (4) and the bending stiffness D = D1 ( 1 − α r a )3 (5) where h1 and D1 are the thickness and flexural stiffness at r = 0, respectively. The plate bending stiffness is D1 = Eh31 12(1−ν2) where E and ν are the modulus of elasticity and the Poisson’s ratio. Solution for normal (transverse) deflection w(r, t) may be approximated by w(r, t) = N1∑ m=1 ηm(t)φm(r) (6) where ηm(t) is the mth generalized coordinate,φm(r) is the mth shape function which satisfies the normal boundary conditions, and N1 is the number of chosen shape function. For an approximation, a single mode is chosen in this study w(r, t) = η(t)φ(r) = η(t) [ 1 − ( r a )2]2 (7) Equation (7) clearly satisfies the clamped edge boundary conditions. Inserting Eqn.(7) into Eqn. (3) and Eqn.(1), then Eqn. (3) into Eqn. (1), with simulated external random pressure pr(t) (Shinozuka and Jan, 1972), Galerkin procedure is used to reduce the system into the following ordinary nonlinear differential equation: ρh1 d2η dt2 + ch1 dη dt + µ1η(t) + µ2η 3(t) = λpr(t) (8) VIBRATIONS OF CIRCULAR PLATES 153 where µ1 = 44D1 a4 (693 − 256α) [ 1680 − 315α2 (−9 + ν) − 288α (15 + 2ν) + 32α3 (−23 + 6ν) ] (9) µ2 = − 18480α2D1 a4h21 [ log(1 − α) + α] (10) λ = 1160 (693 − 256α) (11) The numerical solutions of Eqn.(8) were carried out by using the Runga-Kutta fifth order method. Aluminum plate with h1 = 0.002m , a = 0.5m, α = 0.2 is considered, unless indicated otherwise. Responses are found at the center of the plate. The displacement response time histories for SPL=160 dB is given in Fig. 1. For selected geometry/material properties, RMS responses for different Sound Pressure Levels are presented in Fig. 2, which also includes linear response. Non- linearity seems to gain importance for input loading above 120 dB. Effect of the taper ratio on the responses is shown in Fig. 3. Response spectral density is given in Fig. 4. Responses for different thickness to radius ratio are presented in Fig. 5. Figure 1. Displacement response time histories for SPL=160 dB 154 DOĞAN Figure 2. Linear and nonlinear RMS responses for different Sound Pressure Level Figure 3. The taper ratio effect on the RMS responses VIBRATIONS OF CIRCULAR PLATES 155 Figure 4. Response Spectral Density Figure 5. RMS responses vs. Sound Pressure Levels for different radius/thickness ratios 156 DOĞAN References Chang T. P. (1994) Random Vibration of a geometrically nonlinear orthotropic circular plate, Journal of Sound and Vibration 170 426-430. Dumir P. C., Nath Y., Gandhi M. L. (1984) Non-linear axisymmetric transient analysis of orthotropic thin annular plates with a rigid central mass, Journal of Sound and Vibration 97 387-397. Efstathiades G. J. (1971) A new approach to the large-deflection vibrations of imperfect circular disk using Galerkin’s procedure, Journal of Sound and Vibration 16 231-253. Ghosh S. K. (1986) Large Amplitude Vibrations of Clamped Circular Plates of Linearly Varying Thickness, Journal of Sound and Vibration 104 371-375. Hadian J., Nayfeh A. D. (1990) Modal Interaction in Circular Plates, Journal of Sound and Vibration 142 279-292. Liu C. F. Chen G. T. (1996) Geometrically Nonlinear Axisymmetric Vibrations of Polar Orthotropic Circular Plates, Int. J. Mech. Sci. 38 325-333. Nayfeh A. H., Pai P. F. (2004) Linear and Nonlinear Structural Mechanics, John Wiley Sons, New Jersey. Ramachandran J. (1975) Nonlinear vibrations of circular plates with linearly varying thickness, Journal of Sound and Vibration 38 225-232. Sathyamoorthy M. (1996) Large Amplitude Circular Plate Vibration with Transverse Shear and Rotatory Inertia Effects, Journal of Sound and Vibration 194 463-469. Sathyamoorthy M. (1997) Nonlinear Analysis of Structures, CRC Pres, Boca Raton, Florida. Shinozuka M., Jan C. M. (1972) Digital simulation of random processes and its applications, Journal of Sound and Vibration 25 111-128. Smaill J. S. (1990) Dynamic Response of Circular Plates on Elastic Foundations: Linear and Non- Linear Deflection, Journal of Sound and Vibration 139 487-502. Sridhar S., Mook D. T., Nayfeh A. H. (1975) Non-linear Resonances in the forced responses of plates Part I: Symmetric Responses of Circular Plates, Journal of Sound and Vibration 41 359-373. 157 157–162. DYNAMIC ANALYSIS OF A HELICOPTER ROTOR BY DYMORE PROGRAM Vedat Doğan and Mesut Kırca Faculty of Aeronautics and Astronautics, Rotorcraft Research and Development Centre, İstanbul Technical University, İstanbul, Turkey Abstract. The dynamic behavior of hingeless and bearingless blades of a light commercial he- licopter which has been under design process at ITU (İstanbul Technical University, Rotorcraft Research and Development Centre) is investigated. Since the helicopter rotor consists of several parts connected to each other by joints and hinges; rotors in general can be considered as an assembly of the rigid and elastic parts. Dynamics of rotor system in rotation is complicated due to coupling of elastic forces (bending, torsion and tension), inertial forces, control and aerodynamic forces on the rotor blades. In this study, the dynamic behavior of the rotor for a real helicopter design project is analyzed by using DYMORE. Blades are modeled as elastic beams, hub as a rigid body, torque tubes as rigid bodies, control links as rigid bodies plus springs and several joints. Geometric and material cross-sectional properties of blades (Stiffness-Matrix and Mass-Matrix) are calculated by using VABS programs on a CATIA model. Natural frequencies and natural modes of the rotating (and non-rotating) blades are obtained by using DYMORE. Fan-Plots which show the variation of the natural frequencies for different modes (Lead-Lag, Flapping, Feathering, etc.) vs. rotor RPM are presented. Key words: bearingless, rotor dynamics, Dymore 1. Introduction There are basically four types of helicopter rotor hubs in use. These are the teeter- ing design, the articulated design, the hingeless design, and the bearingless design (Schindler and Pfisterer, Chopra and Inderjit, Mil et al., 1967). One of the bear- ingless rotor designs is “Elastic Articulation” (EA) Rotor which was originally developed by Lockheed, California in the early 1960s and further developed by Tom Hanson. General overview of EA Rotor system is shown in Fig. 1. In EA rotor system, the flexbeam is responsible to carry out all blade motions (flapping, lead-lag, and feathering). The torque tubes give the pitching input to the rotor system without taking the tensional stresses by using the spherical pivot bearing at the root. A sweep angle exists between the blade c/4 line and the feath- ering axis. This sweep angle is the part of a system that aims to achieve automatic blade stabilization during the articulation of the system. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 158 DOĞAN, KIRCA Figure 1. EA Rotor Exploded View (Hanson, 1998) 2. Analysis In this study, Hanson’s Rotor system (Hanson, 1998) shown in Fig. 2 is adapted as base. The main tasks related with designing a rotor are calculating the clear- ance between blade tip and tail boom, obtaining the mode shapes, plotting all the frequencies in the fan plot, manipulating the flexure and blade properties to find an optimum frequency placement to avoid resonance, investigating the ground and air resonance. DYMORE is a handy tool for the analysis of these types of problems. DYMORE is a multi-body dynamic analysis code developed by Prof. O.A. Bauchau at Georgia Institute of Technology (Bauchaou). The needed in- puts for dynamic analysis from a preliminary design program output file include Figure 2. General view of the bearingless rotor (EA) system DYNAMIC ANALYSIS OF A HELICOPTER ROTOR 159 Figure 3. Flexure-blade assembly Figure 4. Flexure cross-section gross weight, tip speed, number of blades, rotor diameter, blade chord, tip weight, forward sweep angle and blade weight. Present study centers around the fan plot which gives variations of the natural frequencies of rotating blades vs. rotor speed. A CATIA model of flexbeam-blade assembly is shown in Fig. 3 , Fig. 4 and Fig. 5 give the cross sections of flexbeam and blade at a station, respectively. Materials used in blades are also indicated in Fig. 5. Rotor blades and flexure are usually modeled as slender beams with heterogeneous cross sections. VABS (Variable Asymptotic Beam Section) is a computer program that implements a variable asymptotic method for computing the stiffness of a heterogeneous beam at a given cross-section. 160 DOĞAN, KIRCA Figure 5. Blade cross-section materials VABS is capable of computing the full 6-by-6 stiffness matrix (all bending, ex- tension, shear and torsion terms) with asymptotic accuracy. The sectional stiffness can then be used in 3-D flexible multibody analysis tools such as DYMORE . Following main steps are employed to obtain stiffness matrix: − Take cross-sections at a given (chosen) station (Fig. 7) − Import those sections to ANSYS − Create ANSYS areas in each cross section. − Generate a 2-D mesh within each area and nodal compatibilities between areas. − Write macros to define reference curves and element material orientations and 3-D material properties − Prepare VABS input file − Run VABS, and get VABS Output file. Output (Figure 8) includes “stiffness matrix” which can be readily used within DYMORE model. Steps are shortly repeated in Fig. 6. Figure 6. Flow diagram of the analysis DYNAMIC ANALYSIS OF A HELICOPTER ROTOR 161 Figure 7. Blade and flexure cross sectional stations from CAD model To avoid resonance, natural frequencies must not cross 1P to 4P lines in the vicin- ity of 1P of rotor speed. If it does, as in Fig. 9, 2nd Lead-Lag frequency and 4P lines cross each other, resonance occurs. By changing geometric and material properties as well as other design parameters, “optimum” frequency placement can be achieved. Figure 8. VABS Output required by DYMORE 162 DOĞAN, KIRCA Figure 9. Fan-plot (Natural Frequencies vs. Rotor Speed) obtained by DYMORE References Bauchaou O. A. DYMORE: User’s Manual and Theory Manual. Chopra, Inderjit. Helicopter Dynamics, University of Maryland, Center for Rotorcraft Education and Research. Hanson T. F. (1998) Designer Friendly Handbook of Helicopter Rotor Hubs. Mil M. L., et al. (1967) Helicopter Calculation and Design, Volume 2- Vibrations and Dynamic Stability, Moscow. Schindler R., Pfisterer. Impacts of Rotor Hub Design Criteria on the Operational Capabilities of Rotorcraft, AGARD-CP-354. 163 163–174. VIBRATION SUPPRESSION ∗ Horst Ecker Vienna University of Technology,Institute of Mechanics and Mechatronics, A-1040 Vienna, Austria Abstract. This contribution presents an overview of recent research activities, with the goal of uti- lizing parametric excitation as a novel method to suppress self-excited vibrations by amplifying the damping properties of the system. The basic idea of the method is discussed and different methods to analyze parametrically excited systems are reviewed. Three different examples are presented, where parametric stiffness excitation is applied and either stabilizes the system or improves the damping behavior. Key words: parametric resonance, self-excitation, stability of mechanical systems 1. Introduction If an excitation is applied to a mechanical system it will, in general, vibrate. Most of the time these vibrations are unwanted and require countermeasures if the vibration amplitudes exceed a certain level. Depending on the nature of the excitation, different strategies and methods of vibration reduction are in use. This contribution focuses on a rather new idea to reduce, and in some cases even cancel vibrations in mechanical systems by means of parametric excitation. In mechanical systems different types of excitation are observed. A frequently encountered type is the external excitation, also named forced excitation. An unbalance excitation as known from rotating machinery is one example, the exci- tation of a structure which is attached to a vibrating foundation is another. A mechanical system under the influence of an external excitation will per- form forced vibrations. Vibration amplitudes are determined by the dynamical properties of the system and the amplitude of excitation. In the case of a lin- ear system, vibration amplitudes are proportional to excitation amplitudes and damping properties of the system. Vibration reduction for the system in general is mainly either achieved by choosing these parameters appropriately or by tuning the system as to avoid resonances. ∗ This work was partly funded by the Austrian Science Fund under Grant P-16248 Vibration Problems ICOVP 2005, c© 2007 Springer. PARAMETRIC STIFFNESS EXCITATION AS A MEANS FOR (eds.), . şE. Inan and A. Kırı 164 ECKER Although mentioned in first place, the reduction of vibrations caused by exter- nal excitation will only play a minor role in this paper. We will almost exclusively deal with mechanical systems where two other types of excitation mechanisms are prevailing: Self-Excitation (SE) and Parametric Excitation (PE). As in most practical cases, self-excitation will be the source of unwanted vibrations. It will be shown that parametric excitation can be employed as a mean to suppress self- excited vibrations. 2. Parametric excitation - past and presence Parametrically excited systems have been of interest since a long time, and date back as far as to the 19th century, when M. Faraday investigated sloshing liquids in a container and, about 30 years later, when E. Mathieu established the famous equation given his name. Since then, parametric excitation (PE) has attracted much interest, mainly because it may lead to a unique type of resonances, called parametric resonances. Given the available space, it is virtually impossible to even give a brief overview and just name the important contributions in this field of ongoing active research. From an application point of view and focussing on mechanical engineering two different aspects of parametric excitation can be extracted from the numerous references: how to avoid or reduce the effect of parametric excitation in a system and how to take advantage of PE, especially PE-resonances. The second aspect is much less popular and not very many applications of PE in this sense are known. Within this area, the idea prevails, to make use of the large amplitudes which will occur when a system is operated at a parametric resonance. For instance, recent works focus on micro-electromechanical systems (MEMS) as a possible application, see e.g. (Shaw, et al., 2004). However, not all PE-resonances lead to large amplitudes, since some of them may be non-resonant. This special case has not been studied at all, until Tondl found out about an interesting phenomenon associated with non-resonant parametric resonances (Tondl, 1998). In his paper Tondl shows early results obtained from analog computer simu- lations of an unstable, non-linear, parametrically excited system, see Fig. 1. The surprising detail in this result is a frequency interval of the PE, where the self- excited vibration amplitudes of the system are completely suppressed. Since this occurs at the frequency of a parametric resonance, the phenomenon was named parametric anti-resonance. This pioneering work triggered research efforts at various places. It led to a growing number of contributions related to this phenomenon, only some of them shall be mentioned here. Analytical methods and bifurcation analysis have been applied in Utrecht by Verhulst and his students (Fatimah, 2002), (Abadi, 2003). Very comprehensive investigations, both analytically and numerically were carried out in Vienna by Dohnal (Dohnal, 2005) and this author (Ecker, 2003). PARAMETRIC EXCITATION AND VIBRATION SUPPRESSION 165 Figure 1. Simulation result (obtained with an analog computer) of a self-excited system exhibiting vibration suppression near the parametric combination resonance frequency η0 = Ω2 − Ω1. From (Tondl, 1998) Parametric excitation of a more general type has been applied by Makihara in Tokio (Makihara et al., 2005) and valuable contributions have been made also by Nabergoj from Trieste (Nabergoj and Tondl, 2001). Last but not least, Tondl himself has continued to study the effect of vibration quenching by parametric excitation (Tondl et al., 2005) and e.g. has also investigated parametric damping and parametric mass excitation, see (Tondl, 2001). 3. Modelling systems with parametric stiffness excitation The generic equations of motion of a mechanical system with parametric stiffness excitation can be written in a rather general matrix form as Mẍ + [ C + G(ν) + CZ(x) ] ẋ + [ K + N(ν) + KZ(x) ] x + KPE(t)x = Fex. (1) The vector of deflections is denoted x. For a linear, homogeneous system with constant system matrices, only the following matrices would be needed and there- fore non-zero: mass matrix M, damping matrix C, stiffness matrix K. Parametric stiffness excitation (PSE) is introduced by matrix KPE(t) with time-periodic co- efficients according to harmonic functions cos(ωt + pi j). Only single-frequency PSE with frequency ω is considered for this system but multi-location parametric excitation is not excluded. Phase relations between different locations of PE are introduced by phase angles pi j KPE(t) = cos(ωt + pi j)PE. (2) 166 ECKER The number of degrees of freedom of the system determines the size of the sys- tem matrices. It is pretty obvious how to establish these matrices for simple two or three mass chain systems, as used in (Tondl, et al., 2005) and several other references by the author. Self-excitation can be introduced to the system by setting elements of the damping matrix C to a negative value. Negative damping is one of the common methods to represent the effect of flow-induced self-excitation (Blevins, 1977). Basic non-linear behavior can be represented by the additional stiffness and damping matrices CZ(x) and KZ(x), which may depend in an arbitrary way on vector x and also, if required, on ẋ. Matrix G(ν) is a function of a system parameter ν and is needed in mechanical systems to represent gyroscopic forces, which would depend on a rotational speed ν. Also frequently encountered in rotor systems are non-conservative forces N(ν)x, created by bearings and seals. Such forces usually increase with increasing rotor speed and may ultimately destabilize such a system. It is of particular interest to investigate rotor system, since the effect of a parametric anti-resonance could improve the performance of rotating machinery quite significantly. Finally, to take into account the effect of external forces, Fex appears on the right hand side of Eq.(1). 4. Parametric resonance frequencies It is widely known, see e.g. (Cartmell, 1990), that a system with parametric ex- citation may exhibit Principle Parametric Resonances at frequencies ηprj/n and Parametric Combination Resonances at frequencies ηcrj±k/n for the PSE-frequency ω equal to: η pr j/n = 2Ω j n , ηcrj±k/n = |Ω j ±Ωk| n , ( j � k), ( j, k, n = 1, 2, 3, ...). (3) Symbols Ω j and Ωk denote the j-th (k-th) natural frequency of the system. The denominator n represents the order of the parametric resonance. Most of the time only first order resonances n = 1 of the lowest frequencies Ω1,Ω2 are significant. The effect of a parametric anti-resonance can only occur for parametric combi- nation resonances. It depends on the system, whether the difference type or the summation type is non-resonant and can be used to achieve vibration suppression. It can be shown (Ecker, 2003) that for a symmetric stiffness matrix K = KT parametric vibration suppression will occur for the difference-type combination resonance ηcr( j−k)/1 = (Ω j − Ωk) and that an interval of instability will be observed at the summation-type combination resonance ηcr( j+k)/1 = (Ω j + Ωk). To predict the appropriate PE-frequency for vibration suppression it is neces- sary to know the natural frequencies of the system. Therefore, the lower undamped PARAMETRIC EXCITATION AND VIBRATION SUPPRESSION 167 natural frequencies Ω1,2,3,... have to be calculated from system Eq.(1) by solving the eigenvalue problem for Mẍ + Kx = 0. 5. Analysis of systems with parametric excitation In its most general version Eq.(1) defines a set of non-homogeneous, non-linear, time-periodic differential equations of second order. Also because of its basically unlimited complexity with regard to the size of the system there does not exist a single method to suit all kind of problems. Depending on the actual size of the problem, the presence of non-linearities and inhomogeneous terms, different methods are advantageous to be applied. Another factor is also if time series of system states of the original problem are sought or if only the local stability of such a solution is of interest. 5.1. NUMERICAL SIMULATION METHOD The most direct method, which can be applied to virtually any kind of such prob- lems is numerical simulation. By integrating the system equations in the time domain, starting from initial conditions, the solution is computed. Nonlinearities and large matrices only affect the computational speed, but would not prevent using simulation. Of course, appropriate integration methods have to be applied, to balance computational effort and accuracy. Nevertheless, CPU-time may become still a problem, when the stability of a system near the stability threshold shall be investigated and very slowly changing transients have to be followed. 5.2. ANALYTICAL METHODS A number of analytical and semi-analytical methods have been developed to deal with time-periodic systems. Even trying to briefly introduce the most interest- ing ones would exceed the length of this overview by far. Therefore, only one method is explicitly mentioned, since it has been used by the author and his co- workers and others quite successfully in this context. This method is nowadays mostly called Method of Averaging (MoA). However, based on being promoted by Krylov, Bogoliubov and Mitropolski, in the past the method is also associated with these names. A rather detailed comparison of three distinctively different methods is presented in (Ecker, 2003). The Method of Averaging is applicable to a linear(ized) and homogeneous subset of Eq.(1) and will primarily provide information about the stability of the system. It can be implemented as a first order method, as well as for higher orders. However, deriving a first order solution can be already cumbersome for a low-dimensional system. This, and the need to identify a small parameter in the system, are the major disadvantages of this method. But to be fair, a price has to be 168 ECKER paid with practically every of the analytical methods. A very recent and detailed presentation of AoM is found in (Verhulst, 2005). The application to PE-systems of various complexity is thoroughly discussed in (Dohnal, 2005). The advantages of analytical methods can be seen easily by the following example. Equations (3) and (4) are a simplified and normalized version of Eq.(1), which have been used by Tondl and others to investigate the stability of PE-excited two-degree of freedom systems. u′′ +Ω2u = −ε (Θu′ + cos ητQcu) , (4) u = [ u1 u2 ] , Θ = [ Θ11 Θ12 Θ21 Θ22 ] , Ω2 = [ Ω21 0 0 Ω22 ] , Qc = [ Qc11 Q c 12 Qc21 Q c 22 ] . (5) By application of the Method of Averaging two necessary conditions are ob- tained for stability at a parametric excitation frequency η = (Ω2 −Ω1): Θ11 + Θ22 > 0, (6) Θ11Θ22 + 1 4Ω1Ω2 Qc12Q c 21 > 0. (7) Not only the stability of the system can be examined for a certain PE-frequency. It is also possible to calculate the frequency interval of stability in the vicinity of this frequency. The interval is defined as η0 + εσlo < η < η0 + εσhi. (8) with σlo,hi = ∓ (Θ11 + Θ22) 2 √ − Qc12Q c 21 4Ω1Ω2Θ11Θ22 − 1. (9) It is interesting to note that first order averaging leads to exactly the same results as obtained in (Tondl, 1998) by a different method based on Floquet theory. In exchange for compact results, one has to accept, that accuracy is degraded as soon as the parameter ε ≮ 1 cannot be considered as small anymore, at least if only a first order approximation is used. PARAMETRIC EXCITATION AND VIBRATION SUPPRESSION 169 5.3. NUMERICAL STABILITY ANALYSIS The stability of the trivial solution x = 0 of system Eq.(4),(5) can be investi- gated also numerically by means of Floquet-theory, see (Verhulst, 2000). Floquet’s theorem postulates that for a system of first order differential equations ẏ = A(t) y, A(t) = A(t + T ), (10) with a T -periodic matrix A(t) each fundamental matrix M(t) of the system can be represented as a product of two factors M(t) = Q(t) exp(tC), (11) where Q(t) is a T -periodic matrix function and C is a constant matrix. Stability of the time-periodic system can be determined either from the eigen- values of the Floquet exponent matrix C or from the monodromy matrix M(T ), which is in fact the state transition matrix evaluated after a period T . The mon- odromy matrix can be calculated numerically by repeated integration of the sys- tem equations over one period T , starting from independent sets of initial condi- tions. It is convenient to use the columns of the identity matrix I as initial vectors to start from. By solving n initial value problems over one period T ẏ = A(t)y, [y(0)1, y(0)2, ..., y(0)n] = I, t = [0,T ], (12) and by arranging the results as follows M(T ) = [y(T )1, y(T )2, ..., y(T )n] (13) the monodromy matrix is obtained. Finally the eigenvalues of the monodromy matrix Λ = eig(M(T )), (14) are calculated numerically. The system is unstable if any of the eigenvalues are larger than one in magnitude max(|Λ1|, |Λ2|, ..., |Λn|) { < 1 stable system > 1 unstable system. (15) This procedure leads to a very efficient numerical method, which allows to ex- amine the stability of a parametrically excited system much faster than with direct numerical integration. This method, however, cannot be used when nonlinearities are present and will not give vibration amplitudes. 6. Examples In this section three examples of mechanical systems with parametric excitation are presented, to demonstrate the capabilities of the proposed method. Due to 170 ECKER D c02c01 k12 x2x1 gxd3k01ec01 m1 m2 Figure 2. Two-mass system with self-excitation because of negative damping c02 < 0 and parametric stiffness excitation at k01(t) limited space, not all the details for every model can be included and therefore references are given where to find the full documentation for each example. 6.1. TWO-MASS SYSTEM In Figure 2 a schematic is shown of a two-mass system with one stiffness element being periodically changed according to k01(t) = k01(1+ec01 cos(ωt). It is assumed that the damping element c02 < 0 and therefore the system may exhibit self- excited vibrations. To demonstrate the stabilizing effect of parametric excitation, the system was investigated for the critical value of parameter c02 at the stability threshold. In Fig. 3 the stable area is filled white and the unstable area is grey. This result was obtained by a numerical stability analysis. One can see easily that significantly lower values of the critical parameter c02 are possible at the combination resonance frequency of the system Ω2 −Ω1. This indicates that PE at this frequency and with a sufficiently high amplitude will allow for an increase of the self-excitation parameter without becoming unstable. If the wrong frequency is chosen, however, then the effect is turned upside down, as one can see from the result in the vicinity of the parametric combination resonance of the summation type. The stability threshold, separating stable and unstable regions, is also plotted as a broken line, near the parametric resonances. These two lines have been ob- tained by analytical formulas for the interval of stability, similar to Eqs.(8) and (9). Note that also the interval of instability near Ω1 + Ω2 can be obtained with rather high accuracy, compared to the numerical solution. The full set of data for this example can be found in (Ecker, 2003). PARAMETRIC EXCITATION AND VIBRATION SUPPRESSION 171 Parametric excitation frequency ω SE − Pa ra m et er c 02 Ω 2 −Ω 1 Ω 1 +Ω 2 2 2.5 3 3.5 4 −0.2 −0.15 −0.1 Figure 3. Typical stability map for a system as shown in Fig. 2. White areas indicate a stable system, grey areas instability due to self-excitation. Stability threshold is calculated numerically (solid line) and by analytic formulas (broken lines) 6.2. ROTOR SYSTEM The results in this section are obtained from a rather elaborate model of a Jeffcott- rotor with 4 degrees of freedom for the rigid disk and 2-dof at each of the bear- ing stations. Parametric stiffness excitation is created by time-periodic bearing forces, see (Ecker, 2003). Self-excitation forces are acting on the rotor due to in- ternal damping of the shaft. To limit the vibration amplitudes when self-excitation 1 1.5 2 2.5 3 3.5 4 0.5 0.75 1 1.25 1.5 0 0.5 1 1.5 Parameter ν PE−frequency η m ax (x r) Figure 4. Parameter study for a rotor with internal damping and parametric stiffness excitation by the bearing forces 172 ECKER occurs, progressive damping is introduced and consequently the rotor reaches a limit cycle, when becoming unstable. This nonlinear model can only be investigated by numerical simulation, when none of its features shall be neglected or approximated. On the other hand, the case where vibration suppression occurs can be seen directly from the results for the amplitudes, which vanish after a transient period. Figure 4 shows the deflection at the rotor disk station as a function of the rotor angular speed ν and the frequency of the simultaneous harmonic stiffness variation in both bearings. The amplitude of the stiffness variation is about 20% of its nominal value. The region where self-excitation does not create instability is the gray shaded area. Without any PE this region would reach almost to a rotor speed of ν = 1.0). Now with PE tuned to the right frequency, which would be near 1.7, a rather narrow stretch of stability and no amplitudes reaches up to ν = 1.45). This means that in this example the rotor speed could be increased by 50% if PE is employed to cancel self-excited vibrations. 0 0.2 0.4 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Rotational speed νue m ax (|z r|) Figure 5. Rotor amplitudes as a result of unbalance ue, self-excitation due to internal damping and parametric excitation The result in Fig. 5 shows primarily the unbalance resonance peak as a func- tion of the unbalance parameter ue and the rotor speed ν. Parametric excitation is in effect and increases the onset of instability from about ν ≈ 1 to ν = 1.4 at zero rotor unbalance. Obviously and as expected, there is no significant influence of the PE on the resonance peak. This observation holds in both directions and therefore also no adverse effect due to parametric excitation on the unbalance response is PARAMETRIC EXCITATION AND VIBRATION SUPPRESSION 173 Fo rc e A m pl . PE−Frequency ω 50 100 150 200 250 300 350 400 0 10 20 30 40 50 60 Figure 6. Stability map for cantilever beam with periodic axial force excitation. Shaded areas indicate instability. Enhanced damping is obtained in the center of the contour lines near ω = 140 observed. In fact, there is even a minor increase of the stability threshold, as the unbalance increases. 6.3. CANTILEVER BEAM WITH PERIODIC AXIAL FORCE EXCITATION The last example shall point out, that parametric excitation, when adjusted to a non-resonant PE-resonance frequency, introduces additional damping to the sys- tem. This is basically the mechanism why self-excited systems can be stabilized by this method. The system investigated now is a uniform, continuous slender cantilever beam in an upright position. A vertical load of constant direction is applied to the tip of the beam and this load has a time-periodic component. Al- though the force is an external force, the system equations turn out to be those of a system with parametric excitation, see (Ecker et al., 2005). As there is no 0 0.2 0.4 0.6 0.8 1 −0.4 −0.2 0 0.2 0.4 Time T ip D ef le ct io n 0 0.2 0.4 0.6 0.8 1 −0.4 −0.2 0 0.2 0.4 Time T ip D ef le ct io n Figure 7. Vibration signal at the tip of the cantilever beam. (Left) free oscillations without PE, (right) with optimal harmonic axial force excitation 174 ECKER self-excitation in the system, instabilities are only created by resonant parametric excitations. See Fig. 6 for a stability chart of the system, where the parameters of the diagram are PE amplitude and frequency. The damping effect of parametric excitation can be observed as an area of reduced eigenvalues of the monodromy matrix, near ω ≈ 140 as indicated by the contour lines in Fig. 6. Taking the best combination of the force amplitude and the frequency one can study the decay of free oscillations of the free system. In Fig. 7 left, transversal vibrations at the tip of the beam decay only very slowly due to little structural damping. If parametric excitation is applied, how- ever, the situation can be improved quite significantly, as one can see from Fig. 7 (right). References Abadi (2003) Nonlinear Dynamics of Self-excitation in Autoparametric Systems, Ph.D. Thesis, Utrecht University. Blevins R. D. (1977) Flow-induced vibration, Van Nostrand Reinhold, New York. Cartmell M. (1990) Introduction to Linear, Parametric and Nonlinear Vibrations, Chapman and Hall, London. Dohnal F. (2005) Damping of mechanical vibrations by parametric excitation, Ph.D. Thesis, Vienna University of Technology. Ecker H. (2003) Suppression of Self-excited Vibrations in Mechanical Systems by Parametric Excitation, Habilitation Thesis, Vienna University of Technology. Ecker H., Dohnal F., Springer H. (2005) Enhanced Damping of a Beam Structure by Parametric Excitation, Proc. of European Nonlinear Oscillations Conf. (ENOC-2005) Eindhoven, NL. Fatimah S. (2002) Bifurcations in dynamical systems with parametric excitation, Dissertation, Utrecht University. Makihara K., Ecker H., Dohnal F. (2005) Stability Analysis of Open Loop Stiffness Control to Suppress Self-excited Vibrations, J. Vibration and Control 11. Nabergoj R., Tondl A. (2001) Self-excited vibration quenching by means of parametric excitation, Acta Technica ČSAV 46 107-211. Shaw S. W., Rhoads J. F., Turner K. L. (2004) Parametrically excited MEMS oscillators with filtering applications, Proc. 10th Conf. on Nonlinear Vibrations, Stability, and Dynamics of Structures Blacksburg, VA. Tondl, A. (1998) To the problem of quenching self-excited vibrations, Acta Technica ČSAV 43 109-116. Tondl A. (2001) Systems with periodically variable masses, Acta Technica ČSAV 46 309-321. Tondl A. (2002) Three-mass self-excited systems with parametric excitation., Acta Technica ČSAV 47 165-176. Tondl A., Ecker H. (2003) On the problem of self-excited vibration quenching by means of parametric excitation, Archive of Applied Mechanics 72 923-932. Tondl, A., Nabergoj, R., Ecker, H. (2005) Quenching of Self-Excited Vibrations in a System With Two Unstable Vibration Modes, Proc. of 7th Int. Conf. on Vibration Problems (ICOVP-2005) Istanbul, Turkey. Verhulst F. (2000) Nonlinear Differential Equations and Dynamical Systems, Springer-Verlag. Verhulst F. (2005) Methods and Application of Singular Perturbations. Boundary Layers and Multiple Timescale Dynamics, Springer. 175 175–180. A FORMULATION OF MAGNETIC-SPIN AND ELASTO-PLASTIC WAVES IN SATURATED FERRIMAGNETIC MEDIA A. Ünal Erdem Department of Mech. E., Gazi University, 06570, Maltepe, Ankara, Turkey Abstract. In this work our objective is to obtain a mathematical model for the formulation of propagation of magnetic spin waves as coupled with elasto-plastic waves in magnetically saturated ferromagnetic insulators . Elasto-plastic deformations are considered to be infinitesimal (classical theory). Key words: magnetic-spin, elasto-plastic waves, ferrimagnetic media 1. Introduction In view of the two significant shortcomings of the stress-space formulation , we shall adopt the strain-space formulation (Khan and Huang, 1995). At this point, prior to proceeding with plasticity , it will be useful to expose the results of the related portion of magneto-elasticity. For the case of infinitesimal strains, the constitutive equations for magnetically saturated insulators are given by (Eringen and Maugin, 1990). Eti j = ∂Σ ∂ εi j = ci j k l εk l + κk i j µk + 1 2 λi j k l µkµl =E t j i , ( κk i j = κk j i ) (1) bi = −ρ− 10 ∂ Σ ∂ µi = ( − ρ− 10 χi j µ j + κi j k ε j k + λi j k l εk l µ j ) , ( χi j = χ j i ) (2) σ (µ) i j = Ai k µ j, k , (Ai k = Ak i) (3) ti j =E ti j − ρo bi µ j (Balance o f angular momentum) (4) ti j = ci j k l εk l + κk i j µk + 1 2 λi j k l µk µl + ( κi k l εk l µ j + χi k µkµ j + λk l i m εk l µmµ j ) (5) Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 176 ERDEM Σ ≡ ρ0 Ψ ≡ 1 2 ci j k l εi j εk l+ 1 2 χi j µi µ j+κi j k µi ε j k+ 1 2 λi j k l εi j µk µl+Ai j µk , i µk , j (6) with ∑ , being the free energy density for the unit undeformed mass, µ ≡ Mρ magnetization per unit mass, ε infinitesimal strain tensor, t total strain tensor , b local magnetic induction, and σ(µ) magnetic spin stress tensor (Tiersten ten- sor). At the right-hand side of Eq. (6) we can see the elastic energy, magnetic anisotropy energy, piezomagnetic energy, magnetostriction energy, and magnetic spin-exchange energy, respectively. 2. Formulation Strain-space formulation can now be summarized as follows: For any given load- ing history, the boundary between elastic and plastic strain-states is the current yield strain εY . With σe being a fictitious stress obtained by assuming a totally elastic deformation represented by σ e = E ε , in 1-D case, the current stress is given by σ = E ε − σP ,where σ e = E εP (εP , current plastic strain).In 3-D case this is generalized to the expressions σP = C : εP (7) σ = C : ( ε − εP ) = C : ε − σP (8) Here Ck l m n stands for the stiffness tensor in the case of pure linear elastic defor- mations. Accordingly we have, σi j = Eσi j − ρobiµ j = ti j − σPi j � σ j i (9) where ti j is given by Eq. (5). As for the term σPi j we proceed as follows: The yield surface, with the same order of magnitudes involved, will be given by ϕ = ϕ(εk l , µk , σ P k l , αk l , κ) − λ ( µk µk − µ 2 S ) = 0 , (10) where λ is the Lagrange multiplier which takes care of the restriction µk µk = const. , due to saturation. The relation d ϕ = 0 is to be satisfied on the yield sur- face (consistency condition). Here εk l is infinitesimal strain tensor, αk l and κ rep- resenting kinematic and isotropic hardening parameters, respectively. Evolution equations (Khan and Huang, 1995) consist of the Flow rule: dσP = d η dϕ d ε , or • σ P = • η d ϕ d ε , (11) Isotropic hardening: d κ = N ( ε , µ , σP , α , κ ) : d σP ≈ d κ d l √ (2/3) dσP : dσP , (12) MAGNETIC-SPIN WAVES AND ELASTO-PLASTIC WAVES 177 κ = κ (l), l ≡ ∫ √ (2/3) ( d σP d t : d σP d t ) dt , (13) and kinematic hardening: d α = L ( ε , µ , σP , α , κ ) ≈ m C−1 : dσP (m , aparameter) (14) where ≈ signifies the assumption of theory to be that of infinitesimal strains. Use of evolution equations in the consistency condition d ϕ = 0 determines the value of dη as follows. Differentiating d ϕ = ∂ ϕ∂ εk l d εk l + ∂ ϕ ∂ µk d µk + ∂ ϕ ∂ σPk l d σPk l + ∂ ϕ ∂αk l d αk l + ∂ ϕ ∂ κ d κ − λ (2 µk d µk). We note that coefficient of Lagrange multiplier λ does not contribute here, since µk dµk is identically zero due to saturation restriction. Now using Eq.’s (11-14), we get d η = 1 D ( ∂ ϕ ∂ εk l : d εk l + ( ∂ ϕ ∂ µk d µk )) . (15) D ≡ − ∂ϕ ∂σP : ∂ ϕ ∂ ε + m ∂ϕ ∂α : C−1 : ∂ ϕ ∂ ε + √ 2 3 ∂ ϕ ∂ κ ∂ κ ∂ l √ (∂ ϕ / ∂ ε) : (∂ ϕ / ∂ ε) (16) With this value of dη flow rule takes the form of, d σPi j = 1 D [ ( ∂ ϕ ∂ εi j ∂ ϕ ∂ εk l d εk l ) + ( ∂ ϕ ∂ εi j ∂ ϕ ∂ µk ) d µk ] (17) in which all the increments are fully independent. Total stress increment is given by Eq.’s (9), (5) and (17). From differentiation of Eq. (5), substituting from (17) yields d σi j = [ ci j k l − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ εk l ] d εk l + [ κk i j − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ µk ] d µk + 1 2 ( λi j k l + 2 χi k δl j ) d (µk µl) + κi k l d ( εk l µ j ) + λi m k l d ( µm εk l µ j ) (18) Divergence of the total stress can be evaluated as σi j , i = ∂σi j ∂ εk l εk l , i + ∂σi j ∂ µk µk , i + ∂σi j ∂ (µk µl ) (µk µl ), i + ∂σi j ∂ (εkl µr ) (εkl µr), i + ∂σi j ∂ (µm εkl µr ) (µm εkl µr), i (19) 178 ERDEM σi j , i = [ ci j k l − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ εk l ] uk , l i + [ κk i j − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ µk ] µk , i + 12 ( λi j k l + 2 χi k δl j ) (µk µl), i + κi k l ( εk l µ j ) , i + λi m k l ( µm εk l µ j ) , i (20) This last expression is to be used in the equations of motion for the ferrimag- netic material medium. ρ0 ∂2u j ∂ t2 = ρ0 µkb j , k + σk j, k (b, : Local magnetic induction), (21) along with the value of the local magnetic induction given by Eq. (2). As for the magnetic-spin equation of motion consistent with the magnetoelastic formulation, we have, ∂ µi ∂ t = γ εi j k µ j ( Hk + bk + 1 ρ0 σ (µ) r k , r ) (22) in which the third term is supplied by (3). The unknowns are the fields u j and µi. Upon substituting the constitutive equations (2) and (3) into (22); and (2) and (20) into (21), respectively, we get equations of motions ∂ µi ∂ t = γ εi j k µ j Hk − ρ−10 γ εi j k µ j ( χk p µp + κk r p ur , p + λk l r pµlur, p − Ar pµk ,r p ) (23) ρ0 ∂2 ui ∂ t2 = −(χ j rµiµr, i + κ j r sµiur, s i + λ j r s pµi ( µrus, p ) , i + + [ ci j k l − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ εk l ] uk , l i + [ κk i j − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ µk ] µk , i + 12 ( λi j k l + 2 χi k δl j ) (µk µl), i + κi k l ( uk, l µ j ) , i + λi m k l ( µm uk, l µ j ) , i (24) Quasi-magnetostatic field will be governed by ∇ · B = 0, ∇ × H = 0, B ≡ H + M = H + ρ0µ (25) H = −∇Φ , ∇2 Φ = ρ0µk, k or Φ, k k = ρo µk , k (26) Integrating Eq. (26)3 over the space variables and assuming that initial mag- netization is the saturation magnetization of magnitude µs,we get µk = µ o k + ρ −1 0 Φ, k ( µ o k ≡ initial magnetization, ∣∣∣µo ∣∣∣ = µs (27) MAGNETIC-SPIN WAVES AND ELASTO-PLASTIC WAVES 179 In the undisturbed state, µo is parallel to H due to magnetic saturation.). Now a magneto-elastoplastic plane wave, which is a disturbance of initial magnetization µo from the direction parallel to H, propagating in the direction of a unit vector n can be represented by, uk = uk (n · x − c t) and Φ = Φ (nk xk − c t ) (28) (x : spatial position on the wave front) By the use of Eq.’s (28) with the substituting of the spacial and temporal derivatives into Eq. (22), noting also the fact that propagation direction n is per- pendicular to the initial saturation magnetization, namely, µoi ni = 0 , the magnetic spin equation of motion we get cΦ� − ρ−10 γεi j kniχk p µ o jnpΦ � = γ εi j kni [ κk r p + λk l r pµ o l ] µojnpu � r + ( γ εi j kniλk l r pρ−10 µ o jµ o l np ) Φ� u�r (29) with (.)�and (.)� denoting first and second derivatives with respect to the argument of the functions defined by Eq. (31). As for the displacement equation of motion, ρoc2 u�j = −χ j rρ −2 0 nr Φ �Φ� − κ j r s ρ−10 ns Φ � u�r + + λ j r s p ( ρ−10 µ onp Φ� u�s + ρ −2 0 nr np Φ � 2 u�s ) + [ ci j k l − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ εk l ] nlni u�k + [ κk i j − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ µk ] ρ−10 nkniΦ � + 12 ρ −1 0 ( λi j k l + 2 χi k δl j ) ( µoknl + µ o l nk + 2 ρ −1 0 nknl ) ni Φ� +κi k lninl µoju � k + κi k l ni nl ρ −1 0 n j Φ � u�k + κi k l ninl ρ −1 0 n j Φ � u�k + λi m k l ni nl (µojµ o mu � k + ρ −1 0 µ o mn j Φ � u�k + ρ −1 0 µ o mn j Φ � u�k + ρ−10 µ o jnm Φ � u�k + ρ −1 0 µ o jnmΦ � u�k + ρ −2 0 nm n j Φ � 2 u�k +2ρ−10 nm n ju � kΦ �Φ�) (30) These two equations are strongly coupled and highly nonlinear. For further linearization, in (29) we ignore the term Φ� u�r as compared to the others, then take the derivative of both sides with respect to the common argument of Φ and solve this for Φ�� as a function of Φ� and u�r. Setting Φ �� = 0 (meaning to ignore its 3rd 180 ERDEM variation near the others) and solving for Φ� yields, Φ� = Rr u � r , Rr ≡ γ εi j k niµ0jnp ( κk r p + λk l r p µ 0 l ) γ εi j k niµ0jnp ( ρ−10 c −1 0 χk p ) = const. (31) Now ignoring all the nonlinear terms in Eq. (30), we get an eigenvalue prob- lem for a nonsymmetric matris Q, ( Q j r − qδ j r ) u�r = 0, q ≡ ρ0 c2 , c = √ q ρ0 , (32) Q j r ≡ ni nl [ (ci j r l − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ εr l + ρ−10 ( κl i j − 1D ∂ ϕ ∂ εi j ∂ ϕ ∂ µl ) Rr + ( κi r l − λi m rl µom ) µoj + ρ −1 0 ( λi j k l + χi k δl j + χi l δk j ) ( µok + 2 ρ −1 0 nk ) Rr] (33) This is nothing but the acoustical tensor for this particular problem. 3. Conclusion The format of the solution is that, positive eigevalues q(i) of matrix Q yield pos- sible wave speeds c(i) and corresponding eigenfunctions u� (i)r . The displacement field ur is obtained by integrating twice over the phase variable z ≡ n · x−c t . Then µk (x , t) are determined from Eq.’s (33) and (29). Further linearization will, of course, be necesseray when we introduce the yield function ϕ ( εk l, µk, ε p k l, αk l, κ ) , in terms of its arguments, explicitly. Also, most of the ferrimagnetic materials consist of the centrosymmetric cubic and uniaxial, that is, transversely isotropic crystals. Accordingly, various order tensors in the expression of Qi j are to be writ- ten in appropriate forms (Eringen and Maugin, 1990). On the other hand, Q is not a symmetric tensor, and the contribution of the terms involving the yield function ϕwill reflect the effects of plastic deformation, as well as elastic deformation. This explicit investigation necessitates a numerical work on the subject which we postpone until later for a separate study. References Eringen, A. C. and Maugin, G. (1990) Electrodynamics of Continua II, Fluids and Complex Media, Springer Verlag. Khan, A. and Huang, S. (1995) Continuum Theory of Plasticity, John Wiley. 181 181–194. EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES Mustafa Erdik1 and Nurdan Apaydın2 1 University, İstanbul, Turkey 2General Directorate of State Highways, 17th Division, İstanbul, Turkey Abstract. Suspension bridges represent critical nodes of major transportation systems. Bridge failure during strong earthquakes poses not only a threat of fatalities but causes a substantial inter- ruption of emergency efforts. Although wind induced vibrations have historically been the primary concern in the design of suspension bridges, earthquake effects have also gained importance in recent decades. This study involves ambient vibration testing and sophisticated three-dimensional dynamic finite element analysis and earthquake performance assessment of Fatih Sultan Mehmet and suspension bridges in Istanbul under earthquake excitation. Nonlinear time history analysis of 3D finite-element models of Fatih Sultan Mehmet and suspension bridges included initial stresses in the cables, suspenders, and towers in equilibrium under dead load condi- tions. Multi-support scenario earthquake excitation was applied to the structure. Suspension bridges are complex 3-D structures that can exhibit a large number of closely spaced coupled modes of vibration. The large number of closely spaced modes, spatially different ground motion charac- teristics, and the potential for nonlinear behaviour complicate the seismic response of suspension bridges. Spatial variation of ground motion exhibits itself with different ground motions at supports due to the wave passage, incoherence and local site effects. The source of nonlinear behaviour is due to geometric nonlinearity and the presence of cables that can only carry tensile forces. The natural frequencies of vibration and the corresponding mode shapes in their dead load and live load configurations are determined. Displacement time histories and stresses at critical points of the bridges are computed and their earthquake performance under the action of scenario earthquake are estimated. Key words: Earthquake, suspension bridge 1. Introduction The main structural elements of a suspension bridge are: the main cable system, the towers or pylons, the anchorage and the deck (or girder). Suspension bridges are flexible structural systems susceptible to the dynamic effects of traffic, explo- sions wind and earthquake loads. The assessment of the dead and live load effects on the structure, the response of the structure to wind and earthquake loads, and the subsequent design of an optimized lateral and vertical load resisting system, require sophisticated analytical methods. Although the computational power of today’s hardware and software elim- inated most of the analytical difficulties encountered in the assessment of the Vibration Problems ICOVP 2005, c© 2007 Springer. Boğaziçi Boğaziçi Boğaziçi (eds.), . şE. Inan and A. Kırı 182 ERDİK, APAYDIN dynamic response of such structures, the following general issues still remain to be elaborated and developed for on the wind and earthquake response/performance and design of suspension bridges. • The structural system and its characteristics, • The nature of wind forces and earthquake forces, • The computer modelling and the limitations of the software utilized, • Wind tunnel and shaking table experiments • Strong Motion Instrumentation There exists two suspension bridges in crossing the Bosphorus straits: the Bosphous and Fatih Sultan Mehmet Bridges. Past investigations on these two suspension bridges concentrated on: Analytical assessment of dynamic properties; Forced and Ambient Vibration Testing; Wind induced vibration measurements and Structural Vibration (Health) Monitoring. 2. Description of the suspension bridges in İstanbul Bosporus Suspension Bridge, commissioned in 1973, joins the European and Asian Continents through Ortakoy and Beylerbeyi districts of It is a gravity anchored suspension bridge with steel pylons and inclined hangers. The main span is 1,074 m (World rank 12th). It consist of one main and two side spans. Side spans are not suspended and rest on slender columns. The cost of the bridge amounted to USD 200 million. Fatih Sultan Mehmet Suspension Bridge, located about 5km north of the Bosphorus Bridge, also spans the Bosporus strait between sion bridge with no side spans and with steel pylons and double vertical hangers The main span is 1,090 m (World rank 11th). It was completed in 1988 at a cost of USD 130 million. Both bridges were designed by Freeman Fox & Partners. Table 1 provides a summary of the structural characteristics of both bridges. Figures 1 and 2 respectively shows the general structural sections and elevations of the Bosporus and Fatih Sultan Mehmet Suspension Bridges. 3. Full scale testing Vibration characteristics and the dynamic properties of the suspension bridges are important design parameters controlling their wind and the earthquake safety. Full-scale dynamic testing of suspension bridges is a reliable method for the assessment of the free vibration characteristics (i.e. vibration mode shapes, fre- quencies and the associated damping rations), and the calibration of the analytical and finite element models. Furthermore the detection of the changes in these vi- bration characteristics are one of the tools used for structural health monitoring İstanbul İstanbul. Hisar (European Side) and Kavac k (Asian Side). It is a gravity-anchored suspen-ı EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 183 Figure 1. Bosphorus Suspension Bridge General Structural Sections and Layout purposes. The size and the very low frequencies of vibration of the suspension bridges disqualifies the use of forced vibration techniques (such as harmonic ex- citation or impulsive loading) commonly used for other bridges and viaducts.One of the best methods utilized to assess the dynamic characteristics of such massive structures is the measurement of the structural response to the ambient excita- tions, such as, wind and traffic noise. The structural vibrations caused by such excitations are termed the “Ambient Vibrations”. Since ambient vibration tests provide the “output-only” type measurements, the system identification techniques based on Fourier Spectral Analysis tech- niques are generally employed. In the Fourier Spectral Analysis technique the modal vibration frequencies and the shapes are obtained on the basis of Fourier Amplitude and Phase Spectra obtained at measurement points. The so-called “half- power bandwidth” procedure using the Fourier Amplitude Spectra at the identified modal frequencies of vibration is employed to obtain the modal damping charac- teristics. Recent applications of system identification also include the so-called Natural Excitation Techniques paired with the Eigensystem Realization Algo- rithms (NexT-ERA), initially developed by James et al. (1993). The basic principle behind this method relies on the fact that the theoretical cross-correlation func- tion between two response data has the same analytical form as the impulse re- sponse function of the bridge. This technique is claimed to be especially suited to 184 ERDİK, APAYDIN TABLE I. Comparison of the Structural Characteristics of the Bosporus and Fatih Sultan Mehmet Suspension Bridges. BOĞAZİÇİ FATİH SULTAN MEHMET Span Main Span 1074 m Main Span 1090 m Asian Side Span 255m European Side Span 231 m Deck Width: 28 m. (33.4 m total) 33.8 m (39.4 m total) Substructure 2 Steel Towers (165 m) 2 Steel Towers (107 m) 2 Anchorages 2 Anchorages Spread Foundation Spread Foundation Dead Load per Unit Length: 143 kN/m 216 kN/m Areas per Cable: 0.205 m2 0.365 m2 Gross x-sectional area 0.85 m2 1.26 m2 Deck Moment of Inertia Ixx=1.24 m4 Ixx=1.73 m4 Iyy=63.6 m4 Iyy=129.2 m4 Modulus of Elasticity of Cables 193 kN/mm2 205 kN/mm2 Sag: 93 m+7.1 m 91 m+6.3 m Hanger Geometry: Inclined Vertical (Double) Clearance from the sea level 64 m 64 m extract modal parameters from a structure with closely spaced modes. The “ambi- ent vibration” tests which used the ambient and non-measurable excitation (such as wind and traffic) as the source have been successfully applied to several impor- tant suspension bridges all over the world such as Vincent Thomas (Abdel-Ghaffar and Housner 1978), Golden Gate (Abdel-Ghaffar and Scanlan 1985), Roebling (Ren et al., 2004) and Humber (Brownjohn et al., 1987). Similarly, in Turkey vi- bration tests were conducted in Bosporus Bridge (Petrovski et al., 1974; Tezcan et al., 1975; Brownjohn et al.,1989, Erdik et al.,1988; Beyen et al., 1994 and Kosar, 2002). Table 2 and Table 3 provides the empirical frequencies and general shapes associated with first several modes of vibration of Bogazici and Fatih Sultan Mehmet Suspension Bridges. 4. Wind induced vibration measurements The response of the and Fatih Sultan Mehmet Suspension Bridges to strong winds on February 12-13, 2004 were recorded by the strong motion ac- celerograms installed in the north and south sides at the center of the main span of the bridges at every half hour for 5 minutes. The wind speeds ranged between 60 to 90 km/hr. The RMS value of the accelerations vary between 0.02g and 0.06g in the lateral and 0.1g and 0.3g in the vertical direction. Whereas peak ı2003) and in Fatih Sultan Mehmet Bridge (Brownjohn et al.,1992 and Apayd n, Boğaziçi EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 185 Figure 2. Fatih Sultan Mehmet Suspension Bridge General Structural Sections and Layout TABLE II. Experimental Modal Frequencies of Vibration of Suspension Bridge Ranges in Kosar (2003) refer to light and heavy traffic conditions Mode Shape Frequency (Hz) Frequency (Hz) Erdik et al., 1988 Kosar, 2003 First Lateral Symmetric 0.072 0.073-0.063 First Vertical Asymmetric 0.13 0.163-0.155 First Vertical Symmetric 0.16 0.171-0.169 Second Vertical Asymmetric 0.19 First Lateral Asymmetric 0.22 0.215-0.22 The damping ratios in these modes vary between 7% (first mode) and 4% (others). accelerations, velocities and displacements reached respectively to 0.1g, 0.1m/s and 0.2m in the horizontal and 0.6g, 0.2m/s and 0.3m in the vertical direction (Kaya and Harmandar, 2004). Figure 3 provides the Fourier Amplitude Spectra in transverse, longitudinal and vertical directions at the center of the main span (north side) for the Suspension Bridge. The peaks of the spectra can be associated with the modal frequencies of vibration. Boğaziçi Boğaziçi 186 ERDİK, APAYDIN TABLE III. Experimental Modal Frequencies of Vibration of Fatih Sultan Mehmet Suspension Bridge Mode Shape Frequency (Hz) Frequency (Hz) Apayd n, 2002 Brownjohn et al., 1992 First Lateral Symmetric 0.068 0.076 First Vertical Asymmetric 0.102 0.108 First Vertical Symmetric 0.118 0.125 Second Vertical Asymmetric 0.154 0.145 Figure 3. Fourier Amplitude Spectra of wind response at mid-span of Suspension Bridge (After Kaya and Harmandar) 5. Analytical studies In addition to the in-plane and sway vibrations of the cable subsystem and due to the simultaneous occurrence of lateral, vertical and torsional modes, dynamic behaviour of suspension bridges exhibit great vibrational complexities. Closed form solutions to the governing dynamic equilibrium equations in lateral direction are studied by Silverman (1957), Selberg (1957) and Moisseif (1933) before the computer era. Assuming inextensible cables and suspenders, perfectly flexible deck and rigid towers, the deformations of the deck follows the cable deflections. The frequencies of free vibration of a simply supported parabolic catenary with a mass distribution equal to that of the total axial mass distribution of the suspension bridge approxi- mately yields the frequencies of vibration of the pure vertical modes of the bridge. ı Boğaziçi EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 187 Pugsley (1949) provides closed form expressions for these frequencies in terms of cable length, cable sag, mass per unit length and horizontal tension. On the basis of such a simplified approach the frequencies of vibration for the first three vertical vibration modes can be calculated as 0.12, 0.16 and 0.23Hz (Beyen et al., 1994). These compare well with the empirical results listed in Table 1. A two dimensional finite element solution of the continuum approach was developed by Abdel-Ghaffar (1978) for the lateral vibration of suspension bridges. This simple approach is based on the Hamilton’s Principle with the basic assump- tions of: (1) vibration amplitudes are sufficiently small to remain in linear ranges, (2) Coupling between lateral and torsional modes are ignored and (3) the cable, ends are immovable. Due to the high lateral stiffness of the towers, supports of the cable and the deck are also assumed to be immovable. Analytically derived energy expressions are then used to express the stiffness and consistent mass matrices for the finite element application. Erdik et al., (1988) computed the lateral modes of vibration of Suspension Bridge using only eight elements for the span. The analytical results were fount to be in good agreement with the experimental values. The modal vibration frequencies and the shapes of the Bosporus Suspension Bridge were determined on the basis of a three-dimensional finite element model in Beyen et al., (1994) and Kosar (2003). Beyen et al., (1994) used LUSAS (1993) finite element program where, the bridge deck is modelled using equivalent shell elements and the effects of axial forces on the stiffness of the cables are considered through the use of equivalent frame elements. The model neglects the towers and the side spans. However, the boundary conditions are chosen to simulate the effect finite element analysis program (Wilson and Habibullah, 1989). To account for the geometric nonlinearity P-Delta effect is modeled using the geometric stiffness matrix, which incorporates the influence of axial forces on transverse stiffness. An estimate of the appropriate magnitude of tension was obtained from a static analysis of the bridge under self-weight. The resulting tension was then added to the suspension and hanger cables as a directly specified P-Delta axial force. The finite element model includes the following elements and boundary conditions: All portal beams and towers are represented as frame elements; Constraints are used to enforce certain type of rigid-body behavior, to connect together different parts of the model and to impose certain types of symmetry condition; Deck is defined as a frame element and body constraints are used to minimize relative displacement in the deck elements and the camber of the deck, geometry of deck and towers are defined in the model; Main cables and hangers are modeled with frame elements with hinged ends and; At the deck-tower connection the rocker frames are connected to the deck with a hinge with a moment release only in longitudinal direction. Kosar (2003) also utilized SAP-90 finite element analysis program (Wilson and Habibullah, 1989). The finite element model incorporates Boğaziçi of the deck and cable supports at the towers. Apayd n (2002) utilized SAP-90ı 188 ERDİK, APAYDIN the rocker frames are connected to the deck with a hinge with a moment release only in longitudinal direction. To provide an example, the computed modal vibra- tion shapes of the Suspension Bridge are provided in Figure 4 for the first six lateral vibration modes. In connection with the earthquake performance and retrofit studies of the and Fatih Sultan Mehmet Suspension Bridges, Japanese Bridge and Structure Institute (JBSI, 2004) conducted analytical studies to gain an understanding of the dynamic seismic behavior of the bridges. Figure 5 illustrates the vibration shape of the first three vertical modes of free vibration for the Suspension Bridge. Figure 4. lateral vibration modes (After Beyen et al., 1994) For the earthquake response of suspension bridges different ground motion at sive assessment of the response of the Fatih Sultan Mehmet Suspension Bridge under excitation by spatially varying (asynchronous and/or travelling) ground motion. Bridge responses to incoherent earthquake ground motions are due to a dynamic portion and differing support movements (denoted as “pseudo-static” displacements). It is computationally equivalent to decompose the total (absolute) considerations similar to those of Apayd n (2002). At the deck-tower connectionsı Boğaziçi Boğaziçi 6. Earthquake response analysis Computed modal vibration shapes of the Suspension Bridge for the first six each support may need to be specified. Apayd n (2003) provides a comprehen-ı Boğaziçi Boğaziçi EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 189 Figure 5. Vibration shape of the first three vertical modes of free vibration for the Suspension Bridge (After JBSI, 2004) displacement into pseudo-static displacement and relative (vibrational) displace- ment resulting from inertial effects. Pseudo-static displacements are those re- sulting from the static application of support displacements at any time without inertial effects. For the analysis a unit displacement (in each direction-vertical, longitudinal, transverse) is applied to the bridge at one of the ground points and its static response displacement vector is calculated. under spatially varying ground motion the ground motions obtained during 1979 Imperial Valley earthquake at the El Centro Array Stations 7 and 6 are applied respectively to the east and west tower and anchorage foundations. El Centro Array Stations 7 and 6 are located at each side of the Fault and represent extensive spatial variance in a distance comparable to the length of the bridge. The response of the Fatih Sultan Mehmet Suspension Bridge to travelling ground motion is analyzed using an artificial earthquake record as the input. Input motion is selected to represent a full cycle displacement pulse with a period of 1.2s. For this analyses, the ground motion is assumed to travel from Asia to Europe. The lateral response at the top of the European and Asian towers are illustrated in Figure 6 for a phase difference of 1.00 s (propagation velocity of about 1000m/s). 7. Structural health monitoring Structural health monitoring can be defined as the diagnostic monitoring of the in- tegrity or condition of a structure. Early detection of damage or structural Boğaziçi To provide an example Apayd n (2003) studied the response of the bridgeı 190 ERDİK, APAYDIN Figure 6. Lateral Displacement at the top of the Asian and European Towers of the Fatih Sultan Mehmet Suspension Bridge excited by a full cycle displacement pulse with a 1s phase difference degradation prior to local failure can prevent catastrophic failure of the system. The system should be able to automatically detect, locate and assess structural damage anywhere within the bridge system (health monitoring), and to commu- nicate the status (alerting) to responsible authorities. The diagnosis techniques generally includes finite-element modelling of the suspension bridge, measure- ment of strong wind, earthquakes and traffic loadings, analysis of the ambient vibration data through the use of the appropriate system identification techniques and selection of sensors and techniques to detect localized damage and defects. A vibration monitoring system encompassing 12 tri-axial acceleration transducers and a data acquisition unit is installed along the Asian half span of the Fatih Sultan Mehmet Suspension Bridge (Apayd n and Erdik, 2001; Apaydin, 2002). A general layout of the system is illustrated in Figure 7. 8. Earthquake performance and rerofit investigations An earthquake response and performance analysis of the and Fatih Sul- tan Mehmet Suspension Bridges was performed by Japanese Bridge and Structure Institute (JBSI, 2004) for the General Directorate of State Highways (Turkey) to determine the seismic vulnerability and retrofitting requirements of the bridges. In this connection comprehensive analytical studies were also conducted to gain an understanding of the dynamic seismic behavior of the bridges. The analysis was performed using detailed three-dimensional models that included geometric and material non-linearity and soil-structure interaction. The performance-based ı Boğaziçi EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 191 Figure 7. Structural Health Monitoring System installed on Fatih Sultan Mehmet Suspension Bridge seismic analysis and retrofit design provides the means for designing structures with measurable levels of structural response. It leads to structures with seismic load resisting systems of specified strength, stiffness and energy dissipation ca- pacity. For earthquake resistant retrofit analysis and design the following criteria were considered: Under exposure to the Functional Evaluation Earthquake (FEE) Ground Motion the damage level will be minimal (essentially elastic performance) and the functionality of the bridge will continue without interruption. This ground motion refers to the high probability earthquakes that can affect the structure one or twice during its lifetime. This earthquake is generally associated with a 50% probability of exceedance in 50 years. Due to specific seismo-tectonic the FEE earthquake will be taken as the site-specific ground motion that would re- sult from the Mw=7.5 scenario earthquake on the Main Marmara Fault. Under exposure to Safety Evaluation Earthquake (SEE) ground motion only repairable damage is allowed, such that, the damage can be repaired with a minimum risk of losing functionality without endangering and lives. In consideration of re- gional earthquake occurrences this level of ground motion will be associated with 192 ERDİK, APAYDIN a site-specific probabilistic ground motion associated with a 2% probability of occurrence in 50 years. The performance criteria for both suspension bridges under SEE require that: No damage to main and hanger cables (within tensile strength) No slipping of cable clamps Light damage in Towers with no local buckling No damage at rocker bearings (within yield point) No damage due to collision between tower and deck No damage at expansion joints. Although both suspension bridges were originally designed for much lower earth- quake loads under SEE earthquake (Peak Ground Acceleration of about 0.4g- 0.5g) they exhibited very good performance. The retrofit actions recommended for the main spans of both suspension bridges consisted of the reinforcement of the tower legs below the portal beams (for collision between deck and tower) and installation of falling down prevention devices (cables), and reinforcement of the rocker bearings. At the Asian and European joints of the approach viaducts (side spans) of the Suspension Bridge the retrofit actions will encompass replacement of the elastomeric bearings, construction of new diaphragms, and installation of hysteretic dampers. 9. Comments on the earthquake response of the suspension bridges Following general comments on the dynamic characteristics and the earthquake response of the suspension bridges can be made: Suspension bridges are complex 3-D structures that can exhibit a large number of closely spaced coupled modes of vibration. Time domain analysis is the most rational method for the earthquake response analysis. The characterization and effects of variable support input motions and the influence of nonlinearities in suspension bridges are two areas which are not well understood and which will require significant additional research. For suspension bridges exposed to strong earthquake motions, significant changes in the initial bridge geometry will result in geometric nonlinearities as the stiffness of the suspension cables and deck system change appreciably with large displacements of the structure. Once the analysis is carried out under the deformed geometry (under dead load) the variation of geometry due to lateral earthquake loads are quite negligible and the total system can be treated without consideration of the any geometric non-linearity. Theoretical analysis of the free vibration of suspension bridges indicate that the modes of the structure can be divided into three groups with respect to the dominance of the respective displacements. In the first group the displacements Boğaziçi EARTHQUAKE RESPONSE OF SUSPENSION BRIDGES 193 of the deck is important (Deck Modes), in the second the displacements of the cables are important (Cable Modes) and in the third group the displacements of the towers become important (Tower Modes). The deck modes can also be classified as symmetric and asymmetric or as lateral, vertical and torsionally dominant. A relatively large number of modes may be necessary to obtain a reasonable representation of the response. Studies reported herein and the similar ones in- dicate that finite element analysis techniques can be used satisfactorily for the analytical assessment of these mode shapes, frequencies and the participation factors. Earthquakes have a complicated interaction with the whole structure. For ex- ample vertical vibrations of these structures are induced to a great extent by the lateral shaking in addition to the vertical excitation. In vertical modes of vibration the earthquake induced cable tensions may reach considerable values. In coupled vertical-tensional vibrations, either large amplitude vertical or large amplitude torsional vibration may dominate the mo- tion. The first symmetric torsional modes appear to dominate the torsional re- sponse because of its high participation factor. It becomes necessary to consider asynchronous excitation at the each and every contact point of the structure with the ground. This strong ground motion should be specified as three-axial and all possible kinds of wave fields (especially SH, Rayleigh and Love) should be considered in the analytical techniques used in their simulation. Under multi-support excitation at both of the towers in the lateral direction, different peak displacements are observed. For the travelling ground motion, as the phase difference increases, the displacements of the deck in the lateral direction become smaller. At the towers, as the phase difference increases, peak displace- ments in the lateral direction increase. The two towers may move in different directions. For suspension bridges located in the fault near-field where large ground dis- placement pulses and deterministic wave radiation patterns are known to be a possibility, special ground motion simulations should be considered. References Abdel-Ghaffar A. M., Housner G. W. (1978) Ambient Vibration Tests of Suspension Bridges,, Journal of Engineering Mechanics Division, ASCE, No.EMS, Proc. Paper 140b5, 104 983-999. Abdel-Ghaffar A. M., Scanlan R. H. (1985) Ambient Vibration Studies of Golden Gate Bridge: I-Suspended Structure, J. Engrg. Mech., ASCE, No. EM4, 111 463-482. Apaydin N., Erdik M. (2001) Structural Vibration Monitoring System for the Bosporus Suspension Bridges in Strong Motion Instrumentation for Civil Engineering Structures (NATO Science Series), Ed.By. M.Erdik et al., Springer-Verlag New York, LLC, 373. Apaydin A. (2002) Seismic Analaysis of Fatih Sultan Mehmet Suspension Bridge, Ph.D. Thesis, Department of Earthquake Engineering, Bogazici University, Istanbul, Turkey. 194 ERDİK, APAYDIN Beyen K., Uckan E., Erdik M. (1994) Ambient Vibration Investigation of the Bogazici Suspension Bridge, Istanbul, Turkey, Proc. Earthq. Res. Construction and Design, Balkema, Rotterdam. 915-922. Brownjohn J. M. W., Dumanoglu A. A., Severn R. T., Taylor C. A. (1987) Ambient Vibration Survey of the Humber Suspension Bridge and Comparison with Calculated Characteristics, Proc. Instn. Civ. Engrs. 83 561-600. Brownjohn J. M. W., Dumanoglu A. A., Severn R. T. (1992) Ambient Vibration Survey of the Fatih Sultan Mehmet (Second Bosphorus Suspension Bridge), Earthquake Eng. Struct. Dyn. 21 907-924. Erdik M., Uckan E. (1988) Ambient Vibration survey of the Bogazici Suspension Bridge, Report No: 89-5, Department of Earthquake Engineering, Bogazici University, Istanbul, Turkey. Japanese Bridge and Structure Institute - JBSI (2004) Project Reports and Basic Design Documents for the project entitled “Seismic Reinforcement Of Large Scale Bridges In Istanbul”, prepared for the General Directorate of State Highways, Turkey. Kaya Y., Harmandar E. (2004) Analysis of Wind Induced Vibrations on Bogazici and Fatih Sultan Mehmet suspension Bridges, Internal Report, Department of Earthquake Engineering, Bogazici University, Istanbul, Turkey. Kosar U. (2003) System Identification of Bogazici Suspension Bridge, M. Sc. Thesis, Department of Earthquake Engineering, Bogazici University, Istanbul, Turkey. LUSAS (1993) Finite Element Analysis Computer Program, Finite Element Analysis Ltd., Surrey, U.K. Petrovski J., Paskalov T., Stojkovich A., Jurokovski D. (1974) Vibration Studies of Istanbul Bogazici Suspension Bridge, Report OIK 74-7, Institute of Earthquake Engineering and Engineering Seismology, IZIIS, Skopje, Yugoslavia. Ren W. X., Harik I. E., Blandford G. E., Lenett M., Baseheart T. M. (2004) Roebling suspension bridge. II: Ambient testing and live-load response, Journal of Bridge Engineering 9 119-126. Tezcan S., Ipek M., Petrovski J., Paskalov T. (1975) Forced Vibration Survey of Istanbul Bogazici Bridge, Proc. 5th ECEE, Istanbul Turkey 1.2. Wilson L. E., Habibullah A. (1989) SAP90: A Series of Computer Programs for the Static and Dynamic Analysis of Structures, Computers and Structures Inc., Berkeley, CA., USA. 195 195–204. THE MICROWAVE SENSOR OF SMALL MOVING FOR THE ACTIVE CONTROL OF VIBRATIONS AND CHAOTIC OSCILLATIONS MODES ∗ V. A. Fedorov1, S. M. Smolskiy1, A. V. Mizirin1, V. V. Shtykov1, A. V. Fomenkov1 and S. M. Kaplunov2 1Moscow Power Engineering Institute (Technical University), Russia 2Mechanical Engineering Research Institute of Russian Academy of Sciences, Russia Abstract. In the report results of practical development of the modeling vibrations test bench and the experimental check of a new class of vibrations and the midget moving sensors are given on the basis of the microwave radar, capable to fix the object moving in space with the high linear resolution up to 10-6 m. Such sensors was named “Pulsar” and authors for the first time used it for remote non-contact registration of fluctuations of a person body sites surface under influence of various physiological processes - breath, pulse blood capillary filling, clonus, acoustic vibrations etc. Simultaneously with the sensor development laboratory researches on modeling vibrations test bench were carried out with the excitation by various complex signals. At work with biological objects the majority of the signals registered by microwave radar- tracking sensor “Pulsar” and caused by physiological processes, belong to a class of the signals most adequately described by the theory of nonlinear dynamic chaos (Glass and Mackey, 1991), (Moon, 1987). Therefore authors have put before themselves the purpose of creation of the com- puter analysis method of the signals received on the “Pulsar” sensor output at its any application, including at research of functional conditions of various biological objects. The offered report is devoted to results of vibrations studying in mechanical and biological objects with the help of the radar-tracking sensor of small movings. The specified purpose is achieved with the help of computer mathematical model of chaotic signals, transfer of these signals on the vibrations test bench activator and registration of the vibrations amplitude by remote non-contact “Pulsar” system. Key words: microwave, active control, chaotic oscillation 1. The basic preconditions of researches At present chaotic signals draw the increasing attention of researchers in the di- versified areas of a science and engineering. Systems, developing such signals, are studied as with the help of classical methods of the nonlinear oscillations theory, and with the help of the new approaches using such concepts as chaotic process, “strange attractor”, a bifurcation mode, a set fractional dimension etc. Therefore ∗ Not presented Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 196 FEDOROV, SMOLSKIY, MIZIRIN, SHTYKOV, FOMENKOV, KAPLUNOV development of devices and methods of precision registration and the analysis of small moving with the help of the sensors, which are not deforming these small movings, is the important and actual problem of researchers. Most fully these requirements are answered with the non-contact sensors working in a microwave range. Such sensors do not depend on object of studying, do not render physical influence on it, can function on appreciable distance from an object during mea- surement, they are compact, have small power consumption, are capable to work continuously long time. They are especially perspective at research of functional conditions of alive objects, including the person as for them the usual clothes are not an obstacle and besides researches can be carried out on real objects during their active vital phase and in real time. It is necessary to note, that microwave sensors are easily integrated to a com- puter which can undertake a significant role in processing the radio signals re- flected from object and in the automatic analysis of measurements results with the analysis results output in real or quasi-real time. During researches authors have applied the complicated chaotic signal which distinctive feature is its rather complex mathematical description and technically complex formation. Physiological processes in alive organisms are so complex, that very seldom it is possible to give their unequivocal theoretical interpretation. In these conditions in an arsenal of the researcher it is important to have mathematical models of physiological processes with which help it would be possible to receive a pic- ture of complex interactions of separate sub-systems of an organism. The present report is devoted exactly to this question. 2. Computer modeling of chaotic process One of ways of chaotic process modeling with the help of a computer is a process examination on low frequency. This way has the next two problems. The first one is a complexity of behavior in time of the signals reflected from biological objects with account of their individual specificity and characteristic features for different patients. The second one is a necessity of transfer of the low-frequency chaotic process, modeled by a computer, in microwave area (on the carrying frequency of “Pulsar” equal 60 GHz) with the subsequent allocation, a filtration and an amplification of the signal low-frequency Doppler component, that is necessary at synthesis of mathematical model of real physiological processes. These signal transformations demand high linearity and a minimization of noise parameters of a signal passing path. Such necessary quality requirements of send-receive and amplifying paths of “Pulsar” sensor were successfully achieved. Since the reflected signals with chaotic structure from biological objects are difficult to simulate, it was necessary to fulfil a measurement technique on ACTIVE CONTROL OF VIBRATIONS 197 mechanical vibrations test bench, having excited its mechanical vibrations by a signal, synthesized on a computer with the help of mathematical model of chaotic process. In the theory of nonlinear fluctuations there are examples of mathematical modeling of signals with chaotic dynamics (Chernobaev, 2002), (Tomashevskiy, 2003). As such examples systems with phase control which represent on structure the system of automatic control with a phase feedback can quite serve. For ex- periments with “Pulsar” the computer model of chaotic fluctuations was used on the basis of the theory of electronic amplifiers with phase control (Chernobaev, 2002). At first the opportunity of chaotic signals generating with the help of ordi- nary electronic amplifiers was analyzed. According to it we shall enter, following (Chernobaev, 2002), the general name for researched system: the amplifier with phase control (PCA), generalized block diagram of which is presented on Figure 1. In this figure it is shown, that on an input of an amplifying path the input quasi- harmonic signal u1(t) is acting (here amplitude and phase functions U1(t), ϕ(t) are assumed slow, i.e. poorly varying for the period of carrying frequency ωc), and on an PCA output a target quasi-harmonic signal u2(t) is developed . The main reason causing of an oscillations phase deviation ϕ2 on an PCA output compared to a phase ϕ1 of an input signal, the central frequency detuning ξosc of the oscillatory circuit, included in phase discriminator PD, is concerning of an input signal frequency. In the PCA model submitted by the block diagram on Figure 2, the detuning ξosc plays the role of the basic destabilizing factor. Figure 1. Structural scheme of an amplifier with phase control In a mode of the out- put phase stabilization the PCA operates on a princi- ple of an automatic control system, reducing the cur- rent mismatch of an out- put phase ϕ2 relative to the input one ϕ1 which is measured with the help of phase discriminator (PD). Produced in FD the mis- take signal as slowly vary- ing voltage e(t) is used by a control circuit (CC) for regulation of tuning fre- quency of amplifying path with the help of the special frequency controller (FC), for example, a varactor. Let’s take advantage, following (Chernobaev, 2002), mathematical models of an examined path as the symbolical abbreviated differential equations of the 3-rd 198 FEDOROV, SMOLSKIY, MIZIRIN, SHTYKOV, FOMENKOV, KAPLUNOV order minimally necessary for occurrence in examined system of dynamic chaos interesting us. Let on a system input the signal of a voltage operates u1(t) = U1(t) cos [ωct + ϕ1(t)] = Re[U1(t)e jωcte jϕ1(t)], (1) and on an output a signal with the same frequency operates u2(t) = U2(t) cos [ωct + ϕ2(t)] = Re[U2(t)e jωcte jϕ2(t)]. (2) Let’s assume, that complex amplitudes of input U1(t)e jϕ1(t) and output U2(t)e jϕ2(t) signals are the functions of time slowly varying for the period of high frequency Tc = 2π/ωc, and we shall connect these complex amplitudes through symbolical transfer factor of an amplifying path: U2(t)e jϕ2(t) = K[p, ξ?, ξ(t)] U1(t)e jϕ1(t), (3) where p ≡ d/dt is the differentiation operator of slowly varying functions, ξosc is the own frequency detuning of the PCA oscillatory system concerning frequency of an input signal, ξ is the resulting detuning inserted in PCA from FC side. Expression in the operational form for inserted detuning will look so: ξ = ΩW(p)F(ϕ1, ϕ2,U1,U2), (4) where F is the normalized PD characteristic, W(p) is the transfer factor of the control circuit, dependent on the differential operator, Ω is the area of stable frequency detuning of system. Expressions (3) and (4) are the differential equations which have been written down in the operational form which describe the PCA operation, and the equation (3) is the complex one, i.e. it breaks up to two real equations. Three real equations allow the determination of three unknown time functions of a problem: a phase and amplitude of an output signal and the inserted frequency detuning. Studying of transient modes has the big importance at revealing dynamic properties of system as really the amplitude U1(t) and a phase ϕ1(t) of an input signal bear the helpful information on vibrations of mechanical system or on a functional condition of the patient for our device. Research of transients is meant a supervision over processes of an establishment in time of amplitude U2(t) and a phase ϕ2(t) of the output oscillations, occuring in the constructed models at various initial conditions and at values of systems parameters. Such dynamic processes are usually displayed in “phase space” of system with the indication of coordinate axes and grid where time is the independent parameter varying along integrated curve of the initial differential equations system. It appears that in the described system at quite certain combinations of pa- rameters there can be repeated chaotic oscillations. The received mathematical ACTIVE CONTROL OF VIBRATIONS 199 model of this chaotic process has permitted to design and make measuring vibra- tions test bench with chaotic processes and to carry out measurements, using the following procedure. First with the help of a computer and the appropriate mathematical methods the low-frequency chaotic process in the digital form with use of the equations (1)- (4) is modeled. Parameters of system and, hence, parameters of chaotic process can be changed during measurements. Then with the help of a computer sound card and the special software this digital process is transformed on an exit of digital-to-analog converter (DAC) to a real electric signal after that this signal is passed through the linear amplifier and moves the vibrations test bench activator. In result this test bench starts to make low-frequency fluctuations according to a signal received on mathematical model. At the second stage modeling vibrations test bench, excited by a low-frequency chaotic signal, is irradiated from distance of 1-2 m with the microwave short- range radar “Pulsar” working in a millimeter wave band. The signal reflected from a vibrating surface, which phase changes under the law set by synthesized chaotic process, accepted by the radar antenna and due to the subsequent fre- quency transformation the initial chaotic signal can be transferred in any area of a spectrum. Taking into account, that properties of a chaotic signal of computer model can be changed at the request of the experimenter, we receive the mechanism of technical formation of a wide class of regular and chaotic processes and their carry to any frequency area necessary for experiments, i.e. a way of creation of some generator of real chaotic signals. Thus, the first object in view is achieved - to create a method of a chaotic signal formation by computer means and to compare its key parameters (for ex- ample, the time form, auto correlation function of process and its phase portrait) to the same parameters received as a result of signal processing in “Pulsar” radar. Thus it is supposed that at concurrence of the specified characteristics of initial chaotic process and characteristics of process, past through all radar transforma- tion blocks, a validity of carry of chaotic process in a radio range it will be proved and remote sensor “Pulsar” can be with good reason used as the non-contact sen- sor of the signals carrying the undistorted information on physiological processes in an alive organism. 3. Displays of a functional condition of complex system to a phase plane Researches on studying behavior of complex objects including alive systems (Glass and Mackey, 1991) with application of the mathematical approach of non- linear dynamic processes have shown, that in a normal functional condition in characteristics of their behavior is present an irregular component with a high de- gree of complexity. On the other hand, at aging and deterioration of any system the 200 FEDOROV, SMOLSKIY, MIZIRIN, SHTYKOV, FOMENKOV, KAPLUNOV well defined periodicity of characteristics accompanying with some “decrease” of a randomness degree and a complexity degree of dynamics of investigated para- meters are shown. Thus, in the complex hierarchical system, normally function- ing, characteristics of its behavior can be submitted by chaotic processes. In com- plex system which parameters have deviated norm as have shown experiments, the obvious periodic component gradually superseding chaotic characteristics of its behavior is shown. The method of phase space allows the description of a phase portrait of ana- lyzed dynamic system if one spatio-temporal sequence as any parameter is known only, describing behavior of system. The analysis of a phase portrait permits to determine the type or prominent features of system dynamics. Authors carry out the description of a measurement method of the vibrations test bench oscillations amplitude with the help of the remote radar-tracking sensor of a millimeter range which construction is discussed below. This test bench sim- ulates complex oscillations in space according to the given mathematical model of oscillatory (chaotic) process and an active test bench surface which points are oscillating and investigating, makes moving to space with chaotic movement parameters. To register the moving character and amplitude of test bench surface points by a traditional way, it is necessary to arrange moving sensors on a researched surface. For this purpose strain gauges (accelerometers) are usually used. In a case when it is necessary to place some several strain gauges, there are complexities with fastening contact conductors and gauges on an active surface of test bench. Authors offer to use as researched model of chaotic vibrations a usual powerful loudspeaker with big diffuser in the sound coil of which the chaotic process of an electric current generated by a sound card of a personal computer is created and it creates with usage of the mathematical model a researched chaotic signal. Under action of a chaotic sound current a loudspeaker diffuser makes chaotic vibrations which further are fixed by any moving sensor. Authors used as measuring system the microwave radar-tracking sensor of high resolution on space (device “Pulsar”) which can be arranged on distance from 0.1 up to 5.0 m from a surface of the modeling vibrations test bench and to investigate dynamic properties of any cho- sen point on it. The output information signal of the radar-tracking sensor through a cable or the radio channel goes to a computer on which monitor it is possible to observe in a real time mode results of measurement process and the analysis of registered moving of a test bench surface. 4. Initial preconditions of mathematical modeling In laboratory experiment with research of vibrations test bench parameters the re- mote millimeter range radar-tracking sensor of spatial moving “Pulsar” developed ACTIVE CONTROL OF VIBRATIONS 201 in MPEI with the resolution on space about 1 micron was used. The block diagram of this experiment is given on Figure 2. The personal computer on this circuit forms in software the mathematical model of chaotic process and directs an output electric signal on the vibrations test bench driver. The device “Pulsar” irradiates a vibrating element, perceives the reflected radio signal and at first transfers by a radar a signal to the necessary area of a spectrum, and then processes a signal in the same computer and creates on a monitor screen a phase portrait of the chaotic process accepted by a radar. Figure 2. The block diagram of the experiment The method of phase space is widely used at the analysis of complex non- linear dynamic systems as it gives an evident picture of system behavior. Be- sides well developed ap- proaches of this method can be used for interpre- tation of research results, for example, a system lo- cal stability. On Figure 3 one of possible mathematical so- lutions of the differential equations is given, de- scribing behavior of an output signal of the circuit, shown on Figure 1, as the diagram of time function for a case when the amplitude and the phase of the output process have obviously expressed chaotic dynamics. On Figure 4 the auto correlation function of an output signal in this complex mode is given, and on Figure 5 the phase portrait of this process is shown. Figure 3. A chaotic process in time Figure 4. Auto correlation function Radar Antenna A Radiation diagram Excitaition system for shaking table Model shaking table The program which is carrying out registration of voltage U(t) Analysis software of a vibrating signal The converter of function A(t) into voltage U(t) Mathematical model of chaotic signals generator Personal computerSoftware for vibrating process phase portrait drawing 202 FEDOROV, SMOLSKIY, MIZIRIN, SHTYKOV, FOMENKOV, KAPLUNOV The given characteristics illustrate dynamic properties of computer mathe- matical model of oscillatory system as the nonlinear differential equations of 3-rd order. The mathematical solution of these equations, received on a personal com- puter, can be transformed into chaotic voltage U(t) with the help of the built-in standard computer sound card and through digital-to-analog converter (DAC) can be sent on an input of the linear activator resulting in movement the modeling vibrating test bench (as shown in Figure 2). Thus, real chaotic fluctuations of the test bench are activated. The next problem was a registration of vibrations test bench fluctuations with the help of the remote microwave radar-tracking moving sensor “Pulsar” [5], which represents the small-sized send-receive radar device of Doppler type. Its block diagram is given on Figure 6. Figure 5. The phase portrait in chaotic mode Figure 6. The block diagram of “Pulsar” radar Figure 7. Time signal at Pulsar output “Pulsar” device operates in this application as follows. The transmitter antenna radi- ates in a test bench direction a narrow directed signal with 60 GHz frequency. At this time the test bench is already ex- cited by the chaotic signal re- ceived with the help of com- puter mathematical model of chaotic oscillations, and os- cillates in a chaotic mode. The radio transmitter signal is reflected from activated test bench and gets in the recep- tion antenna of a microwave measuring instrument. On an output of a reception path of device “Pulsar” the low-frequency in-phase and quadrate components are developed (Figure 6) which are digitized and move on a computer (Figures 2 and 6). The special computer ACTIVE CONTROL OF VIBRATIONS 203 software developed for “Pulsar” registers test bench fluctuations and makes the analysis of these fluctuations by a technique of the chaotic fluctuations analysis. Characteristics of the signal, reflected from test bench, and a phase portrait of test bench oscillatory process are the results of this analysis. Phase processing of the signal reflected from the test bench, permits to receive the high resolution on space and to register thus the minimal vibrations amplitude about 1 micron. For confirmation of an opportunity of chaotic process carry on frequency in a radio range it is enough to compare its characteristics (Figure 3-5) for computer mathematical modeling to characteristics and the phase portrait, received at the analysis of an output signal of a radar-tracking measuring instrument “Pulsar”. Figure 8. Auto correlation function Figure 9. A phase portrait of the output process Comparing in pairs Figures 3 and 7, 4 and 8, 5 and 9, we find out, that diagrams of correlation functions of the modeling and measured processes almost ideally coincide, that confirms identity of both processes. Some differences in the figures concerning mathematical model and received on measured signal speak about a difficulty of full synchronization on time of a modeling signal and the measured signal among themselves. So, a comparison of results shows that computer modeling of initial processes in a mode of chaotic oscillations and experimental results on carry of these processes through a sound card and the activator on modeling test bench, which fluctuations are analyzed by precision radar-tracking approach, gives a similar picture on a radar output. 5. Conclusion Thus, as a result of the carried out researches we receive the following important results. 1. The new radar tracking millimeter device “Pulsar” is developed as a new non-contact device for registration of mechanical vibrations as small moving of any type of the diversified alive and lifeless objects. 204 FEDOROV, SMOLSKIY, MIZIRIN, SHTYKOV, FOMENKOV, KAPLUNOV 2. New opportunities of modeling of complex chaotic processes with the help of a computer and carry of this process to area of real radio signals with the aid of the radar-tracking signal transformation are offered. 3. The developed method of mathematical synthesis of signals with chaotic phase change can be used for generating and the analysis of the signals adequately reflecting dynamics of physiological processes of alive organisms at various func- tional conditions of the patients, giving the researcher an opportunity to analyze on a computer a course of patients illness and its treatments results. References Chernobaev, V. G. (2002) Chaotic oscillations generators on the base of phase synchronization systems, (in Russian)., Ph.D. dissertation. MPEI. Moscow. Fedorov, V. A. and Krokhin, L. A. (1993) The method of arterial pulse and breathing frequency registration and the device for Doppler radar location., Russian patent 2000080. Fedorov, V. A., Driamin, M. Yu., Smolskiy, S. M., Shtykov, V. V., Kaplunov, S. M. (2000) New technology of measurement and analysis in medical diagnostics with application of nonlinear dynamic models, International Conference on Advances in Structural Dynamics, 1151-1158. Glass, L., and Mackey, M. (1991) From Clocks to Chaos. The Rhythms of Life (in Russian), MIR Publisher, Moscow. Moon, F. C. (1987) Theoretical and Applied Mechanics., Cornell University. Ithaca, New York. Tomashevskiy, A. I. (2003) Chaotic oscillations formation in amplifying paths with phase control, (in Russian)., Ph.D. dissertation. MPEI. Moscow. 205 205–212. FORCED OSCILLATIONS OF TUBE BUNDLES IN LIQUID CROSS-FLOW ∗ T. N. Fesenko and V. N. Foursov Mechanical Engineering Research Institute of Russian Academy of Sciences, Russia Abstract. The stiffness of tube system for industrial heat-exchangers is usually increased by appli- cation of intermediate supports with corresponding tube-to-support gaps. The cases are frequent when intensive hydrodynamically excited vibrations and tube wear can arise at the location of tube-to-support contacts. In this case under the conditions of tube-to-support vibrocontacts the mechanism of wear during sliding impact takes place. Such a wear is more intensive one than a pure fretting wear at constant contact parameters. The present report considers an algorithm to solve a problem of tube bundle forced oscillations in liquid cross flow with taking into account the intermediate supports and main design parameters. The model of oblique impact with normal and tangential components of the reaction forces is used which takes into consideration tube-to-support impacts. The impulsive reactions of supports are included in the right sides of nonlinear forced oscillation equations. The problem of forced oscillations is solved by use of the known force factors. The present model takes into account the vortex and hydroelastic mechanisms of oscillation excitation in cross-flow. Key words: forced oscillation, tube bundle 1. Introduction Increase in heat-exchanger durability is connected with development of investiga- tion methods for structure vibrations of heat-exchangers subjected to the action of a coolant flow. In designing heat-exchangers it is of importance to predict levers of tube displacements, stresses as well as the quantities affecting tube vibrowear at the location of tube-to-intermediate support contacts. The purpose of the present article is to propose algorithm permitting to determine tube displacements, vibrov- elocities, contact loadings, sliding path, time of contact for heat-exchanger tubes, with taking into account their real intermediate supports with clearances, i.e. non- linear elastic couplings for example for a case when tube-to-baffle clearances exist. ∗ Not presented Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 206 FESENKO, FOURSOV 2. Basis of mathematical model Tube bundles are subjected to simultaneous action of several types of excitation so they must apparently be included into mathematical models of tube bundle vibra- tion . The model used in the present paper is applicable for predicting response of tube bundle cell under the action of vortex and hydroelastic excitation mechanism. The model in full measure takes into account the results on this problem obtained by R.D. Blevins, Y.N. Chen, S.S. Chen, F. Axisa et al. Hydrodynamic forced (g, h) which depend on tube motion are determined by the expression of a form (Chen, 1978): gi = k∑ j=1 ( αi j ∂2u j ∂t + σi j ∂2w j ∂t2 + α′i j ∂u j ∂t + σ ′ i j ∂w j ∂t + α ′′ i ju j + σ′′i jw j ) hi = k∑ j=1 ( τi j ∂2u j ∂t2 + βi j ∂2w j ∂t2 + τ′i j ∂u j ∂t + β ′ i j ∂w j ∂t + τ ′′ i ju j + β′′i jw j ) , (1) where u j, w j are tube displacements in y- and x- directions, αi j, σi j, τi j, βi j are ma- trices of attached masses, α′i j, σ′i j, τ′i j, β′i j, α ′′ i j, σ ′′ i j, τ ′′ i j, β ′′ i j are matrices of hydrodynamic stiffness. All these coefficients depend on reduced flow velocity, mutual arrangement of tubes, cylinder oscillation amplitudes and characteristics of an approach flow. In practice it is often admissible to use linearized equations of motion. In this case the coefficients of hydrodynamic forces are thought to be independent on oscillation amplitudes. About calculation of hydrodynamic force coefficients we shall say in detail below. To describe tube bundle vibration we use equations for small bending oscilla- tions of rods. The following expressions for the lift force and drag force acting on i-th tube are taken: f Li = 0, 5ρQv 2DCysin ( ωst + φLi ) f Di = 0, 5ρQv 2DCx [ 1 + � sin ( 2ωst + φDi )] , (2) φLi , φ D i − describe phase dependence for tubes. The values of Sh, Cx, Cy were taken in accordance with the data in (Makhutov et al., 1985, Belocerkovskiy et al., 1983) where Sh and Cy are shown to depend on a relative pitch of the bundle; ωs= 2πSh v/D is frequency of vortex shedding. In terms of the abovementioned general equations of motion for tube bundles subjected to the action of cross flow will be EiIi ∂4ui ∂z4 + 2εimi ∂ui ∂t + mi ∂2ui ∂t2 + k∑ j=1 ( αi j ∂2u j ∂t2 + σi j ∂2w j ∂t2 + α′i j ∂u j ∂t + + σ′i j ∂w j ∂t − α ′′ i ju j − σ′′i jw j ) = 0, 5ρQv2DCysin ( ωst + φLi ) + N∑ l=1 Ryil (t) δ (z − zl) (3) FORCED OSCILLATIONS OF TUBE BUNDLES 207 EiIi ∂4wi ∂z4 + 2εimi ∂wi ∂t + mi ∂2wi ∂t2 + k∑ j=1 ( τi j ∂2u j ∂t2 + βi j ∂2w j ∂t2 + τ′i j ∂u j ∂t + β′i j ∂w j ∂t −τ ′′ i ju j − β′′i jw j ) = 0, 5ρQv2DCx [ 1+?sin ( 2ωst+φDi )] + N∑ l=1 Rxil (t) δ (z − zl) , (4) where Rxil and R y il are projections onto x- and y-axes of reactions of l-th interme- diate support for i-th tube. 3. Calculated analysis of tube bundle vibrations Instead of finite-element discretization of the Equations (3), (4) which is usually used by many investigators (Makhutov et al., 1989) it is expedient to project them onto a basis of natural oscillation modes of the related linear system. The main advantage of such an approach is saving of calculation time since it is sufficient to construct solutions only at the locations of supports with clearances and at some intermediate points. As far as in most real heat-exchanger structures all the tubes are of the same length and have identical type of boundary conditions in x- and y-directions the modal functions for the tubes vibrating in x- and y-directions will be identical. Thus let: ui (z, t) = ∞∑ m=1 αimΦm (z) wi (z, t) = ∞∑ m=1 bimΦm (z) , (5) where Φm(z) is m-th mode of natural oscillations of the related system in vacuum, i.e. L∫ 0 Φm (z)Φn (z) dz = δmn. If we project the equations (3), (4) onto this basis we obtain the following matrix equations: [C1]{Ä} + [C2]{Ȧ} + [C3]{A} + [C4]{B̈} + [C5]{Ḃ} + [C6]{B} = {P1} [D1]{B̈} + [D2]{Ḃ} + [D3]{B} + [D4]{Ä} + [D5]{Ȧ} + [D6]{A} = {P2}. These equations may be written as a single equation: [M]{S̈ } + [C]{Ṡ } + [K]{S } = {F}, (6) where: [M] = [ C1 C4 D4 D1 ] , [C] = [ C2 C5 D5 D2 ] , [K] = [ C3 C6 D6 D3 ] , {F} = { P1 P2 } , {S} = { A B } 208 FESENKO, FOURSOV M, C, K are matrices of masses, damping and stiffness correspondingly and S, F are vectors of displacement and external loading. Matrices M, C, and K have dimensions 2k×2k; k is a number of tubes in a bundle. To solve the problem stated we used the Wilsons method of step-by-step integration (Bathe and Wilson, 1972). 4. Determination of hydrodynamic coupling matrixes For determination of matrix elements M, C, K the following assumptions are introduced: − length of tubes is considerably higher than radius and tube-to-tube spaces; − plane flow; − ideal and incompressible liquid; − vortex-free flowing over bodies. Hydrodynamic force is determined by distribution of pressure about the surface of moving profile; the pressure is connected with flow potential Cauchy-Lagrange integral. In terms of these assumptions we obtained the following expressions for coef- ficients of added masses, hydrodynamic damping and hydrodynamic forces at the first approximation (Nikolayev and Smirnov, 1989) αi j = −2πρQD2ε2i jcos2γi j, σi j = −2πρQD2ε2i j sin2γi j, τi j = −2πρQD2ε2i j sin2γi j, βi j = −2πρQD2ε2i jcos2γi j, i � j, αii = βii = πρQD2; τii = σii = 0; (7) α′i j = 4πρQDv2ε3i jcos3γi j, σ′i j = 4πρQDv2ε3i j sin3γi j, τ′i j = 4πρQDv2ε3i j sin3γi j, β′i j = 4πρQDv 2ε3i jcos3γi j, i � j, β′ii = σ ′ ii = τ ′ ii = β ′ ii = 0; (8) β′′i j = α ′′ i j = −24πρQv2ε4i j sin4γi j, σ′′i j − τ′′i j = −24πρQv2ε4i jcos4γi j, α′′ii = β ′′ ii = 24πρQv 2∑ j�i ε4i j sin4γi j, σ′′ii = −τ′′ii = 24πρQv2 ∑ j�i ε4i jcos4γi j. (9) FORCED OSCILLATIONS OF TUBE BUNDLES 209 To investigate small oscillations of elastic tube bundles we took the values for parameters and in the expressions for the coefficients in stationary configuration. A momentary configuration is taken in terms of. Figure 1 When it is necessary one can use the second approximation for the predeter- mined coefficients. 5. Forces of contact interaction between tube and support In real heat exchangers tubes are being spaced by use of spacer grids with clear- ances. Taking into consideration impact of tubes with ring constraints we took the model of oblique impact with normal and tangential components of support reaction force. Description of contact interaction in a normal direction includes that energy of dissipation at an impact is not taken into account, and expression for normal forces will be of a form: RilN (t) = −C[ril (t) − δil]η[ril (t) − δil] , where ril(t) is radial displacement at l-th support; δil is clearance at l-th support; η is Heavyside’s function. To calculate tangential forces of oblique impact we used dry friction hypoth- esis, according to which the normal force by coefficient of friction and is directed countermotion Rilτ = fT R il N . The total reaction for i-th tube at l-th intermediate support is determined by geo- metric summation of forces �Ril = �R il N + �Rilτ . An expression for projection of support reactions onto axis x and y in the equations (3) and (4) will be of a form: Rxil (t) = −C[ril (t) − δil] (cosϕ + fT sinϕ) η (ril − δil) , Ryil (t) = −C[ril (t) − δil] (sinϕ + fT cosϕ) η (ril − δil) , 210 FESENKO, FOURSOV where φ is phase angle, C is equivalent stiffness for bodies of a contact pair. Satisfactory results are obtained if the following stiffness assessment is used: C = 1, 9 ( Eh2 / D ) √ h/D , where h is thickness of a tube wall. Contact duration for every impact depends on value of C. As a rule these durations are very short, less than 10−3 s. However demand of high accuracy for that assessment is not often necessary, due to the fact that influence of the processes is averaged in considerably higher time scales being determined by bending oscillations of tubes. 6. Results As it is known from the experiments (Makhutov et al., 1985), only interactions between the nearest tubes of a bundle are of essential importance, which is a feature of short-range action of hydrodynamic couplings. For heat-exchanges with great number of tubes used in practice one can distinguish a system fragment consisting of 4 – 7 tubes being nearest to i-th tube. We shall call this system including i-th tube “a cell” of bundle and shall take into account the interaction of i-th tube only with the tubes of this cell. In cross sections of real bundles tubes displacement symmetry often takes place i.e. for all the inner tubes of the bundle geometry of the corresponding typical cells is identical. This symmetry can give analogous symmetry in hydro- dynamic couplings. In this case all the internal tubes of the bundle are under the identical condition of flow over bodies, so hydrodynamic couplings in the entire typical cell will be identical. For large bundles with a regular layout of cross-section it is sufficient to perform investigations of tube vibrations on model bundle with geometrically similar typical cell and consisting of a smaller number of tubes. In conclusion we present the computational results for vibration of 3 different cells of a bundle (Fig. 2). Each tube is hinged at the ends, Intermediate supports have coordinates z1 = 0,515 m; z2 = 1,215 m; they have clearances δ = 0,0005 m. At midspan of tube- systems there is a cross flow of steam. FORCED OSCILLATIONS OF TUBE BUNDLES 211 Figure 2 Figure 3 The main parameters are the following: length of tubes l = 2, 13 m Young’s modulus E = 1,1•105MPa Internal diameter d = 0,0176 m External diameter D = 0,019 m. The tubes are displaced relatively to the centre of openings at intermediate supports; the eccentricity down the flow is taken to be e = 0,005 m. The spacing T between the tubes is taken to be equal to 0,028 m for all the cells shown in Fig. 2. Fig. 4-6 presents the obtained dependences of maximum displacements along and across the flow for three various types of cells. The first maxima of displacements (Fig. 4-6) correspond to the first natural frequencies of a bundle. The right boundary value of approach flow velocity on graphics corresponds to initiation of hydroelastic instability induced by tube-to-tube interaction. Figure 4 Figure 5 212 FESENKO, FOURSOV Figure 6 In this case flow velocity is equal to 9,2 m/s for a cell of three tubes (Fig. 2a), to 9,4 m/s for a cell of 6 tubes (Fig. 2b) and to 8,0 m/s for a cell of 7 tubes (Fig. 2c). As it can be seen the difference in values of critical velocity is not great as well as the difference in oscillation amplitude level which in fact confirms a possibility to study tube bundle vibrations on the cell consisting of a small number of tubes. In our case it was a cell of three tubes. References Bathe, K. I., Wilson, E. L. (1972) Large eigen value problems in dynamics analysis, J. Eng. Mech. Div. 98, 1471-1485. Belocerkovskiy, S. M., Kotovskiy, V. N., Nisht, M. I., Fedorov, V. M. (1983) Mathematical modeling of unsteady separated flow of circular cylinder (in Russian), News of AS of USSR. Mechanics of Liquid and Gas 4, 138-148. Chen, S. S. (1978) Cross flow Induced Vibrations of Heat Exchanger Tube Bundles, Nuclear Engineering Design 47, 67-86. Makhutov, N. A., Kaplunov, S. M., Prouss, L.V. (1985) Vibration and longevity of the marine power equipment (in Russian), Sudostroenie, Leningrad, USSR. Makhutov, N. A., Fesenko, T. N., Kaplunov, S.M. (1989) Dynamic of Systems in a Liquid Flow and Structure Durability, Intern. Conf. EAHE, Prague, 148-156. Nikolayev, N. Ya., Smirnov, L.V. (1989) Determination of boundary of hydroelastic excitation of oscillation for the single-row tube bundle streamlined by cross flow (in Russian), Inter-High- School Collection, Gorky, USSR, 38-45. 213 213–218. THE EXACT AND APPROXIMATE MODELS FOR THE VIBRATING PLATE PARTIALLY SUBMERGED INTO A LIQUID George V. Filippenko and Daniil P. Kouzov Institute of Mechanical Engineering of Russian Academy of Sciences, Russia Abstract. The problem of free oscillations of the plate partially submerged into the layer of liquid is considered in the rigorous mathematical statement. The exact analytical solution of the problem is constructed. The eigen frequencies and the eigen functions of vibrating plate basing on analyses of exact solution are calculated. The influence of liquid’s level on eigen frequencies and on eigen functions is analysed. Key words: plates, membranes, acoustic field, wave propagation, boundary-contact problems 1. Introduction The problem of oscillations of elastic constructions partially submerged into the water is one of the actual problems of modern technics. Ships, oil platforms, sea airports are the examples of such bodies. However the exact calculation of such bodies vibrations is rather complicated. So it is useful to explore the pos- sible oscillations in these objects taking as an example more simple mechanical systems. The aim of this work is to build the exact solution of free oscillations problem of the most simple mechanical model of this class - the plate partially submerged into liquid. Then we compare the exact solution for plate with the solution of the approximate model. The related problems of acoustic waves passing through elastic body have been described in articles (Romilly, 1964)–(Belinsky, 1984). In these articles, the plate overlapped the waveguide channel without passing over its bounds. Free oscillations of thin elastic body protruding above the surface are de- scribed below. The model is shown on Fig. 1. The plate is orthogonally rigidly fixed to the bottom of water reservoir. It crosses the layer of liquid and pro- trudes over its upper free surface. We restrict ourselves to two-dimensional model, considering the processes independent on the coordinate z. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 214 FILIPPENKO, KOUZOV The similar problem solved for the vibrating membrane one can find in (Filip- penko and Kouzov, 2001). Our considerations are based on the precise analytical representations for acoustic field in the fluid and for the field of bending displacements of the plate. In order to get this representation the rigorous mathematical statement of this problem is given. It is considered to be a boundary-contact value problem in a medium, i.e. the problem that besides conditions on the bounds of the medium (on the bottom, on the free surface, on the plate) - has contact conditions at the points of the plate fixing with the bottom and at the point of submerging of the plate into the water. z x y H l Figure 1. Physical model We satisfy these conditions at the final stage of the solution’s construction. In this case that leads to the finite system of linear algebraic equations for the determination of several constants. The size of this system is defined by the num- ber of boundary-contact conditions. This number does not depend on the num- ber of normal modes of the waveguide we take into consideration for numerical calculations. 2. The exact statement of the plate oscillations problem Let’s start considering a layer of an ideal compressible fluid, where the acoustic pressure P(x, y) is described by the Helmholtz equation (∆ + k2)P(x, y) = 0, (−∞ THE VIBRATING PLATE PARTIALLY SUBMERGED INTO A LIQUID 215 and bottom of the reservoir is rigid ∂P(x, 0) ∂y = 0, (−∞ 216 FILIPPENKO, KOUZOV It’s not difficult to consider any other boundary conditions. We supplement boundary conditions by the condition at the infinity. We take into account the waves in the liquid radiating from the plate and the waves damp- ing as the distance from it increases. The problem is in the determination of resonances of the plate and its shapes in the domain below the cutoff frequency of the waveguide. The exact expression for displacements of the plate can be derived only after defining a field in the medium. So we come to the boundary-contact problem (Kouzov, 1997)–(Kouzov et al., 2003) for the Helmholtz equation, for the acoustic pressure in the fluid. We search the solution as an expansion in plane waves (normal modes of the waveguide). At first, according to the traditional scheme of solving of boundary - contact value problems, we construct the general solution, satisfying all requirements, except for the boundary - contact conditions. Such solution should contain four unknown constants (according to the number of boundary - contact conditions). We search acoustic field to the left and to the right of the plate in the form: P(x, y) = sign(x) ∞∑ n=0 An fn(y)e iλn |x|, (10) where fn(y) = cos(µny), µn = π H (n + 1 2 ), λn = √ k2 − µ2n, n = 0, 1, 2, ... The general solution ζ0(y) of the relevant homogeneous equation (5) has the form ζ0(y) = A cos κy + B ch κy + C sin κy + D sinh κy (11) Similarly the displacements ζ(y) of protruding part of plate H THE VIBRATING PLATE PARTIALLY SUBMERGED INTO A LIQUID 217 Hence, we obtain the dispersion equation for eigen frequencies of the plate. After defining constants and frequencies, we have complete solution of the problem in terms of displacements of the plate ζ(y) and pressure P(x, y) in the medium P(x, y) = sign(x) ∞∑ n=0 (ζ0 , fn) bn fn(y)e iλn |x|, with the exactitude to within an arbitrary constant. Here bn = 2 Dc(µ4n − κ4) + iλn ρwω2 , (ζ0 , fn) = 2 H ∫ H 0 ζ0(y) fn(y)dy. 4. Approximate model The exact solution and numerical calculations based on it allows us to carry out the comparison of various approximate methods for obtaining numerical and analytical results and estimate their precision and applicability. Let’s consider the case, when frequency ω is less than the cutoff frequency ωcut of the waveguide. In this case, separating from boundary effects, it is possible to accept, that the main influence on oscillations of the underwater part of plate are rendered by associated mass of water. For approximate calculation of the field in the plate and in the fluid we replace our initial model by the “dry” plate, consisting of two pieces. For the upper plate (l>y>H) we have a homogeneous equation (3), where κ is the wave number in the “dry” plate (4). The similar equation we have for the lower plate (H>y>0) with replacement of κ by wave number λ, which we take from dispersion equation of a problem for infinite plate in the fluid λ4 − κ4 ω2 − 2ρw Dc √ λ2 − k2 = 0. The equations are supplemented by contact conditions (7) of the plate rigid fixing in the point y=0, and conditions (8) in the point of linking of the upper and lower parts of plate. Here takes place the continuity of displacements, moments and cutting forces. We take the boundary conditions on the upper extremity of plate y=l as (9). The free oscillations of the system are considered. 5. Numerical calculations The evaluations were carried out according to the obtained formulas. The con- structed analytical solution of the problem is correct for the whole interval of 218 FILIPPENKO, KOUZOV frequencies where the physical model is correct. But we are interested in the frequencies below the cutoff frequency of the waveguide, when the system liquid – plate has real eigen frequencies. The influence of relative heights of fluid’s level H on the eigen frequencies and the form of the modes of initial model and of the approximate model are shown by the calculations. References Belinsky B. P. (1984) Free oscillations of elastic body in waveguide, Acoustic journal (Russian) 30 14-17. Filippenko G. V., Kouzov D. P. (1998) Sound transmission through a plate dividing the waveguide and protruding above the surface of a liquid, XXV- conference of Non-Linear mechanics, Institute of Mechanical Engineering of Russian Academy of Sciences 281-286, St.Petersburg. Filippenko G. V., Kouzov D. P. (1998) The income impedance of a plate submerged in a waveguide, XXVI- conference of Non-Linear mechanics, Institute of Mechanical Engineering of Russian Academy of Sciences 286-295, St.Petersburg. Filippenko G. V., Kouzov D. P. (2001) On the vibration of a membrane partially protruding above the surface of a liquid, Journal of Computational Acoustics (JCA) 9 1599-1609. Kouzov D. P., Pachin V. A. (1976) Diffraction of acoustic waves in plane semiinfinite waveguide with elastic walls, J. Appl. Math. R. 40 89-96. Kouzov D. P. (1997) Boundary-contact problems of acoustics and methods of factorisation, Methoden und Verfahren der Mathematischen Physik 42 102-113. Kouzov D. P., Veshev V. A., Lavrov Y. A. (2003) Analytic solutions in boundary-contact problems of acoustics, Recent Research Development Acousyics, ISBN: 81-7895-083-9 1 145-170. Lukianov V. D., Nikitin G. L. (1990) Scattering of acoustical waves on an elastic plate parting two various fluids in a wave guide, Acoustic journal (Russian) 36 68-75. Romilly N. (1964) Transmission of Sound through a Stretched Ideal membrane, JASA 36 1104- 1109. Romilly N. (1969) Sound transmission through a thin plate under tension, Acustica 22 183-186. 219 219–224. A COMPUTATIONAL TOOL BASED ON GENETIC ALGORITHM FOR DETERMINING OPTIMUM SHAPES OF VIBRATING PLANAR AND SPACE TRUSSES M. Tolga Göğüş, Nildem Tayşi and Mustafa Özakça Department of Civil Engineering, University of Gaziantep, 27310 Gaziantep, Turkey Abstract. This paper deals with the structural optimization of vibrating planar and space trusses. The objective of the present study is to develop a robust and reliable shape optimization tool for vibrating trusses. Natural frequencies are determined using standard finite element matrix- displacement method. The fundamental concept is to generate structural shapes for trusses in which certain vibration characteristics are improved. The results of comparative studies of the Genetic Algorithm (GA) against other continuous optimization algorithms for truss problems are reported to show the efficiency of the structural optimization. Key words: truss, free vibration, optimization, genetic algorithm 1. Introduction Structural optimization of linear elastic structures has advanced steadily by deep- ening of its theoretical foundations from different disciplines and widening of its field of applicability. The significant progress in this field in the last two decades is a result of parallel progress in structural analysis, optimization techniques, geometric (or design) modeling and computer hardware. Structural optimization combines mathematics, mechanics with engineering and has become a multidisci- plinary field with applications in areas such as mechanical and civil engineering. In general, optimization techniques used in structural engineering design can be categorized into three distinct approaches: (1) gradient-based techniques, (2) optimality criteria methods and (3) evolutionary algorithms. Several textbooks discuss these methods (Arora, 1989; Rao, 1996). Review papers (Arora and Hsieh, 1994; Vanderplaats and Thanedar, 1991) also illustrate methods applied for mixed discrete-integer, continuous variable nonlinear optimization for structural design applications. The previous research work has focused on the optimization of struc- ture in static situations. However, very few applications appear to have been pre- sented for shape optimization of vibrating structure based on GA (Langley, 2003; Al-Khamis, 1996). Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 220 GÖGÜŞ, TAYŞİ, ÖZAKÇA The present paper focuses on the development of computational tool for find- ing optimum structural shapes of vibrating arches thereby assisting structural engineers and potentially providing considerable insight into dynamic behavior. 2. Genetic algorithm Unlike many mathematical programming algorithms, GAs do not require the eval- uation of gradients of the objective function and constraints. However, GA re- quires more function evaluations than gradient-based methods. GAs are compu- tationally simple, but powerful in their search for improvement, and they are not limited by restrictive assumptions about the search space, such as continuity or existence of derivatives. GAs are search procedures based on the mechanics of natural genetics and natural selection. They combine the concept of the artificial survival of the fittest with genetic operators abstracted from nature to form a robust search mechanism. The GA approach towards structural shape optimization may be given as follows (Tayşi, 2005), Step 1. Define the optimization problem including the type of objective and constraint functions and GA parameters (population size nz, bit string length m, elite ratio Pe, probability of crossover Pc, probability of mutation Pm, crossover operator required and seed number.) Step 2. Generate population of n bit strings (candidate solution to a problem) with number of individuals equals to nz. (i.e. create the population with the number of individuals equals to nz. Step 3. Repeat the following steps for each individual (i = 1, 2, . . . , nz) • Define the geometry. • Carry out finite element analysis and calculate the objective function f (s)i and constraint functions (gi). Step 4. Check of the feasibility of each individual using the predefined con- straints. The constraint violation violi is obtained from the evaluated constraints gi violi = ci − 1.0 f or ci ≥ 1.0 (1) violi = 0.0 f or ci ≤ 1.0 (2) where ci = gi/gT (3) Step 5. Compute the value of the modified objective function f̄ (si). f̄ (si) = f (s)i × (1.0 + PM × ci) (4) ˘ A COMPUTATIONAL TOOL BASED ON GENETIC ALGORITHM 221 Step 6. Search the smallest f̄ (s)min and largest f̄ (s)max value of the modified objective function out of whole population nz. Step 7. Evaluate the fitness function ( f̂i) for all individuals: f̂i = f̄ (s)max + f̄ (s)min − f̄ (s)i (5) Step 8. Sort the whole population ( f̂i) according to the value of fitness function f̂i of each individual where the largest value of is the best. Step 9. Calculate the average fitness value ( fav) using: fav = ∑ndv dv=1 f (s)i nz (6) Step 10. Check of the termination condition. | fav − fbest| fav × 100 < convg (7) where, fbest is the best fitness value in each generation, and convg is the conver- gence value specified by the user. If the conversion is satisfactory, then stop the program. If not, continue. Step 11. Kill all individuals whose fitness value below the average fitness fav. Step 12. In the surviving part ns, finding of the new value of the largest modi- fied objective function f̄ (s)newmin which is slightly above fav. Step 13. Define a new fitness function f̂ newi for only the surviving part of the population. The fitness function is f̂ newi = f̄ (s)max + f̄ (s) new min − f̄ (s)i (8) Step 14. Calculate the probability of selection f iti of all the surviving individ- uals using f iti = f̂i fav (9) Step 15. Create the new population. Step 16. Perform the mutation. The number of binary digits nmb that are changed can be computed from nmb = nzPmndvm (10) Step 17. Repeat the process from Steps 3–13 until a converged solution is obtained, or a prescribed maximum number of iterations have been performed. 222 Figure 1. Nodal and element numbering for nine bar truss 3. Examples 3.1. NINE BAR 2D TRUSS Problem definition: In the first truss example, we consider the sizing optimization of nine bar truss. The geometry of the truss is shown in Figure 1. Material proper- ties for the truss are: Young’s modulus E = 2.0 × 1011 N/m2 and material density ρ = 7860.0 kg/m3. Node 1 acts as a pin-jointed static support, while node 6 is free to move in the x-direction. In this optimization problem, the objective is to maximize the fundamental frequency with a constraint that the total weight of the structure should remain constant. The initial weight is 55215.89 kg. Nine design variables are considered by the GA where design variables are cross-sectional area of each member. The GA pseudo-continuous design variables considered are in the range 0.0005 to 0.5 m2, with initial cross-sections of each bar is equal to 0.2 m2, population size 400, design variable binary string length 10 and number of iterations 100. Discussion of the results: The convergence to optimal design for nine bar truss is achieved after 61 iterations. The fundamental frequency increases from 30.79 Hz to 36.79 Hz. The increase in fundamental frequency from sizing optimization was 16%. The resulting truss design of the GA for pseudo-continuous design variables, is compared with reference in Table 1. The present optimum solution is very close to reference solution. 3.2. THREE BAR 3D TRUSS Problem definition: This example, involves the optimization of the 3D three bar truss shown in Figure 2 was presented by other researchers (Tayşi, 2005). The following material properties for the truss are: Young’s modulus E = 6.7 × 1010 N/m2 and material density ρ = 2700.0 kg/m3. Nodes 1,2,3 are the locations of ball and socket static supports. The objective is to maximize the fundamental fre- quency subject to the constraint that the weight of the truss remains constant. Two GÖGÜŞ, TAYŞİ, ÖZAKÇA˘ A COMPUTATIONAL TOOL BASED ON GENETIC ALGORITHM 223 TABLE I. Comparison of optimum frequencies of nine bar 2D truss DV Element Size opt converged solutions (m2) number Present GA (Al-Khamis, 1996) DOT s1 1 0.3170 0.2720 s2 2 0.3080 0.3250 s3 3 0.1810 0.1430 s4 4 0.0440 0.0340 s5 5 0.0970 0.1010 s6 6 0.1590 0.1520 s7 7 0.0260 0.0310 s8 8 0.2660 0.2710 s9 9 0.2660 0.3060 Optimum freq. (Hz) 36.69 37.38 Figure 2. Nodal and element numbering for nine bar truss design variables are considered by the GA where design variable s1 = A(1) = A(3) and s2 = A(2). The constrained weight is 9.9803 kg. The GA pseudo-continuous design variables considered are in the range 0.0001 to 0.005 m2, with initial cross- sections of each bar is equal to 0.001 m2, population size 200, design variables binary string length 10, number of iterations 100. Discussion of the results: The convergence to optimal design for three bar truss was achieved after 56 iterations. For an initial weight of 9.98 kg, the frequency before optimization was 177.08 Hz. Table II illustrates the results obtained using 224 TABLE II. Comparison of optimum frequencies of three bar 3D truss DV Element Size opt converged solutions (m2) number Present GA (Al-Khamis, 1996) DOT s1 1,2 0.000541 0.000526 s2 3 0.000129 0.000148 Optimum freq. (Hz) 205.022 203.056 the GA for continuous design variables. The increase in fundamental frequency from sizing optimization was 13.6 %. 4. Conclusion In the present work, computational tools have been developed for the geometric modeling, automatic mesh generation, analysis and GA optimization of planar and space trusses. GAs are best suited for structural optimization, since they han- dle the discrete variables efficiently. Note that almost all the design variables in most of the structural optimization problems are discrete in nature. Many optimal design problems solved using GAs give better results compared to the results of traditional methods. GAs are acceptable for engineering design. References Al-Khamis M. (1996) Structural Optimization for Static and Free Vibration Conditions Using Genetic and Gradient-Based Algorithms, Ph.D. thesis, University of Wales, Swansea. Arora J. S. (1989) Introduction to Optimum Design, McGraw-Hill. Arora J.S., Hsieh, C. (1994) Methods for Optimization of Nonlinear Problems with Discrete Variables: A Review, Structural Optimization 8, 69–85. Langley S. (2003) Genetic Algorithm Focused Comparative Optimization Study for a Broad Scope of Engineering Applications, Ph.D. thesis, University of Wales, Swansea. Rao S. (1996) Engineering Optimization, Theory and Practice, John Wiley & Sons. Tayşi N. (2005) Analysis and Optimum Design of Structures Under Static and Dynamic Loads, Ph.D. thesis, University of Gaziantep, Gaziantep. Vanderplaats G. N., Thanedar P. (1991) A Survey of Discrete Variable Optimization for Structural Design, In Proceeding of the 10th ASCE Congress Conference on Electronic Computation, Indianapolis, Indiana. GÖGÜŞ, TAYŞİ, ÖZAKÇA˘ 225 225–230. ON THE DYNAMICAL STRESS FIELD IN THE PRE-STRETCHED BILAYERED STRIP RESTING ON THE RIGID FOUNDATION Coşkun Güler1 and Surkay D. Akbarov2 1Dept. of Math. Eng., Faculty of Chemistry and Metallurgy, YTÜ, Davutpasa Campus, 34010, İstanbul, Turkey 2Inst. of Math. and Mech. of National Academy of Sciences of Azerbaijan, Baku, Azerbaijan Abstract. The forced vibration of the bi-layered pre-stretching slab resting on the rigid foundation is investigated within the framework of the piecewise homogeneous bodies model by the use of the equations Three-dimensional Linearized Theory of Elastic Waves in Initially stressed Bodies (TLTEWISB). Key words: pre-stretched bilayered strip, rigid foundation 1. Introduction The investigations regarding the study of the influence of the initial stresses on the dynamical stress field in the homogeneous and layered materials has a great significance in the theoretical and practical sense. Up to now in this field there (Akbarov and Ozaydın, 2001a) and (Akbarov and Ozaydın, 2001a), (Güler and Akbarov, 2004), (Emiro lu et al., 2004) and some others. The domains consid- ered in these investigations are semi-infinite. Therefore the results obtained in the aforementioned papers can not be applied, for example, in the case where the lay- ered material which rests on the rigid foundation and in the like many others cases. Because of the above discussions in the present paper the investigations carried out in the foregoing papers are developed for the system which comprises bilayered infinite strip and rigid foundation. The studies are made within the framework of the piecewise homogeneous body model with the use of the Three-dimensional Linearized Theory of Elastic Waves in Initially Stressed Bodies (TLTEWISB). Throughout the study, repeated indices indicate summation over their ranges. However, underlined repeated indices are not to be summed. Vibration Problems ICOVP 2005, c© 2007 Springer. ¸ ¸ üare a few investigations presented in the papers by (Akbarov and G ler, 2005), ¨¨ ğ (eds.), . şE. Inan and A. Kırı 226 GÜLER, AKBAROV 2. Formulation of the problem We consider the bilayered strip resting on the rigid foundation. Assume that the thickness of the upper and lower layers of the strip are h1 and h2 respectively. The Lagrangian coordinates 0x1x2x3 which in the natural state coincide with the Cartesian coordinates, are associated with the upper face of the strip. In this case the upper and lower layers in the strip occupy the regions {−∞ < x1 < ∞,−h1 ≤ x2 ≤ 0,−∞ < x3 < ∞} and {−∞ < x1 < +∞,−h2 ≤ x2 ≤ −h1,−∞ < x3 < ∞}, respectively. Assume that the free surface (plane) of the upper layer of the strip the time-harmonic lineal-located normal force acts. We aim that the layers before the compounding with each other and with a rigid foundation be stretched separately along the 0x1 axis. As a result of the acting of the foregoing lineal located force the plane-strain state occurs in the 0x1x2 plane. Below the following notation is used: the values related to the upper and lower layers are denoted by superscripts (1) and (2) respectively: the values related to the initial stresses are denoted by supercripts (m), 0 where m = 1, 2. Consider the case where the initial stresses in the layers of the strip are deter- mined as σ(m),011 = const � 0, m = 1, 2, σ (m),0 i j = 0 f or i j � 11. (1) The linear elastic material of the layers are assumed to be both homogeneous and isotropic, and σ(m),011 is the unique non-zero component of the initial stress tensor of the layers; therefore, σ(m),011 is denoted in short form as σ (m) 0 . According to Guz (2004), for the considered case within of each layer the following equations of the TLTEWISB are satisfied ∂σ (m) i j ∂x j + σ(m)0 ∂2u (m) i ∂t2 = ρ (m) 0 ∂2u (m) i ∂t2 , i; j = 1, 2; m = 1, 2 (2) In equation (2) the density of the m the material in the natural state is denoted by ρ(m)0 . The other notation in equation (2) is conventional. For an isotropic compressible material the mechanical relations can be written as σ (m) i j = λ (m)θ(m)δi j + 2µ (m)ε (m) i j , θ (m) = ε(m)11 + ε (m) 22 , (3) where λ(m) and µ(m) are the Lame’ constants and ε(m)i j = 1 2 ( ∂u(m)i ∂x j + ∂u(m)j ∂xi ). (4) It is assumed that the complete contact conditions u(1)i |x2=−h1 = u (2) i |x2=−h1 , σ (1) i2 |x2=−h1 = σ (2) i2 |x2=−h1 , u (2) i |x2=−h1−h2 = 0 (5) PRE-STRETCHED BILAYERED STRIP 227 are satisfied. Moreover, on the free-face plane of the upper layer of the strip the boundary conditions σ(1)12 |x2=0 = 0, σ (1) 22 |x2=0 = −Pδ(x1)e iwt (6) hold. Within the above-stated the formulation of the considered problem is ex- hausted. 3. Method of solution We attempt to use the well-known dynamical potentials of the classical linear elastodynamics for the solution of the considered problem, i.e. we use the repre- sentations: u(m)1 = ∂ϕ(m) ∂x1 + ∂ψ(m) ∂x2 , u(m)2 = ∂ϕ(m) ∂x2 − ∂ψ (m) ∂x1 , (7) By direct verification it is proved that the potentials must satisfy the equations ∇2ϕ(m) + σ (m) 0 λ(m) + 2µ(m) ∂2ϕ(m) ∂x21 = 1 (c (m) 1 ) 2 ∂2ϕ(m) ∂t2 , ∇2ψ(m) + σ (m) 0 µ(m) ∂2ψ(m) ∂x21 = 1 (c (m) 2 ) 2 ∂2ψ(m) ∂t2 , (8) where ∇2 = ∂ 2 ∂x21 + ∂2 ∂x22 , c (m) 1 = √√ λ(m) + 2µ(m) ρ (m) 0 , c (m) 2 = √√ µ(m) ρ (m) 0 (9) are the speeds of dilatation and distortion waves respectively. It follows from equation (8) that, under the condition σ(m)0 = 0, the equations (8) coincide with the corresponding equations derived in the classical linear theory of elastodynamics. In the hypothesis of time-harmonic motion with circular frequency w, every field g(x1, x2, t) can be expressed as g(x1, x2)eiwt and equation (8) becomes ∇2ϕ(m) + σ(m)0 λ(m) + 2µ(m) ∂2ϕ(m) ∂x21 + (c(1)2 ) 2 (c(m)1 ) 2 Ω2ϕ(m) = 0, ∇2ψ(m) + σ(m)0 µ(m) ∂2ψ(m) ∂x21 + (c(1)2 ) 2 (c(m)2 ) 2 Ω2ψ(m) = 0. (10) 228 GÜLER, AKBAROV Where Ω = wh c(1)2 . For the solution of the equations (10) the exponential Fourier transformation fF(sx2) = ∫ +∞ −∞ f (x1, x2)e isx1dx1 (11) is employed. After applying the transformation ()to equations (10) the functions ϕ (m) F and ψ (m) F are determined as follows: ϕ̄ (m) F = A (m) 1 (s)e γ (m) 1 x2 + A (m) 2 (s)e −γ(m)1 x2 , ψ̄ (m) F = B (m) 1 (s)e γ (m) 2 x2 +B(1)2 (s)e −γ(m)2 x2 , (12) where (γ(m)1 ) 2 = s2(1 + σ(m)0 λ(m) + 2µ(m) ) − (c(1)2 ) 2 (c(m)1 ) 2 Ω2, (γ(m)2 ) 2 = s2(1 + σ(m)0 µ(m) ) − (c(1)2 ) 2 (c(m)2 ) 2 Ω2. (13) The unknowns A(m)1 , A (m) 2 , B (m) 1 and B (m) 2 which enter the expressions (12) are determined from the contact conditions (5) and boundary conditions (6). By usual mathematical procedure, expressions are also determined for u(m)nF , ε (m) n jF and σ (m) n jF (n;j;m=1,2) from which the original of the sought values are represented as {u(m)n , σ (m) n j , ε (m) n j }(x1, x2) = 1 2π ∫ +∞ −∞ {u(m)nF , ε (m) n jF , σ (m) n jF}(sx2)e −isx1ds (14) The algorithm for calculation of the integrals (14) is given in the paper by (Akbarov and Guler, 2005). Therefore, we do not consider here this algorithm. 4. Numerical results and discussions Assume that the values of the Poisson’s ratio for the layers material are equal, i.e. v(1) = v(2) = 0.25. We introduce parameters e = E(1)/E(2), ηm = σ (m) 0 /µ (m), where E(1) and E(2) the Young’s moduli for the layers materials. Moreover, introduce the parameter h = h2/h1 and the notation σ22(x1) = σ (1) 22 (x1,−h1) = σ (2) 22 (x1,−h1). Consider the graphs given in Figs. 1 and 2, which show the dependencies between σ22(0)h/P0 h = 1). Note that under constructions of these graphs it is assumed the initial stretching in the layers of the strip is absent, i.e. η1 = η2 = 0. The jumpings in these graphs correspond to the resonance values of the frequency Ω. It follows from these graphs that the resonance values of Ω decrease with e and h. These results agree well with the well-known mechanical consideration. ¨ and Ω for various h(Fig. 1, under e=2) and for various e (Fig. 2, under PRE-STRETCHED BILAYERED STRIP 229 Figures 3 and 4 show the influence of the parameter η1 (i.e. the influence of the pre-stretching of the upper layer of the strip) to the dependencies between σ22(0)h/P0 and Ω for the case where e = 2, h = 0.1(Fig. 3) and h = 1.0 (Fig. 4). According to these graphs, it can be concluded that the pre-stretching of the upper layer causes to decrease of the absolute values of σ22(0). But the influence of the pre-stretching to the resonance values of Ω is insignificant and for the considered cases can be neglected. Figure 1. Figure 2. Figure 3. Figure 4. 5. Conclusion In the present paper within the framework of the piecewise homogeneous body model with the use of the equations of TLTEWISB the forced vibration of the 230 GÜLER, AKBAROV bilayered slab resting on the rigid foundation is investigated. It is assumed that this forced vibration is caused by the lineal-located time-harmonic, normal force act- ing on the upper plane of the strip. The numerical results are presented according to these results it is established the resonance values of the frequency Ω decrease with increasing e = E(1)/E(2) (where E(1)(E(2)) is the Young’s moduli of the upper (lower) layer material) and with h = h2/h1(where h2(h1) is the thickness of the lower (upper) layer). In this case the pre-stretching of the upper layer causes to decrease of absolute values of the interface normal stress σ22 acting on the plane separated the layers. References Akbarov S. D., Ozaydin O. (2001a) The effect of initial stresses on harmonic stress field within a stratified half-plane, Europ. J. Mechnanics A/Solids 20, 385-396. Akbarov S. D., Ozaydin O. (2001b) On the Lamb’s problem for a prestressed stratified half-plane, Int. Appl. Mech. 37(10), 138-142. Akbarov S. D., Guler C. (2005) Dynamical (harmonic) interface stress field in the half-plane covered by the prestretched layer under a strip load, J. Strain Analysis 40(3), 225-235. Emiroglu I., Tasci T., Akbarov S. D. (2004) Lamb’s problem for a half-space covered with a two- axially pre-stretched layer, Mechanics of Composite Materials 40(3), 227-236. Guler C., Akbarov S. D. (2004) Dynamic (harmonic) interfacial stress field in a half-plane covered with a prestretched soft layer, Mechanics of Composite Materials 40(5), 379-388. Guz, A. N. (2004) Elastic waves in bodies with initial (residual) stresses, Kiev: “A. C. K.” -672p. 231 231–236. DYNAMIC RESPONSE OF A RECTANGULAR PLATE-COLUMN SYSTEM ON A TENSIONLESS ELASTIC FOUNDATION Kadir Güler and Zekai Celep Faculty of Civil Engineering, İstanbul Technical University, Maslak, İstanbul, Turkey Abstract. Response of a plate-column system having five-degree-of-freedom supported by an elastic foundation and subjected to static lateral load and earthquake motion is studied. Two Winkler foundation models are assumed: a conventional and a tensionless model. The governing equations of the problem are obtained, solved numerically and the results are presented in figures to demonstrate the behaviour of the system comparatively for the two types of Winkler foundation models. Key words: plate-column system, tensionless foundation, Winkler foundation, dynamic loads 1. Introduction Earthquake response of plate column systems is usually analyzed under the as- sumption that the base plate is firmly bounded to the soil. However, when the over- turning moment of the inertia forces exceeds the available overturning resistance due to gravity load, a part of the plate loses the contact with the foundation. Stud- ies involving plates on tensionless foundation have been carried out by various authors dealing with the rectangular (Celep 1988) and circular plates (Weisman 1970, Guler and Celep 1995, Celep and 2003, Celep and Guler 2004). Solutions of static problems are accomplished by applying approximate numer- ical techniques to the non-linear governing equations of the problem and in the dynamic problem the contact region of the plate depends on time and usually the analysis is carried out numerically by updating the contact region continuously. (Housner 1963, Meek 1975 and Psycharis 1983) investigated the dynamic behav- iour of the rigid block and flexible structures including the foundation uplift. Yim and Chopra (1984, 1985) investigated a beam-column system and a multi-story system. The circular plate-column system is studied by including free and forced vibrations as well as the earthquake excitation (Celep and Guler 1991 and Celep 1992). The present study deals with a system consisting of a rigid rectangular plate and an elastic column having a tip mass. The plate-column system on the elastic tensionless Winkler foundation is investigated under a horizontal tip load and under the assumption that the system is subjected to ground motion. Vibration Problems ICOVP 2005, c© 2007 Springer. ¨ ¨ ¨ Gençoğlu (eds.), . şE. Inan and A. Kırı 232 GÜLER, CELEP up lif t an d co nt ac t re gion s A A A F igure 1 A oQ Wo M p L y B B Q X x Wo xK x ,C o pM K Y y ,Cy L x U xx M c yL y U y B B cM cM P α o Y X Figure 1. Rectangular rigid plate-column system supported on a tensionless Winkler foundation 2. Statement of the problem The system shown in Fig. 1 is a rectangular rigid plate of dimension 2A and 2B, of mass Mp subjected to a uniformly distributed load Qo(t) in addition to its weight. The plate is on a tensionless Winkler foundation of stiffness K f and supports an axially inextensible and massless column of height L, lateral stiffnesses Kx and Ky and damping Cx and Cy in the directions X and Y , respectively. The column supports a tip mass Mc and the mass is subjected to a horizontal load Po(t) with an angle α. The system is subjected to a horizontal ground excitation in two directions, Ugx(t) and Ugy(t). The displaced configuration of the system can be defined by the horizontal structural displacements Ux(t) and Uy(t), the vertical displacement Wo(t) and the rotation of the rigid plate θx(t) and θy(t) around the axes Y and X, respectively, as shown in Fig. 1. Equations governing the small amplitude horizontal and vertical motions of the system can be written as Mc(Lθ .. x + U .. x + U .. gx) + CxU . x + KxUx = Pocosα Mc(Lθ .. y + U .. y + U .. gy) + CyU . y + KyUy = Posinα (Mc + Mp)(W .. o − g) = 4QoAB − P f (1) McL(Lθ .. x + U .. x + U .. gx) + Mc(W .. − g)(Lθx + Ux) + 1 3 A2Mpθ .. x = PoLcosα + M f y McL(Lθ .. y + U .. y + U .. gy) + Mc(W .. − g)(Lθy + Uy) + 1 3 B2Mpθ .. y = PoLsinα − M f x (2) PLATE-COLUMN SYSTEM 233 where W(X,Y, t) is the vertical displacement of the rigid plate, P f (t) is the vertical load, M f x(t) and M f y(t) are the moments applied to the plate from the foundation. Combination of the governing equations yields the following coupled differ- ential equations mu .. + cu . + ku = p (3) where the dots denote the differentiation with respect to the non-dimensional time τ and uT = [ux uy θx θy wo], m = [mi j(τ)], c = [ci j], k = [ki j(τ)], p = [pi(τ)]. The elements of the matrices can be found in the recent paper of the authors (Guler and Celep 2006). The non-dimensional parameters are defined as ux = Ux L uy = Uy L ugx = Ugx L ugy = Ugy L po = Po KxL kx = Kx K f AB ky = Ky K f AB cx = Cx 2 √ McKx cy = Cy 2 √ McKy qo = Qo K f L mc = Mcg K f LAB mp = Mp Mc Tc = 2π √ Mc Kx τ = t Tc a = A L b = B L (4) Due to the interaction between the free vibration modes, they all related to each other and this interaction depends on the system parameter and it becomes more pronounced, when uplift takes place. Behaviour of the system is represented by the governing equation of the problem (3), which represents the small amplitude motion of the system including the P − ∆ effect of the tip mass. 3. Numerical results and discussion Response of the system to a specified static load and ground motion can be evalu- ated by applying a numerical solution procedure to the governing equation of the problem (3). The equation is non-linear mainly by the dependence of the stiffness coefficients on whether the plate is in full contact with the foundation or it is partially uplifted. In the later case, the coefficients depend continuously on the vertical displacements of the plate. Assuming that the system is under static loads, Fig. 2 shows the variation of the lateral load po a function of the rotation of the plate θx as for various values of the direction angle of the load α. As it is seen, linear variations are obtained in case of the full contact. However, asymptotic variations are displayed, when a partial contact starts to develop. As the load increases beyond at incipient uplift, the plate separates over increasing lift-off region from the foundation. The linear variation becomes stiffer and the asymptotic values of the load increase, as the angle α becomes larger, which indicates that an increase of these parameters delays the onset of the uplift. Consider that the system is subjected to a horizontal load po 234 GÜLER, CELEP Figure 2(a) 0.000 0.002 0.004 0.006 0.008 0.010 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 θx po 0o 15o 30o 45o 60o75 o Figure 2. Variation of the lateral load po with the rotation of the plate θx for various values of the application angle α, cx = cy = 0.1, mc = 0.001, mp = 2, a = b = 0.3, qo = 0.0002 Figure 3(a) 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 0.000 0.002 0.004 0.006 0.008 0.010 po To 75o 60o 45o 30o 15o 0o Figure 3. Variation of the non-dimensional period Towith the lateral loadpo for various values of the application angle α, cx = cy = 0.1, mc = 0.001, mp = 2, a = b = 0.3, qo = 0.0002 and it will start to oscillate after removing it. Fig. 3 shows variation of the non- dimensional period To as a function of the lateral load po for various values of α. The period lengthens as the load, consequently the initial displacement increases and an uplift of the plate takes place. As Fig. 3 reveals, the period displays an asymptotic variation. The above numerical results are presented in a non-dimensional form which applies to all systems having the same non-dimensional parameters, irrespective PLATE-COLUMN SYSTEM 235 Figure 6(b) 0.00 0.04 0.08 0.12 0.16 0.0 0.5 1.0 1.5 2.0 kx=ky vx m ax 30o 15o 0o 45o Figure 4. Spectra of the displacement vx,max tip mass under the El Centro ground motion as a function of the lateral stiffness kx = ky for various values of the angle of excitation direction α for the tensionless foundation model, cx = cy = 0.1, mc = 0.001, mp = 2, a = b = 0.3, qo = 0.0002 of their actual dimensional values. However, because actual earthquake motion records are given dimensionally, the earthquake response of the system can be evaluated within the framework of the present formulation only by assuming a numerical value at least for one parameter having acceleration dimension. In the present case the numerical value T 2c /L = 0.0045s 2/m is assumed. Fig. 4 illustrates the response of the system to the north-south component of the El Centro ground motion for the tensionless foundation. The spectra of the displacement vx,max = [vx(τ)]max = [ux(τ) + θx(τ)]max of the tip mass for various values of the arrival angle of the ground motion α is presented for the tensionless foundation model. The spectral displacement displays a smooth variation with respect to the arrival angle α, i.e., the spectral values decreases as the arrival angle increases. When the lateral stiffness of the column kx and ky is large with respect to the foundation stiffness, then the rocking motion becomes pronounced compared to the column vibration. On the other hand, the column vibration appears to be significant in comparison of the rocking motion, when kx and ky decrease. Further numerical results can be found in the paper of the authors (G ler and Celep, 2006). 4. Conclusions The presented numerical results may lead to the following conclusions: (a) Inclusion of the tensionless response of the foundation softens the static and dynamic behaviour of the system due to the decrease in the support flexibility. (b) The period of the system increases, when the tensionless model of the founda- tion is taken into consideration. However, the dynamic behaviour of the displace- ment and the acceleration becomes more complex. ü 236 GÜLER, CELEP (c) When the uplift develops, the displacements of the system under the ground motion increase, whereas the accelerations and the structural displacements are reduced and the critical frequency shifts by undergoing a gradual decrease. (d) The uplift of the base plate is influenced mainly by the fundamental mode (rocking oscillations) and the higher modes (column vibrations) have less effect on the process. (e) The P − ∆ effect has a negligible effect on the response of the system. References Celep Z. (1988) Rectangular plates resting on tensionless elastic, Journal of Engineering Mechanics 114 2083-2091. Celep Z. (1992) Harmonic and seismic responses of a plate-column system on a tensionless Winkler foundation, Journal of Sound and Vibration 155 47-53. Celep Z., Guler K. (1991) Dynamic response of a column with foundation uplift, Journal of Sound and Vibration 149 285-296. Celep Z., Gencoglu M. (2003) Forced vibrations of rigid circular plate on a tensionless Winkler edge support, Journal of Sound and Vibration 263 945-953. Celep Z., Guler K. (2004) Static and dynamic responses of a rigid circular plate on a tensionless Winkler foundation, Journal of Sound and Vibration 276 449-458. Guler K., Celep Z. (1995) Static and dynamic responses of a circular plate on a tensionless elastic foundation, Journal of Sound and Vibration 183 185-195. Guler K., Celep Z. (2006) Responses of a rectangular plate-column system on a tensionless Winkler foundation subjected to static and dynamic loads. , Structural Engineering and Mechanics, An Int. J. (to be appeared). Housner G.W. (1963) The behaviour of inverted pendulum structures during earthquakes, Bulletin of the Seismological Society of America 53 403-417. Meek J. W. (1975) Effects of foundation tipping on dynamic response, Journal of the Structural Division of the American Society of Civil Engineers 101 1297-1311. Psycharis I. N. (1983) Dynamic behaviour of rocking structures allowed to uplift, Earthquake Engineering and Structural Dynamics 11 501-521. Yim S. C. S., Chopra A. K. (1984) Earthquake response of structures with partial uplift on Winkler foundation, Earthquake Engineering and Structural Dynamics 12 263-281. Yim S. C. S., Chopra A. K. (1985) Simplified earthquake analysis of multistory structures with foundation uplift, Journal of Structural Engineering 111 2708-2731. Weisman Y. (1970) On foundations that react in compression only, Journal of Applied Mechanics 37 1019-1030. 237 237–242. VIBRATION SUPPRESSION OF AN ELASTIC BEAM VIA SLIDING MODE CONTROL Mehmet İtik1, Metin U. Salamcı2 and F. Demet Ülker3 1Yozgat Mechanical Engineering Department, Erciyes University, Yozgat, Turkey 2Mechanical Engineering Department, Gazi University, Ankara, Turkey 3Aeronautical Engineering Department, Middle East Technical University, Ankara, Turkey Abstract. This paper presents experimental results of sliding mode control (SMC) technique ap- plied to an elastic beam. The aim of the controller is to suppress first two vibration modes of the beam. Mathematical model of the beam is a finite dimensional model obtained from the Bernoulli- Euler beam equation. As the system states are to be available in order to design the SMC, an observer has been designed to obtain the states of the system by measuring tip deflection of the beam. By using observed states of the finite dimensional model, SMC is designed and applied to the elastic beam giving thoroughly suppressed vibration modes. Key words: sliding mode control, observer, vibration suppression, flexible structures 1. Introduction Due to use of flexible structures in engineering designs, vibration control and suppression issue is getting more attraction by researchers not only in design- ing different control algorithms but also using various actuators and sensors. The works of Bailey and Hubbard, 1985, Petersen and Pota, 2003, Choi et al., 1999; 2001, and Ulker and others, 2003 are some of the examples in this field. In this study, we shall design and apply Sliding Mode Control (SMC) to sup- press vibrations of an elastic beam. SMC is another active research field because of its ability to control certain type of uncertainties (Utkin, 1992; Spurgeon and Edwards, 1998). Another advantage of SMC is that it can also be applied to non- linear systems (Salamc , 1999; Salamc et al., 2000). Although the beam equation here is described by a partial differential equation, a finite dimensional linear time invariant (LTI) model of the flexible beam is derived in order to simplify the design of SMC. However the states of the model are not measurable and therefore an observer is designed to estimate the state variables by measuring the tip point displacement of the beam by means of a laser displacement sensor. The designed control algorithm is then applied to an experimental flexible beam system. Vibration Problems ICOVP 2005, c© 2007 Springer. ıı (eds.), . şE. Inan and A. Kırı 238 İTİK, SALAMCI, ÜLKER 2. The beam model The flexible structure studied here is shown in Figure 1. As shown, the elastic beam is clamped in one end and free on the other. Two piezo-ceramic patches are bonded to the flexible beam as actuators near the fixed end and laser displacement sensor is used to measure tip point displacement. By using the Euler-Bernoulli beam equation, the system can be described by the following partial differential equation; ∂2 ∂x2 [ EI ∂2y(x, t) ∂x2 ] + ρA ∂2y(x, t) ∂t2 = Ca ∂2Va(x, t) ∂t2 (1) where y(x, t) is the deflection along the x-axis; E, the Young’s modulus; I, the moment of inertia; A, the cross-sectional area, and ρ , the density of the uni- form beam. Va(x, t) is the applied control voltage to piezoceramic actuators. Since piezoceramic actuators are uniform along their lengths, the control voltage can be replaced by Va(t). In order to simplify the design of SMC, Euler-Bernoulli beam equation is trun- cated to finite number of series by using assumed modes approach (Petersen and Pota, 2003) and the following ordinary differential equation set can be obtained. y(x, t) = ∞∑ i=1 qi(t)φi(x) q̈i(t) + 2ζiωiq̇i(t) + ω 2 i qi(t) = Ca ρAL3 [ φ ′ (l1) − φ ′ (l2) ] Va(t) f or i = 1, · · · , n (2) where qi(t), φi(x), φ ′ (x), ωi, ζi are the ith modal co-ordinate, mode shape function, modal slope of the beam, natural frequency, and damping ratio respectively. In this study, first two vibration modes of the beam are controlled (i = 1, 2). The mode shape function for clamped-free beam is given by φi(x) = L (cosh(λix) − cos(λix) − ki (sinh(λix) − sin(λix))) (3) where the constant, ki = cosh(λiL)+cos(λiL)/sinh(λiL)+ sin(λiL). The quantities λi are the real roots of the cosh(λiL)cos(λiL) = −1. For the first two modes of the beam, the quantities are calculated as λ1 = 3.7955 and λ2 = 9.5020.The natural frequencies can be computed from ωi = √ EI ρAλ 2 i where I = bt 3 b/12, A = btb. The natural frequencies are calculated as ω1 = 6.5720 and ω2 = 41.1890 Hz. Damping ratios of the first two modes of the flexible beam are obtained as ζ1 = 0.07 and ζ2 = 0.02 (Onba l , 2000). The constant Ca in eq.(2) is given by Ca = Epd31b(tp + tb) where Ep is the Young’s modulus, tp the thickness, and d31 the electric charge constant of piezoceramic patches. b and tb are the width and ş ı VIBRATION SUPPRESSION OF AN ELASTIC BEAM 239 Figure 1. Flexible beam model thickness of the beam respectively. For the first two modes, the beam model given in Eq.(2) can be written in the state space form as follows; ˙x(t) = 0 1 0 0 −ω21 −2ζ1ω1 0 0 0 0 0 1 0 0 −ω22 −2ζ2ω2 x(t) + Ca ρAL3 0 φ ′ 1(l1) − φ ′ 1(l2) 0 φ ′ 2(l1) − φ ′ 2(l2) u(t) (4) where x(t) = [ q1 q̇1 q2 q̇2 ]T . 3. Sliding mode controller design SMC design can be divided into two parts; design of sliding surface which will satisfy stable motion and synthesis of a control law such that the trajectories of the closed loop motion are directed to the sliding surface. Both design procedures require system states. Since the states of the flexible beam system can not be measured directly, an observer is used to obtain system states by measuring the tip point displacement of the beam. 3.1. SLIDING SURFACE AND CONTROLLER DESIGN SMC design procedure for general LTI systems is well defined (Utkin, 1992) for LTI systems represented by ẋ(t) = Ax(t) + Bu(t) y(t) = Cx(t) (5) The pair (A, B,C) is to be fully controllable and observable. Sliding surface for the system given by Eq. (5) is defined as σ(x, t) = S x(t) where S is the sliding surface matrix and it represents the slope of the sliding surface. This matrix is selected in order to satisfy stable motion for the closed loop system. 240 İTİK, SALAMCI, ÜLKER By the following control action, the system is forced to the sliding surface and kept on it after reaching the surface. u(t) = ueq(t) + un(t) (6) where, ueq(t) = − (S B)−1 S Ax(t) and un(t) = −Ksign(σ). The constant K is a positive number and determines the reaching time to the sliding surface. 3.2. OBSERVER DESIGN The modal co-ordinate and modal velocity which are the states of the flexible beam system can not be measured directly. In order to obtain system states, an observer for linear system can be designed (Utkin, 1992) as follows; ˙̂x(t) = Ax̂(t) + Bu(t) + L(y − Cx̂(t)) (7) where, x̂ ∈ Rn dimensional estimated states. By defining e = x − x̂ as the error between real and estimated states, the error dynamics can be written as follows; ė(t) = (A − LC)e(t) (8) By choosing negative eigenvalues for (A − LC) matrix, stable error dynamics is obtained indicating convergence of the estimates states to the exact values. In the SMC design, we will use these estimated state variables. 4. Experimental setup The beam used in this study is shown in Figure 2. Figure 2(a) gives the exper- imental setup of the flexible beam system. In the experimental setup 507x51x2 mm Aluminum beam and 20x25x0.61 mm dimensional 8 pieces BM500 type piezoelectric patches were used. Parameters of the flexible beam system are given in Table 1. In order to implement the controller designed in section 3, Sensortech SS10 type four-channel programmable controller and data acquisition system was used. This programmable controller system is controlled by a personal computer which runs with Linux operating system. A program was written in C programming language in order to implement SMC algorithm to four-channel programmable controller system. Then the control signal was send to Sensortech SA10 high volt- age power amplifier in order to apply to PZT patches. The controller system can send signal between -10V and +10V, then this signal was amplified 15 times by high voltage power amplifier. Tip point displacement of the beam was measured by Keyence laser displacement sensor and the measured signal was fed back to controller system. VIBRATION SUPPRESSION OF AN ELASTIC BEAM 241 TABLE I. Parameters of flexible beam Beam length, L 0.507 m Beam width, b 0.051 m Beam thickness, tb 0.002 m Beam density, ρ 2480 kg/m3 Beam Young’s modulus, E 70x109 N/m2 PZT position, l1 0.026 m PZT position, l2 0.076 m Charge constant, d31 -200x10-12 m/V PZT Young’s modulus, Ep 60x109 N/m2 PZT width, w 0.051 m PZT thickness, tp 6.1x10-4 m (a) Experimental setup (b) Schematic view Figure 2. Flexible beam system The open loop time response of the beam for 0.04 m initial tip point displace- ment and zero velocity is given in Figure 3(a). The SMC was applied to the beam and controlled time response of the beam is given in Figure 3(b) for the same initial conditions. Applied control input to PZT patches is given in Figure 4. 5. Conclusion In this work, SMC has been designed and applied to a flexible beam to suppress first two vibration modes of the flexible beam. From the experimental results it is seen that the designed controller exhibits satisfactory performance and suppresses the vibration of the beam. 242 İTİK, SALAMCI, ÜLKER (a) Free response (b) Closed loop response Figure 3. Free and closed loop responses of the beam Figure 4. Applied control voltage to PZT patches References Bailey, T. and Hubbard, J. E. (1985) Distributed Piezoelectric-Polymer Active Vibration Control of a Cantilever Beam, Journal of Guidance, Control and Dynamics 8, 605–611. Choi, S. B., Cho, S. S., and Park, Y. P. (1999) Vibration and Position Tracking Control of Piezoceramic-Based Smart Structures via QFT, Journal of Dynamic Systems, Measurement and Control 121, 27–33. Petersen, I. R. and Pota, H. R. (2003) Minimax LQG Optimal Control of a Flexible System, Control Engineering Practice 11, 1273–1287. Salamci, M. U. (1999) Two new switching surface design for nonlinear systems with their applications to missile control, PhD, Middle East Technical University, Turkey. Salamci, M. U., Ozgoren, M. K., and Banks, S. P. (2000) Sliding Mode Control with Optimal Sliding Surfaces for a Missile Autopilot Design, AIAA J. of Guid., Cont. and Dyn. 23, 719–727. Shin, H. C. and Choi, S. B. (2001) Control of Two Link Flexible Manipulator featuring Piezoelectric Actuators and Sensors, Mechatronics 11, 707–729. Spurgeon, S. K. and Edwards, C. (1998) Sliding Mode Control: Theory and Applications, London, Taylor - Francis. Ulker, F. D., Nalbantoglu, V., Caliskan, T., Yaman, Y., and Prasad, E. (2003) Active Vibration Control of Smart Structures, In UMTS 2003 (in Turkish), Ankara, Turkey. Utkin, V. I. (1992) Sliding Modes in Control and Optimization, Berlin, Springer Verlag. 243 243–248. THERMOELASTIC STRESS ANALYSIS FOR LINEAR ELASTIC BODIES J. Ivanova1, N. Bontcheva1, F. Pastrone2 and M. Bonadies2 1 Institute of Mechanics, Bulgarian Academy of Sciences, Sofia, Bulgaria, 2 Department of Mathematics, University of Torino, Italy Abstract. The stress analysis for elastic cylinders made of polymethylmethacrylate (PMMA) under cyclic loading is provided by using FEM. Under the assumption that the elastic moduli depend linearly on the temperature, new fomulae, generalizing the well known Kelvin’s formula, are given such that we can estimate the temperature variation on the external surface of a homogeneous ideal cylinder. Using suitable numerical methods we obtain theoretical results which fit the experimental values for the temperature variation Key words: thermoelastic effects, temperature variation, cyclic loading 1. Introduction The use of high-tech materials for structural applications has shown recently that the study of mechanical behaviour of such materials is often inadequate, ignoring the presence of the temperature changing due to the mechanical loading and to the heterogeneities of material structure. Generalizations of the Kelvin’s formula have been proposed recently, therefore the studies on the interaction between me- chanical and thermal effects in solid bodies have received considerable attention (Berra and Bocca, 1993) ,(Bonadies et al., 1998), (Quaranta, 1992), (Stanley and Chan, 1985),(Stanley and Chan, 1988), (Wong et al., 1987). The aim of this study is to compare the experimental data in (Berra and Bocca, 1993) for variation of temperature with the respective temperature values, ob- tained through new formulae. The numerical results (FEM) for the stress invari- ants of the cylinder made of PMMA under cyclic loading are used to predict the temperature variations. 2. State of art Most results in the literature of thermoelastic stress analysis are obtained within the linear theory of elasticity. For a body under a cyclic loading, the linear relation between the rate of change in temperature and the rate of change of the principal stress sum under adiabatic conditions is given by the well known Kelvin’s formula Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 244 IVANOVA ∆θ = −k∆ (σ1 + σ2 + σ3) , where θ, σi, i = 1, 2, 3 and k denote temperature, principal stresses and thermal constant of the body respectively. In (Bonadies et al., 1998) from the first law of thermodynamic within the lin- ear theory of elasticity for an isotropic continuous body, the differential equation giving the relation between stress and temperature is deduced: θ̇ θ = − ( α ρc − Γ1σ ) σ̇ + Γ2 3∑ 1 σiσ̇i , (1) In equation (1) ρ is the mass density, θ is the absolute temperature, c is the specific heat at constant deformation, σ = trS , σ̇ = trṠ , σi = (S )i, σ̇i = ( Ṡ ) i , S being the first Piola-Kirchhoff stress tensor and Γ1 = 1 ρc ( − ν E2 ∂θE + 1 E ∂θν ) Γ2 = 1 ρc ( 1 + ν E2 ∂θE − 1 E ∂θν ) (2) with E and ν demoting respectively the Young’s modulus and the Poisson’s ratio. Equation (1) is derived by assuming that − the mass density ρ, the thermal expansion α, and the specific heat c are constant with respect to temperature during the thermodynamic process, − the Young’s modulus and the Poisson’s ratio depend only on the variation of temperature. Let us assume that E (∆θ) = E0 + E1∆θ ; ν (∆θ) = ν0 + ν1∆θ ; ∆θ = θ − θ0 , such that the respective first partial derivatives with respect to θ are constant. By assuming such a linear dependence of the elastic moduli from the temperature variation we involve the influence of the possible micro-damage on the macro- level. Let us introduce the thermal constant kρc and let us assume that at very slow linear change of the temperature the functions Γ1, Γ2 can be approximated by Γ1 = 1 ρc E0ν1−E1ν0 E20 [ 1 − 2 E1E0 (θ − θ0) ] Γ2 = 1 ρc E1+E1ν0−E0ν1 E20 [ 1 − 2 E1E0 (θ − θ0) ] , or Γ1 = A1 [1 − 2b (θ − θ0)] and Γ2 = A2 [1 − 2b (θ − θ0)] , where A1 = 1ρc E0ν1−E1ν0 E20 , A2 = 1ρc E1+E1ν0−E0ν1 E20 , 0 < b = E1E2 ≤ 1 . Under the above assumptions we integrate equation (1) in three cases. Case 1. 1D stress case We assume that the principal stresses are σ = σ1, σ2 = σ3 = 0. If at t = 0 θ = θ0, σ (θ0) = σ0, E (θ0) = E0, ν (θ0) = ν0, the integration of (1) under the THERMOELASTIC STRESS ANALYSIS 245 assumption ln θθ0 � ∆θ θ0 gives: ∆θ θ0 = −k (σ − σ0) + A1 + A2 2 [ σ2 − σ20 − 2b ( σ2 − σ2aver ) ∆θ ] , (3) where A1 + A2 = 1ρc E1 E20 and σaver is the average value of the stress for the whole time interval of external loading. Case 2. 2D stress case We assume that the principal stresses are: σ1 � 0, σ2 � 0, σ3 = 0 and we denote by σ and r the first and second invariant of the stress tensor respectively: I1 = σ1 + σ2 = σ, I2 = σ1σ2 = r. If at t = 0 θ = θ0, σ (θ0) = σ0, r (θ0) = r0, E (θ0) = E0, ν (θ0) = ν0, the linearized solution of equation (1) is: ∆θ θ0 = −k (σ − σ0) + A1 + A2 2 [ σ2 − σ20 − 2b ( σ2 − σ2aver ) ∆θ ] −A2 [r − r0 − 2b (r − raver)∆θ] , (4) where raver is the averaged value for the second invariant of the stress tensor for the whole time interval of external loading. Case 3. 3D stress case We assume that the principal stresses are: σ1 � 0, σ2 � 0, σ3 � 0 and we denote by σ and f the first and second invariants of the stress tensor respectively: I1 = σ1 + σ2 + σ3 = σ, I2 = σ1σ2 + σ1σ3 + σ2σ3 = f . If at t = 0 θ = θ0, σ (θ0) = σ0, r (θ0) = r0, f (θ0) = f0, (θ0) = E0, ν (θ0) = ν0, the linearized solution of equation (1) is: ∆θ θ0 = −k (σ − σ0) + A1 + A2 2 [ σ2 − σ20 − 2b ( σ2 − σ2aver ) ∆θ ] −A2 [ f − f0 − 2b ( f − faver)∆θ ] , (5) where faver is the averaged value for the second invariant of the stress tensor for the whole time interval of external loading. 3. Numerical example According to (Berra and Bocca, 1993), we will deal with numerical analysis to find the thermoelastic stress-strain behavior for the cylinder made of PMMA under cyclic loading. The material and thermal parameters are given in Table 1: The cyclic compression load applied to the top of the cylinder is given by: σ = (σmax − σmin) sinωt + σmin , where σmax = aσR, a = 0.2, 0.4, 0.6, 0.8, σmin = 1.6 × 109Pa, ω = 1Hrz, σR = 134 × 109Pa in compression ( σR is a failure stress for PMMA). The time length for each loading level is 120min. 246 IVANOVA TABLE I. Material parameters k ρ c α E ν W/(mK) kg/m3 J/(kgK) 1/K Pa 0.19 1190 2160 7 × 10−6 3.1 × 109 0.2 The elastic cylinder made from PMMA is considered. The height and the diameter of the cylinder are 140mm and 40mm respectively (Berra and Bocca, 1993). FEM is used to calculate the values of the first stress invariant at the chosen points (nodes) of the cylindrical surface, where the thermal sensors are fixed (see Figure 1). We note that according to the experiment conditions each loading level ( σmax = aσR, a = 0.2, 0.4, 0.6, 0.8 ) is kept constant for 120min. Figure 1. FE mesh To facilitate calculations let now assume E1E0 ≈ 1. We obtain from (3) the following equation for the variation of the temperature σ: ∆θ = A B , (6) where A = kσ0 − 12ρcE0σ 2 0 + σ ( −k + 12ρcE0σ ) and B = 1θ0 − 1 ρcE0 σ2aver + 1 ρcE0 σ2. Substituting into formula (6) the numerical values for first invariant we can evalu- ate the temperature variations in the chosen points on the cylindrical surface with initial temperature 200C and σ0 = 1.6 × 109Pa. THERMOELASTIC STRESS ANALYSIS 247 4. Results and discussion According to formula (6) and numericaly calculated variation of the stress invari- ants, the variations of the temperature for node 124 (see Figure 1) are shown in Figure 2b. Comparison of temperature variations at node 124, predicted by means of formula (6) and Kelvin’s formula is given in Figure 2a. At σmax = 0.2σR the experimentally measured temperature variation is 0.0080C, while our predicted value for temperature variation there is 0.0040C. 5. Conclusion In the present study the numerical results for the stress invariants, applied to the analytical formula (6), predict the temperature variation for a PMMA cylinder under cyclic loading. The comparison shows a not complete agreement between the temperature variations (using formula (6)) and the experimental data from (Berra and Bocca, 1993). a. b. 0.20.40.60.8 σ/σR 20 20.004 20.008 20.012 20.01 T e m p e ra tu re , 0C Formula (8) Kelvin's formula (1) 02000400060008000 Time, s 20 20.004 20.008 20.012 20.016 20.02 Tem perature, 0C σ/σR 0.2 0.4 0.6 0.8 Figure 2. a - Comparison between Kelvin formula and (8); b - Temperature variation in time at node 124 248 IVANOVA References Berra, M. and Bocca, P. (1993) Thermoelastic stress analysis: temperature – strain relationships in concrete and mortar, Materials and Structures 26, 395–404. Bonadies, M., Pastrone, F., and Lucia, U. (1998) Thermoelastic stress analysis for linear thermoelas- tic bodies, Quaderni di Matematica pp. 1–15. Quaranta, F. (1991–1992) SPATE: un modello matematico, Master thesis, Mathematical Depart- ment, University of Torino, Italy. Stanley, P. and Chan, W. K. (1985) Quantitative stress analysis by means of the thermoelastic effect, J. Strain Analysis 20, 129–137. Stanley, P. and Chan, W. K. (1988) The application of thermoelastic stress analysis technique to composite materialst, J. Strain Analysis 23, 3, 137–143. Wong, A. K., Jones, R., and Sparrow, J. G. (1987) Thermoelastic constant or thermoelastic parameter, J. Physics and Chemistry of Solids 48, 8, 749–753. 249 249–254. DISCRETE AND CONTINUOUS MATHEMATICAL MODELS FOR TORSIONAL VIBRATION OF MICROMECHANICAL SCANNERS Cihan Kan1, Hakan Urey2 and Erol Şenocak3 1Department of Mechanical Engineering, İstanbul Technical University, İstanbul, Turkey 2 Turkish Airlines Maintenance Center, İstanbul, Turkey 3 Department of Electrical Engineering, Koç University, İstanbul, Turkey Abstract. Micromechanical scanners are used in various industrial scanning applications like display and imaging technologies. The desired vibration mode is often the torsional mode, so derivation of an accurate mathematical model for calculation of torsional mode frequency has great importance. In this work, discrete and continuous mathematical models are given for free torsional vibration of a box shaped scanner suspended with two beams. Numerical calculation of torsional rigidity using energy methods is shown. The derivations are extendible to scanners that have non- rectangular beam cross-sections, orthotropic material anisotropy, and different mirror geometries. Analytical formulas are compared with three-dimensional FEM simulations using ANSYS commer- cial software. The FEM simulations and analytical formulas are verified with experimental results. FEM simulations and experimental results showed that simple discrete models can be used for a wide range of beam dimensions except for the cases where beam inertia is comparable to mirror inertia. Key words: micromechanical scanners, torsional frequency, torsional rigidity, orthotropic material 1. Introduction Micromechanical scanners are used in various industrial applications. The flexure- mounted micromechanical scanners presented throughout this work is used espe- cially for display and imaging applications. The main applications are barcode reading, surgical operations, personal display systems, and data display devices (Urey 2003 and Dickensheets 1998). In display and imaging applications the desired motion is usually torsional mode, so formulation of a reasonable mathe- matical model that give the torsional resonance frequency of the micromechanical scanner is important (Lin 2002). Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 250 KAN, UREY, ŞENOCAK 2. Formulation of mathematical models There are two possibilities for a mathematical model of the micromechanical scan- ner (and of any other physical system under investigation in general), classified ac- cording to the manner in which physical properties of actual system are described discrete systems and continuous systems (Meirovitch 1967). In this work, both methods are used to derive natural frequency equations for torsional vibrations of micromechanical scanners. Second order effects and all type of nonlineari- ties (including material and large deflection induced) are ignored. The rotational and translational motions of flexure cross-sections are assumed to be small. The derivations are given for a linearly elastic constant orthotropic material to give a brief understanding in how anisotropy affects frequency results. In all analytical derivations flexures are assumed to be elastic and they are modeled using uniform torsion theory. Figure 1. Geometrical dimensions of micromechanical scanners The mirror is assumed to be rigid. Warping of mirror is ignored. The geomet- rical dimensions of micromechanical scanner and coordinate system are shown in Figure 1. The natural frequency formula of ODF systems is discussed widely in literature (Meirovitch 1967). d2y dt2 + ω2y = 0 (1) ω = 2π f = √ Ks/Ie f f (2) ω is circular natural frequency, Ks is torsional spring constant for ODF model and Ie f f is effective inertia. TORSIONAL VIBRATION OF MICROMECHANICAL SCANNERS 251 2.1. DISCRETE ANALYTICAL MODELS The simplest mathematical model for vibration of micromechanical scanners is a one degree-of-freedom (DOF) model. The single DOF is torsion of mirror. The flexures are modeled as two discrete torsion springs. Saint Venant theory of uniform torsion (Timoshenko 1951) is used to calculate torsional spring constant. Torsional spring constant is found as a product of torsional modulus (G) and tor- sional stiffness (K) (GK-product) and inversely proportional to the flexure length L f (e.g., single-crystal silicon). The governing differential equation is given as below, (Kan 2004) Jm,xx d2θmech dt2 + 2(GK)θmech L f = 0 (3) where GK is Saint Venant’s constant of uniform torsion. Exact formula of GK is found only for simple cross-sections. The exact solution of GK under torsion is obtained as a serial solution by (Weaver 1990). GK = ( 16 3 ab3Gxy ) 1 − 192 π5µ ( b a ) ∞∑ n=1 (odd) ,3 ( 1 n5 Tanh [nπµa 2b ]) (4) where µ = √ Gxx/Gxy is due to material anisotropy and is equal to 1 for isotropic materials. The natural frequency is found as below (Garnier 2000), f = 1 2π √√ 2GK / L f Jm,xx (5) 2.2. CONTINUOUS ANALYTICAL MODELS In continuous mathematical model the mirror is assumed to be rigid and the flexures are assumed to be encounter small torsional deformations. If a beam is subjected to torsional moment than usually the cross-sections are warped. The continuous model presented here takes into account deformation energy asso- ciated with warping but it neglects the kinetic energy associated with it. If we define θ(x, t) as the angle of the twist of a cross-section at the point x and time t, the relation between the deformation and the twisting moment T is (Meirovitch 1997); T (x, t) = GK(x) ∂θ(x, t) ∂x (6) where GK(x) is the torsional stiffness defined for an anisotropic medium. The tor- sional motion of a cross-section element can be written once the torsional rigidity 252 KAN, UREY, ŞENOCAK is calculated: [ T (x, t) + ∂T (x, t) ∂x ] − T (x, t) + mT (x, t) = ( J f (x) L f ) dx ∂θ2(x, t) ∂t2 (7) where mT (x, t) is distributed external twisting moment applied on the flexure and it vanishes in free vibration of the beams. If we use the relationship between torsional deformation and twisting moment given in Equation 6 and rewrite Equation 7: ∂ ∂x [ GK(x) ∂θ(x, t) ∂x ] − ( J f (x) L f ) dx ∂θ2(x, t) ∂t2 = 0 (8) Solution of Equation 8 is obtained using separation of variables. The following non-algebraic equation is found after some derivations and it has to be solved numerically to find frequency equation (Meirovitch 1997). Tan(βL f ) = J f ,xx( Jm,xx/2 ) 1 βL f Unknown parameter is βL f (9) It has infinity of solutions βL f which must be obtained numerically and are related natural frequencies wr by: wr = βrL f √ GK L f J f ,xx r = 1, 2, 3, . . . (10) 3. Finite element simulations The analytical solutions are compared to 3D FEM results to obtain the validity range of the formulas. The FEM simulations are carried out for Steel and Silicon materials. The material properties of Steel and Silicon (in solution coordinate sys- tem) used in the analytical formulas are given in Table 1. Single crystal < 100 > Silicon has cubic anisotropy in its principal material coordinates. The solution coordinate system used for FEM calculations for silicon structures fabricated from (100) wafers is obtained by a rotation of principal material Cartesian coordinate system of 45◦ around the < 100 > direction, or z-axis and the resulting material model is linear elastic constant orthotropic. An APDL programming language based ANSYS macro is written for easy and quick FEM simulations. Two non-dimensional parameters n = a/b and k = L/2 max(a, b) are defined. The flexure dimensions are adjusted to keep torsional natural frequency of the micromechanical scanner around 20KHz. For our study, we fixed the torsion frequency to be near 20KHz and used the following mirror dimensions in the analytical and FEM calculations using ANSYSTM: TORSIONAL VIBRATION OF MICROMECHANICAL SCANNERS 253 15 different (k, n) combinations for flexure width ratio values of k = [1, 2, 3] and flexure length ratios of n = [2, 4, 6, 8, 10], exploring a large portion of the flex- ure design space. The constant mirror dimensions and varying flexure dimensions used for calculations are given in Table 2. TABLE III. Comparison of results Scanner Model Number SL28 E201-30 CB02 Mirror Shape round square round Flexure width ratio, k = a/b 1/16 2.1 1.9 Flexure length ratio, n 6.7 19.6 6.2 Torsion Formula (Hz) 251 1168 7985 Torsion Experimental (Hz) 270 1069 8450 Torsion ANSYS (Hz) 250 N/A N/A TABLE I. Linear elastic material coefficients of silicon and steel Steel Silicon Ex 2.03 × 1011 1.7 × 1011 Ey 2.03 × 1011 1.7 × 1011 Ez 2.03 × 1011 1.30 × 1011 νxy 0.293 0.064 νxz 0.293 0.36 νyz 0.293 0.36 Gxy 7.85 × 1010 5.1 × 1010 Gxz 7.85 × 1010 7.94 × 1010 Gyz 7.85 × 1010 7.94 × 1010 TABLE II. Mirror and Flexure dimensions used for FEM comparison Dm 1.5mm Lm 1.5mm tm 0.3mm a between 114 and 473µm b between 92 and 194µm Lf between 227 and 4732µm 254 KAN, UREY, ŞENOCAK 4. Experimental results Three different comp-driven micromechanical scanners are tested to validate an- alytical derivations. The tests are performed at Ko University Optical Micro- Systems Laboratory by a Laser Doppler Vibrometer (LDV). The test results are compared with ODF discrete analytical formulation and 3D FEM simulations in Table 3. 5. Conclusions We derived discrete and continuous mathematical models for torsional free vibra- tion of micromechanical scanners. The calculation of torsional stiffness using ap- proximate energy methods is presented. The approximate and exact torsional stiff- ness of rectangular cross-section are given. Analytical derivations are compared with 3D FEM simulations. The analytical and FEM results are partly verified with test results. The FEM simulations and test results have shown that discrete ODF mathe- matical model is a good and reasonable representation of physical system and the need to continuous model for a better flexure inertia calculation is little. References Dickensheets D. L., Kino G. S. (1998) Silicon micromachined scanning confocal optical micro- scope, J. Microelectromechanical Systems 7, 38-47. Garnier A. et al. (2000) A Fast Simple and Robust 2-D Micro-Optical Scanner Based on Contactless Magnetostrictive Actuation, Proc. MEMS’2000, Miyazaki, Japan 715-720. Kan C. (2004) Vibration of Micro Scanners, MS Thesis, Istanbul Technical University Lin L. Y., Goldstein E. L. (2002) Opportunities and challenges for MEMS in lightwave communi- cations, J. Sel. Top. in Quantum Elect. 8 163-172. Meirovitch L. (1967) Analytical Methods in Vibrations, Macmillan, New York. Timoshenko S. P., Goodier J. N. (1951) Theory of Elasticity, McGraw-Hill, New York. Meirovitch, L. (1997) Principles and Techniques of Vibrations, Englewood Cliffs, New Jersey. Urey H. (2003) Retinal Scanning Displays, Encyclopedia of Optical Engineering, Ed. Driggers R., Dekker 3 2445-2457. Weaver W. Jr., Timoshenko S. P., Young D. H. (1990) Vibration Problems in Engineering, Wiley, New York. Young W. C., Budynas R. G. (2002) Roark’s Formulas for Stress and Strain, McGraw Hill, New York, 7th Edition. 255 255–263. PHYSICAL MODELING OF STATIONARY AND IMPULSIVE PROCESSES FOR LARGE-SCALED CONSTRUCTIONS OF FLUID ELASTIC SYSTEMS S. M. Kaplunov, N. A. Makhutov, V. I. Solonin and N. V. Shariy∗ IMASH RAN, MGTU by N.Bauman, OKB GIDROPRESS Russian Federation Abstract. The most applicable method to decide the tasks of practical analysis of structure dynamic response under stationary and nonstationary excitation is physical modeling. The basic points of method are in the simulation of necessary excitation by means of similar medium propelling or of shaking tables with corresponding physical models of structure according to the special relations between main parameters or by accelerogramm (at the second case). The complicated constructions of power equipment are studied in the paper. In this case practice of straight physical modeling confirms the effectiveness of large-scale models application which are produced from the same as prototype material and provided with use of the same technology. Such modeling gives the opportunity to conduct rather representative row of experimental investigations to forecast dynamic behavior and longevity of prototype object. Key words: impulsive processes, large-scaled constraction fluid, elastic systems 1. Introduction Ever growing production and operation of the same type systems of machines and structures undergoing force action of medium flows gave rise to physical modeling of complex dynamic processes and interactions (Alaboujev et al., 1968; Bolotin, 1979; Kaplounov and Makhutov, 1999; Shoulman, 1976; Frolov et al., 2002). Modeling plays a specific role in investigation of the phenomena related to flow separation including self-excited oscillations (Kaplounov and Makhutov, 1999; Frolov et al., 2004). A pure theoretical approach and mathematical models developed recently (Frolov et al., 2002; Frolov et al., 2004). permits to reduce to minimum ne- cessity in experimental data, but nobody can manage without them completely. Aero-hydroelastic damping of oscillations (ξ) and three-dimensional-time distri- butions of pressure pulsations (p’) can only be obtained experimentally as well as clamping conditions and mechanical damping. To some extent the semi-empirical ∗ Not presented Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 256 KAPLUNOV, MAKHUTOV, SOLONIN, SHARIY methods (Alaboujev et al., 1968; Kaplounov and Makhutov, 1999; Frolov et al., 2002), are also based on test results. That is why the experiment is the main way to knowledge and investigation. Thanks to the experiment validity used hypotheses and models are checked and some coefficients are set. Experimental studies performed on models have some essential advantages over the prototypes (Kaplounov and Makhutov, 1999; Frolov et al., 2002; Frolov et al., 2004). The considered direction of physical modeling in aero-hydroelastic systems is based on the main points of the similarity theory (Alaboujev et al., 1968). 2. General approach Figure 1. Shell-tube heat-exchanger (scheme, 1 - case, 2 - input-output, 3 - tube-bundle, 4 - tube-sheets) The structures considered are complex systems con- sisting of elements in the form of shells, tubes, bars, plates with partial or full perforation and with a wide range of clamping condition characteristics (Fig. 1). The known experience of investigation allows us to call force action of flow on the structure as a random sta- tionary process with features of ergodicity and with pe- riodic components of load- ing. Separate components of special interest are those of vortex and vibroacoustic origin (Kaplounov and Makhutov, 1999; Frolov et al., 2002). That is why it is necessary to take into account correlation couplings of pressure pulsations and velocities at similar points of the system. Under separated flow self-excited oscillations of elements or their groups can be observed at qualitative change of damping nature which in fact is connected with changes of hydrodynamic conditions of flow. The main principles of hydroelastic interaction modeling can be obtained on the basis of physical processes analysis at the constant density ρ2 of flow medium (subscripts 1 and 2 – structure and medium correspondingly), Poisson’s coefficient for metal of structure µ1 and temperature of coolant Tcat identical points of the prototype and model structures. The influence of critical phenomena of heat exchange is not taken into account. The liquid is supposed to be incompressible. Firstly we study steady vibrations or periodic processes when the initial conditions are out of consideration. LARGE-SCALED CONSTRUCTIONS OF FLUID ELASTIC SYSTEMS 257 At the interface of two subsystems (structure and liquid) the equilibrium con- dition must be considered as the equality of the corresponding components of stress tensors in the liquid and in deformed solid. For dynamic stresses of the corresponding direction the relation σ′1v = p′2v and σ̄1v = p2v are valid (p2- pressure in liquid). The relations of parameters pv = pp/pmtype are equal to the corresponding scale (subscripts p and m – prototype and model correspondingly). The simultaneity condition at the boundary is determined by the equality of normal interface velocities for the solid v1 and the liquid v2 independently on the boundary layer of liquid related to the added mass (Frolov et al., 2002). The corresponding condition of materials properties provides identical mean- ing of relation between dynamic characteristics of the model and the full-scale system. It is expressed in a ρ1/ρ2 = idem form. If one takes into account that the medium (construction material) density can be determined through the elas- tic modulus E and velocity of sound c - ρ = E/c2 then the expression for the condition of mediums properties correspondence in two - component system is E1c22/E2c 2 1 = idem (E2 for liquid is k2). Condition of the contact similarity in static has a form by N.Prigorovsky (1985) (E1v + E2v) l2v = ( 1 + E12p ) Fv where F is force, E12p = E1p/E2p, subscripts 1, 2 relate here to the contacting parts of the system. The condition is satisfied by corresponding choice of model material and by choice of linear and force scales lv, Fv. In descriptions of dynamic characteristics of contact in the similar systems the relation of test duration on full-scale objects and on a model corresponds to the time scale tv = lv and it is also provided by the choice of materials, mediums, quality of surface and scales lv, Fv. In a complex dynamic model the vibroimpact process is supposed to be ac- companied by oblique impacts. Here according to Newton’s second hypothesis (Frolov et al., 2002) as dynamic contact similarity we suppose proportionality of tangential and normal components of the impact impulse. If it is necessary stress in the contact may be taken into account on the basis of the known models. 3. Main physical (criteria) model Considering all the main physical features of the phenomena, including buffeting, vortex resonance, hydro elastic self-sustained oscillations and galloping we get the main equation: σ1y = F(d, ρ1, ρ2, E, E’, k2, p̄2, p′2, p′2a, Fc, ∆s, bs, ls, kvw, Ls,v̄2, v′, v′2a, R, h, ν, t, ξ, f, µ,L, εσ, K’s, ψ, q), (1) where d, L – diameter and length, k2 - liquid compression modulus, E’ – dynamic elasticity modulus, p̄ and p′ - average and pulsating magnitudes of parameter; Fc – contact force; ∆s, bs, ls – tube-support gap, width of support and length of span, 258 KAPLUNOV, MAKHUTOV, SOLONIN, SHARIY kvw - vibrowear coefficient, Ls - total length of conjugate surface slip, R and h – shell radius and thickness, ν - liquid kinematic viscosity coefficient, ξ - damping coefficient, f – frequency, εσ- relative deformation, K’s- parameter of roughness; ψ, q – angle of flow and tube-bundle density. Here x and y are directions along and perpendicular to flow. In accordance with π-theorem of similitude theory (Alaboujev et al., 1968) for a standard devia- tion of relative amplitude (σ̄1y = σ1y/d) we obtain: Π’1 = σ1y/d = ϕ(Π’2, Π’3,. . . ,Π’28 ) (2) The whole procedure of getting the relation (2) is shown in Table 1 in (Frolov et al., 2002). At the first stage in the Expression (1) completeness of the charac- teristic parameters complex is assessed on the basis of analysis of known com- plexes in (Alaboujev et al., 1968; Kaplounov and Makhutov, 1999; Frolov et al., 2002; Frolov et al., 2004) taking into account mathematical models in (Al- aboujev et al., 1968; Bolotin, 1979; Kaplounov and Makhutov, 1999; Shoulman, 1976; Frolov et al., 2002; Frolov et al., 2004; Abe et al., 1978). The formal approach applied to find a complex of fundamental criteria permits to obtain a traditional form in addition to the known simple algebraic transforma- tions of nondimensional values. Correctness of transition from one main criteria complex to another is also determined by their independence. Analysis of deter- mining parameters and criteria completeness was obtained by authors in (Frolov et al., 2002). For 20 years we have been studying a problem of determining dynamic stresses in the structures of water-water reactors (high-head model), liquid-metal-water reactors (low-head model), multispan tube heat exchangers (vibrowear model) and of different types heat exchangers vibroacoustical processes (Frolov et al., 2002; Frolov et al., 2004) (vibroacoustical model). During all these investigations we successfully used the general criteria model as a basis (Table in (Frolov et al., 2002)). In all cases the equivalent transformation for initial fundamental criteria deter- minant was applied. In such a way we received the traditional criteria form which permitted us to develop a further analysis of hydrodynamic processes similarity by use of available experimental data bank (Alaboujev et al., 1968; Kaplounov and Makhutov, 1999; Shoulman, 1976; Frolov et al., 2002; Frolov et al., 2004). Equivalence of the transition was determined by independence condition satisfac- tion. 4. Added masses analysis If we apply a model medium imitator of different density we must keep in mind the influence of the density on frequency scaling. It can be done in accordance with LARGE-SCALED CONSTRUCTIONS OF FLUID ELASTIC SYSTEMS 259 Figure 2. Eigenfrequency parameter dependence on added mass coefficient χm and relative medium density ρ2v for shell systems general formula of the corresponding oscillation modes in “elastic shell-liquid- rigid shell” system (Frolov et al., 2004), where f12 is frequency of structure-and- added mass; æ or χ is coefficient of added mass. If one uses the models of the same material as full-scale ones (ρ1v = 1) fre- quency scaling is to be made according to the relations as in Fig. 2. The figure shows that as medium density in full-scale system decreases its corresponding natural frequencies rise (curves ρ2v ¡ 1). It is necessary to point out that construction eiugenfrequencies depend on added mass effect which is as well as damping are relevant for tube-bundles with small clearances in intermediate supports and for coaxial shells in liquid. 5. Assessment of physical modeling error Figure 3. Comparison of autocorrelation func- tions of input pressure fluctuations for VVER-440 model and prototype On the whole the total er- ror of modeling depends on the accuracy of simi- larity criteria reproduction which in its turn auto- matically includes the er- rors due to the taken as- sumptions, to the measur- ing process and to the rel- ative influence of the con- stant parameters reproduc- tion. That is why to solve the engineering problems by similarity method it is 260 KAPLUNOV, MAKHUTOV, SOLONIN, SHARIY sufficient to highlight main typical processes. So modeling of the phenomena to be investigated is always approximate to some extent. According to V.A. Venikov’s results in (Frolov et al., 2002) consideration of approximated similarity gives us a criterial equation of regression on the basis of statistical analysis of criteria relations. A relative effect of deviation for separate criteria magnitudes is determined here in terms of standardized coefficients of regression for this equation and according to it one finds the necessary accuracy of the similarity criteria reproduction. Physical modeling approximation and efficiency of implementation are as- sessed in each separate case on the basis of the known experience of experimental studies as it was made by M.Wambsganss (1977) and D.Gorman (1982). For the low-head criteria model (criteria No. 2,3,5,7,15,17-25 in Table in (Frolov et al., 2002)) we find ε0 as general modeling approximation. Final value of modeling relative error is about 15-20%, that is comparable with typical error of experiment. 6. Modeling of impulsive processes Described experimental-and-numerical approach is based on straight decision of dynamic equations or with use of the pseudo dynamic testing method in case of impulsive loading when we can determine the restoring force and solve dynamic equations. The basic points of method are in the simulation by means of shaking tables with special physical models of structure according to the special accelerogramms (recalculated or really characteristic for prototype one) (Shoulman, 1976; Abe et al., 1978). As we study unsteady vibrations or processes the initial conditions are into consideration. The general reception to define similitude conditions of stressed-strained con- struction condition under the impact (Alaboujev et al., 1968; Shoulman, 1976) consists of ratios AD = æ Ast; σD = æ σst; RD = æ Rst (3) where æ – dynamics coefficient; Ast, σst , Rst – static displacement, strain and force reaction, correspondingly. In this case considered system (Alaboujev et al., 1968) consists of rigid body A with weight P without accounting of deformation, which falls from the height H on the stable fixed elastic system B. Dynamics coefficient (Alaboujev et al., 1968) is nondimensional magnitude. That is why similitude under the shaking excitation may be provided with some limitation by observance of similitude conditions in static action of loading with condition of dynamics coefficient identity æ = idem. (4) LARGE-SCALED CONSTRUCTIONS OF FLUID ELASTIC SYSTEMS 261 Next step to traditional modeling of seismic durability when the initial magnitudes of acceleration Ẍ0 = ( d2X dt2 ) 0 and displacement X 0 are known. In this case the next conditions tẌc P0X = idem, ωX V = idem, kX P0X = idem, mẌ P0X = idem, (5) kmAm P0m = kpAp P0p , mmamax m P0Xm = mpamax p P0Xp gives us the opportunity to define the connection between coordinates and time for the model and prototype for any directions (P0- force amplitude, k – rigidity, c – damping, m – mass, ω – frequency, a – acceleration, t - time). Obviously that instead of usual coordinates we can apply the characteristic ones namely oscilla- tions amplitude A, maximum acceleration amax and period of the oscillations T. The last one will be defined as: ωpTp = ωmTm . In dynamic analysis equilibrium condition of a system is expressed while considering inertia force, damping, resistance and applied force (Frolov et al., 2004; Abe et al., 1978; Chen et al., 2000) as [M] {ẅ} + [C] {ẇ} + {R} = {P} (6) where [M] and [C] are the mass and damping matrix respectively, {P} is the external load vector; {ẅ} and {ẇ} are the respect acceleration and velocity vectors. Modern experimental methods give a possibility to receive rather correct re- sults on the structure models or model - fragments with use of shaking tables and different investigation technique with FEM or FSEM application. The investiga- tion procedure (Frolov et al., 2004) is the next: digital accelerogram - accelero- gram modeling with use of similarity theory - pc - code/analogue - analogue signal tape-recording - transformation of seismic force influence accelerogram according to corresponding scale. If we can determine the restoring force {R}, we can solve equation (6). That is the main point of the pseudodynamic testing method. In our case the mass, damp- ing and stiffness matrix differs from (6) by additional hydroelasticity interaction coefficients (Frolov et al., 2002) and become asymmetric in flow. 262 KAPLUNOV, MAKHUTOV, SOLONIN, SHARIY Figure 4. Seismic stand (shaking table – 1, vibrators – 2, model – 3, pneumo-support - 4) Figure 5. Simulated realizations of seismic excitation (240 points, step of figure proce- dure for a model 1:20 is 0.00095 s; a and – accelerograms No. 1 and 2) Figure 6. Calculation overloading coefficients kgof reactor model (Frolov, 2004) in dependence of supporting scheme ( a – reactor model with main elements 1-10, b – model with two supports, c – one support model, 1 and 2 – calculation for accelerograms No. 1 and 2 correspondingly for and ; 3 – experimental data (pointed by crosses)) LARGE-SCALED CONSTRUCTIONS OF FLUID ELASTIC SYSTEMS 263 During pseudo dynamic test especially for inelastic behavior of construction (Chen et al., 2000) the restoring force {R} is directly measured by use of modern interface by National Instruments (LabView-5-7) from the model at each time step and this force is substituted into equation (6) to solve the solution for the next time increment. In such a way combines the reality of shaking table tests with calculation approach. The main advantage here is over quasi-static method in fact that the real restoring force of the model is used in the calculation instead of the hypothetical model. 7. Conclusion Wide spread physical modeling approach may considerably decrease the costly prototypes experiments to the minimum necessary to verify only some basic pa- rameters magnitudes in three or four similar points of the structure. Besides only physical modeling gives us the opportunity to investigate carefully the phenom- enon in the whole, particularly in the critical conditions that can’t be performed in prototype. References Alaboujev, P. M., Geronimus, V. B., Minkevich, L. M. and Shekhovzev, B. A. (1968) Theory of similitude and dimensions. Modeling., Moscow, Vishay Shkola. Bolotin. V. V. M. (1979) Handbook “Vibration in technique” in 6 V. (in russian), Mashinostroenie. Kaplounov, S., Makhutov, N. (1999) Physical modeling of Dynamics and Strength of AeroHy- droElastic System Structures, 3rd EAHE Conf., Czech. Rep., Prague 215-222. Shoulman, S. G. (1976) Handbook Calculation of hydraulic structure seismic strength with account of influence of water medium. M.,Energiy, (in russian). Frolov, K. V., Makhutov, N. A., Kaplounov, S. M. (2002) Dynamics of hydroaeroelastic system structures Ed. By S. Kaplunov and L. Smirnov. M. (in russian), Nauka. Frolov, K., Dragounov, J., Makhutov, N., Kaplunov, S. (2004) Dynamics and Strength of Light Water Reactors Ed. By N. Makhutov. M. (in russian), Nauka. Abe, T., Fujikawa, T., Kurohashi, M., Inoue, Y. (1978) An Approximate Analytical Method for Seismic Response of Structure Having Elastoplastic Behavior, Proc. Inst. Cont. “Vibration in Nuclear Plant”, Keswick, U. K. paper 9:4. Chen, A., Lam, E. S. S. and Wong, Y. L. (2000) Verification of a Pseudodynamic Testing System, Proc. of the Int.Conf. on Advances in Structural Dynamics (ed. By J. M. Ko and Y. L. Xu), Hong Kong, China II. 851-859. 265 265–269. THERMAL STRESSES AND NONLINEAR THERMAL DEFORMATION ANALYSIS OF SHALLOW SHELL PANEL B. Karmakar1, P. Biswas2∗, R. Kahali3 and S. Karanji4 1Research Scholar, Von Karman Society, Jalpaiguri, India 2Founder Member (ICOVP), Old Police Line, Jalpaiguri, India 3Dept. of Physics, P. D. Women’s College, Jalpaiguri, India 4Dept. of Maths, University of North Bengal, Darjeeling, India Abstract. Modern aerospace structures such as high-speed spacecrafts, missiles and engineering and nuclear structures are often subjected to severe thermal loads and reveal a clearly nonlinear response. In such situations the associated strains and stresses are usually determined from Von Karman coupled nonlinear partial differential equations extended to thermal loading in terms of transverse displacement and stress function. Although citations of several authors may be made who have employed the method, only a few is mentioned here which contains several cross-references (Chang and Jen, 1986), (Biswas, 2001) and (Mansfield, 1982). The purpose of the present paper is to further generalize the equations for the case of a simply- supported rectangular panel under thermal loading. Applications of Galerkin’s Procedure ultimately lead to a cubic equation involving several parameters. Key words: thermal stress, shallow shell, thermal loading 1. Basic governing equations Following (Donnell, 1976) and (Timoshenko and Woinowisky-Krieger, 1959) and with usual notations, basic governing equations for transverse displacement (w) and stress function (F) can be derived in the forms: D∇4w + α1E 1 − ν (∇ 2MT ) = ∂2F ∂x2 · ∂ 2w ∂y2 − 2 ∂ 2F ∂x∂y · ∂ 2w ∂x∂y + ∂2F ∂y2 · ∂ 2w ∂x2 (1) ∇4F = Eh ( ∂2w ∂x∂y )2 − ∂ 2w ∂x2 · ∂ 2w ∂yz − αtE ( ∇2NT ) − Eh R · ∂ 2w ∂x2 (2) where MT and NT are given by (Nowachi, 1962). ∗ Note: Keeping main essence, the Paper is substantially reduced to five pages. Interested readers may contact biswas
[email protected] for more details. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 266 KARMAKAR, BISWAS, KAHALI, KARANJI 2. Rectangular panel simply-supported at the edges To approach the particular problem concerning a rectangular panel simply- supported at the edges origin being located at one corner of the shell in the middle surface. Let a, b be the length and peripheral width of the shell and are taken as the x and y axes, z - axis being normally downwards. The distribution of temperature in the direction of z axis is taken as linear in the form (Nowachi, 1962). T (x, y, z) = τ0(x, y) + zτ(x, y) (3) where τ0(x, y) = T1 + T2 2 , τ(x, y) = T1 + T2 h (4) and T1 = T ( x, y, h 2 ) , T2 = T ( x, y,−h 2 ) (5) Since MT is constant one can express it in the form of the Fourier series. MT = α∑ m=1.3... α∑ m=1.3... amn sin mπx a sin mπy b (6) where amn = 16MT mnπ2 . The deflection w is assumed in the form w = w0 sin πx a sin πy b (7) which satisfies the boundary conditions for simply-supported edges: w = 0 = ∂ 2w ∂x2 + MTD(1−ν) atx = 0, a w = 0 = ∂ 2w ∂y2 + MTD(1−ν) aty = 0, b (8) Since NT is a constant and appears in the boundary conditions for inplane dis- placements, we take ∇2(NT ) = 0 in a equitation (1) from which the stress function is obtained in the form F(x, y) = A x 2 2 + B y2 2 + Ehw20 32 [ a2 b2 cos 2πxa + b2 a2 cos 2πyb ] + Ehw0 Ra2π2 ( 1 a2 + 1 b2 ) sin πxa sin πy b (9) SHALLOW SHELL PANEL 267 where, A and B are arbitrary constants to be determined from inplane boundary conditions. In accordance with the conditions occurring in airplane structures, the shell is considered rigidly framed, all edges remaining unaltered after deformation. The elongations of the shell in the directions of x and y are independent of y and x respectively. [(Timoshenko and Woinowisky-Krieger, 1959), P - 426] one gets the constants A and B as: A = Ehw20π 2 8 ( 1 − ν2) [ 1 b2 + ν a2 ] − Eα1NT 1 − ν B = Ehw20π 2 8 ( 1 − ν2) [ ν b2 + 1 a2 ] − Eα1NT 1 − ν (10) Applying Galerkin’s procedure in equation (1) a Cubic Equation is obtained for central deflections in the following non-dimensional form: C1. (w0 h )3 − C2 (w0 h )2 + C3 (w0 h ) − C4 = 0 (11) where, C1,C2,C3 and C4 are known constants. Median surface membrane stresses Nx and Ny are given by: (Nx) a 2 , b 2 Eh = C5 (w0 h )2 − C6 (w0 h ) − αt(T1 + T2) 2(1 − ν) (12) ( Ny ) a 2 , b 2 Eh = C7 (w0 h )2 − C8 (w0 h ) − αt(T1 + T2) 2(1 − ν) (13) where, C5,C6,C7 and C8 are known constants. 268 KARMAKAR, BISWAS, KAHALI, KARANJI 3. Numerical computations TABLE I. Exhibits the variations of non-dimensional central deflections (w, h) for a square panel (b/a = 1) and for b/a = 2 considering a/h = b/h = 10 and αt = 11.9 × 10/◦C (Steel) w/h (T1 + T2 = 10◦) (T1 − T2)/h (T1 + T2 = 50◦) (T1 − T2)/h (T1 + T2 = 100◦) (T1 − T2)/h 0 0 0 0 .2 .008168(R/h=10)∗ .007734(R/h=10)∗ .007295(R/h=10)∗ .009248(R/h=20)∗ .008814(R/h=20)∗ .008375(R/h=20)∗ .008732(R/h=10)∗∗ .008293(R/h=10)∗∗ .007853(R/h=10)∗∗ .009838(R/h=20)∗∗ .009398(R/h=20)∗∗ .008958(R/h=20)∗∗ .4 .025690(R/h=10)∗ .024820(R/h=10)∗ .023940(R/h=10)∗ .026120(R/h=20)∗ .025250(R/h=20)∗ .024375(R/h=20)∗ .030582(R/h=10)∗∗ .029700(R/h=10)∗∗ .028820(R/h=10)∗∗ .031024(R/h=20)∗∗ .030140(R/h=20)∗∗ .029260(R/h=20)∗∗ .6 .052654(R/h=10)∗ .051350(R/h=10)∗ .050035(R/h=10)∗ .053626(R/h=20)∗ .052320(R/h=20)∗ .051007(R/h=20)∗ .069162(R/h=10)∗∗ .067840(R/h=10)∗∗ .066520(R/h=10)∗∗ .070150(R/h=20)∗∗ .068830(R/h=20)∗∗ .067520(R/h=20)∗∗ .8 .096900(R/h=10)∗ .095160(R/h=10)∗ .093410(R/h=10)∗ .098630(R/h=20)∗ .096880(R/h=20)∗ .095140(R/h=20)∗ .036040(R/h=10)∗∗ .034280(R/h=10)∗∗ .132523(R/h=10)∗∗ .137810(R/h=20)∗∗ .136050(R/h=20)∗∗ .134290(R/h=20)∗∗ 1.0 .115801(R/h=10)∗ .113660(R/h=10)∗ .111466(R/h=10)∗ .142830(R/h=20)∗ .140660(R/h=20)∗ .138460(R/h=20)∗ .191080(R/h=10)∗∗ .188880(R/h=10)∗∗ .186680(R/h=10)∗∗ .218710(R/h=20)∗∗ .216510(R/h=20)∗∗ .214310(R/h=20)∗∗ ∗ (results for a square panel when a/b = 1) ∗∗ (results for a square panel when b/a = 2) SHALLOW SHELL PANEL 269 4. Observations In brief, it is observe from the above table for numerical results that deflections increase with increase of temperature along the thickness of the panel and such increase is higher for shell geometries other than a square panel. It is also observed that when T1 + T2 increases deflection decreases with the increase of (T1 − T2)/h for any shell geometry. References Biswas, P. (2001) Analytical and Computational Approach on Some Nonlinear Thermoelastic Plate and Shell problems, 5th ICOVP Moscow (IMASH). Chang, W. P., and Jen, S. P. (1986) Nonlinear Flexural Vibrations of Heated Orthotropic Plates, Int. J. Solid & Structures 22, 267-281. Donnell, L. H. (1976) Beams, Plates and Shells, Mc. Graw Hill Pub. Co., New York. Mansfield, E. H. (1982) On the Large-Deflection Vibrations of Heated Plates, Proc. Royal Society (London) A 379, 15-39. Nowachi, W. (1962) Thermoelectricity, Addison Wesley Pub. Co., New York. Timoshenko, S., and Krieger, S. W. (1959) Theory of Plates and Shells, Mc. Graw Hill Pub. Co. 271 271–278. NONLINEAR THERMAL VIBRATIONS OF A CIRCULAR PLATE UNDER ELEVATED TEMPERATURE B. Karmakar1, S. B. Karanji2, R. Kahali3 and P. Biswas4 1Vibration Research Group, Von Karman Society for Advanced Study and Research, Residence : Deshbondhu Para, Jalpaiguri-735101, West Bengal, India 2Dept. of Maths, University of North Bengal, Darjeeling, India 3Dept. of Physics, P. D. Women’s College, Jalpaiguri, India 4Principal (Retired), A. C. College of Commerce and Head, Vibration Research Group, Von Karman Society for Advanced Study and Research, Old Police Line, Jalpaiguri - 735 101, West Bengal, India. Abstract. Following a new approach, proposed and employed by earlier authors, nonlinear thermal vibrations of a circular plate under elevated temperature have been analysed. Key words: thermal stress, shallow shell, thermal loading 1. Introduction Problems of Mechanics of Thermal Vibrations have been analysed by many au- thors by using the classical von Karman field equations and Berger’s approxima- tion extended to the dynamic case with the inclusion of thermaI-loading (1973, 1980, 1983 and 1999). Merits (advantages) and demerits (disadvantages) of the two methods have been discussed in the literature by many authors. In this paper “a new approach” proposed by Banerjee and Dutta (1981) and employed by Sinharay and Banerjee (1985) and others (1995) will be applied to derive the basic governing equations with the inclusion of thermal loading in the dynamic case and to make an emperical study of the nonlinear thermal vibrations of a circular plate with clamped immovable edges. The basic governing equations so derived have been solved by Galerkin’s pro- cedure. Numerical computations have been presented graphically and followed by observations and discussions. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 272 KARMAKAR, KARANJAI, KAHALI, BISWAS 2. Derivation of governing field equations We consider a circular plate of radius a and subjected a temperature distribution (Nowacki, 1962) T (r, z) = φ0(r) + zφ(r), (1) The potential energy due to bending and stretching undergoing large deflections in the absence of any external mechanical load, may be expressed in the form V = D 2 a∫ 0 ( ∂2ω ∂r2 )2 + 2ν r ∂ω ∂r ∂2ω ∂r2 + ( 1 r ∂ω ∂r )2 + 12 h2 { e21 + 2(ν − 1)e2 } rdr (2) where D = flexural rigidity of the plate = Eh 3 12(1−ν2) , ω is Central deflection, ν is Poisson’s ratio, E is Young’s modulus, u is inplane displacement and e1, e2 are middle surface first and second strain invariants given respectively by e1 = ∂u ∂r + u r + 1 2 ( ∂ω ∂z )2 , e2 = u r + ∂u ∂r + 1 2 ( ∂ω ∂r )2 , (3) Putting the expressions for el and e2 into equation (2) and rearranging one gets V = D 2 a∫ 0 ( ∂2ω ∂r2 )2 + 2ν r ∂ω ∂r ∂2ω ∂r2 + ( 1 r ∂ω ∂r )2 + 12 h2 { ē21 + (1 − ν 2) u2 r2 } rdr (4) where ē1 = ∂u ∂r + 1 2 ( ∂ω ∂r )2 + ν u r (5) The kinetic energy K. E. and WT , the energy contribution due to heating effect in the plate are given by K.E. = ρh 2 � S ( ∂u ∂t )2 ( ∂ω ∂t )2 rdrdt (6) WT = − � S h/2∫ −h/2 αtET 1−ν ( ē1 − z∇2ω ) dxdydz [Basuli 1968] = − � S [ ē1NT − ∇2ωMT ] rdrdθ [Expressed in polar coordinates] (7) where (Nowacki, 1962) NT = αtE 1 − ν h/2∫ −h/2 T (r, z)dz = αtE 1 − νφ0(r)h, (8) NONLINEAR THERMAL VIBRATIONS OF A CIRCULAR PLATE 273 MT = αtE 1 − ν h/2∫ −h/2 zT (r, z)dz = αtEh3 12(1 − ν)φ(r), (9) Following the ‘new approach’ (1981), the term (1 − ν2) u2 r2 is replaced by ł4 ( ∂ω ∂r )4 or in the expression (3) where ł is a factor depending on the Possion’ s ratio of the plate material forming the Lagrangian L = K.E. − V − WT and applying Euler’s variational equations one gets the following two differential equations in the decoupled form ē1 = NT h2 12D + C f (t)rν−1h2 12D (10) ρh ∂2ω ∂t2 + D∇4ω − 6łD h2 ( ∂ω ∂r )2 { ∇2ω + 2∂ 2ω ∂r2 } − C f (t)rν−1 { ∂2ω ∂r2 + ν r ∂ω ∂r } + ∇2 MT = 0 (11) where C is a constant of integration and f (t) is some unknown function of time. 3. Method of solution For free vibrations MT = 0 and NT � 0, [Mazumdar, et al., (1981)] For a clamped circular plate of radius a, the deflection function ω satisfying the boundary conditions ω = 0 = ∂ω ∂r at r = aand u = 0 at r = 0, a (12) is assumed in the form ω = AF(t) ( 1 + 2P r2 a2 + Q r4 a4 ) , P = 1 = Q (for clambed edges) (13) ‘A’ being the amplitude of vibrations. Considering (5), (10) and (13) and integrating over the area of the plate one gets r2u = NT h2 12D rν+1 ν + 1 + C f (t)h2 12D r2ν 2ν − 8A 2F2(t) a4 [ rν+3 ( P2 ν + 3 − 2PQ a2 r2 ν + 5 + Q2 a4 r4 ν + 7 )] + C1 (14) where C1 is a constant of integration. Considering the in-plane boundary condi- tions u = 0 at r = 0, a, the constant C1 is eliminated and one finally gets c f (t) = 192A2F2(t)a−(ν+1) h2 Dν ( 1 ν + 3 − 2 ν + 5 + 1 ν + 7 ) − 2NTν ν + 1 a1−ν (15) 274 KARMAKAR, KARANJAI, KAHALI, BISWAS Applying Galerking’s Procedure in equation (11) and eliminating c f (t) with the help of (15), one gets the well -known time-differential equation in the form: . . . F(t) + αF(t) + βF3(t) = 0 (16) where α = ( D a4ρh ) 320 3 − 2560a2ν ( NT D ) (ν + 1)(ν + 3)(ν + 5)(ν + 7) (17) β = 10 ( D a4ρh ) [ 256ł 35 + 196608ν (ν + 3)2(ν + 5)2(ν + 7)2 ] [A h )2 (18) 18) In the limiting case in the absence of temperature, the result (16) is in exact agreement with that found in the literature (1985, 1995) The solution of equation (16) with the normalized conditions F(O) = 1, Ḟ(0) = 0 is given by Nash and Modeer (1959) in the form F(t) = Cn(ω�t,R), where ω�2 = α + β, R2 = β 2(α + β) (19) Cn being the Jocobian Elliptic Function. The corresponding time - period is T� = 4K/ω� where K is the complete elliptic integral of the first kind. The usual linear time - period is given by T ′ = 2π Ω0 where Ω0 is obtained from (16) by dropping the nonlinear term so that Ω0 = √ α. Relative time-period is given by T�/T ′ = 2K π ( 1 + β α )− 12 (20) 4. Numerical results and discussions Numerical results have been computed and presented graphically with the non- dimensional amplitude ( A h ) along the horizental axis and the relative time periods( T� T ′ ) along the vertical axis as shown in figures 1-14 considering the following set of values αt = 1.2 × 10−5, ν = 0.3, ł = 2ν2 [Ref. (1981)] 5. Observations and discussions � Figure-3 is the combination of Figures 1 and 2. From this figure one can observe that, if the temperature parameter remains fixed at the level of 250K, relative NONLINEAR THERMAL VIBRATIONS OF A CIRCULAR PLATE 275 time periods diminish with the increase of non-dimensional amplitudes fr both the cases of aspect ratios 10 and 20. � Figure-6 is the combination of Figures 4 and 5. From this figure one observes that, if the temperature parameter is kept fixed at the level of 500K, reverse sit- uation develops, i.e., time periods diminish with the increase of non-dimensional amplitudes fr both the cases of aspect ratios 10 and 20. (i) relative time-period decreases with the increase of non-dimensional ampli- tude when aspect ratio is 10. (ii) relative time-period increases with the increase of non-dimensional ampli- tude when aspect ratio is taken as 20. Similar situation also exists for higher temperature levels 750K and 1000K as shown in Figures 9 (combination of Figures 7 and 8) and 12 (combination of Figures 10 and 11). So there should be a transition phase at certain temperature level beyond 250K as aspect ratio changes from 10 to 20. � Figure-13 is the combination of Figures 1,4,7 and 10 for aspect ratio 10 at temperature levels 250K, 500K, 750K, 1000K whereas Figure 14 is the combina- tion of 2, 5, 8, and 11 for aspect ratio 20 at temperature levels 250K, 500K, 750K, 1000K. These figures represent what have been presented in Figures 1-12. Figure 1. Shows variations of T � T ′ vs A h for aspect ratio ah = 10, and φ0 = 250k Figure 2. Shows variations of T � T ′ vs A h for aspect ratio ah = 20, and φ0 = 250k 276 KARMAKAR, KARANJAI, KAHALI, BISWAS Figure 3. Shows comparative variations of T � T ′ vs Ah for fixed φ0 = 250k and a h = 10, 20 Figure 4. Shows variations of T � T ′ vs A h for aspect ratio ah = 10, and φ0 = 500k Figure 5. Shows variations of T � T ′ vs A h for aspect ratio ah = 20, and φ0 = 500k Figure 6. Shows comparative variations of T � T ′ vs Ah for fixed φ0 = 500k and a h = 10, 20 Figure 7. Shows variations of T � T ′ vs A h for aspect ratio ah = 10, and φ0 = 750k Figure 8. Shows variations of T � T ′ vs A h for aspect ratio ah = 20, and φ0 = 750k NONLINEAR THERMAL VIBRATIONS OF A CIRCULAR PLATE 277 Figure 9. Shows comparative variations of T � T ′ vs Ah for fixed φ0 = 750k and a h = 10, 20 Figure 10. Shows variations of T � T ′ vs A h for aspect ratio ah = 10, and φ0 = 1000k Figure 11. Shows variations of T � T ′ vs A h for aspect ratio ah = 20, and φ0 = 1000k Figure 12. Shows comparative variations of T � T ′ vs Ah for fixed φ0 = 1000k and a h = 10, 20 Figure 13. Shows comparative variations of T� T ′ vs A h for fixed a h = 10, and φ0 = 250k, 500k, 750k, 1000k Figure 14. Shows comparative variations of T� T ′ vs A h for fixed a h = 20, and φ0 = 250k, 500k, 750k, 1000k 278 KARMAKAR, KARANJAI, KAHALI, BISWAS References Banerjee, B., Dutta, S. (1981) A new Approach to an Analysis of Large Deflection of Elastic Plates, Int. J. Nonlinear Mech. 16, 47-52. Basuli, S. (1968) Large Deflection of Elastic Plates under Uniform Load and Heating, Indian J. of mech. And Maths 6, 22. Biswas, P. (1980) Large Delection of heated Orthotropic Cylindrical shallow shell, Defence Science J. 30, 87-92. Biswas, P. (1983) Nonlinear Vibrations of Heated Elastic Plates, Indian J. of Pure and Appl. Maths. 14, 119-1203. Jones, R., Mazumdar, J., Chaung, Y. K. (1980) Vibrations and Buckling of Plates at Elevated temperature, J. Solids and Structures 16, 61-70. Mondal, U. K., Biswas, P. (1999) Nonlinear Vibrations of Heated Elastic shallow Spherical under Linear and Parabolic Temparature Distributions, J. Apl. Mechs (ASME) 66, 814-815. Nowacki W. (1962) Thermoelectricity, Pergamon Press, Oxford. Pal, M. C. (1973) Static and Dynamic Nonlinear behaviour of Heated orthotropic Circular Plates, Int. J. Nonlinear Mech. 8, 480-494. Paliwal, D. N., Kanagasabapathy, H., Gupta, K. M. (1955) Vibrations of an Orthotropic Shallow Spherical Shell on a Pasternak Foundation, J. of Comp. Structure 33, 135-142. Sinharay G., Banerjee, B. (1985) A new Approach to Large Deflection Analysis of Spherical and Cylindrical Shells under Thermal Load, Mech. Res. Comm 12, 53-64. 279 279–284. FLEXURAL-TORSIONAL COUPLED VIBRATION ANALYSIS OF A THIN-WALLED CLOSED SECTION COMPOSITE TIMOSHENKO BEAM BY USING THE DIFFERENTIAL TRANSFORM METHOD Metin Orhan Kaya and Özge Özdemir Faculty of Aeronautics and Astronautics, İstanbul Technical University, Maslak, İstanbul, Turkey Abstract. In this study, a new mathematical technique called the Differential Transform Method (DTM) is introduced to analyse the free undamped vibration of an axially loaded, thin-walled closed section composite Timoshenko beam including material coupling between the bending and torsional modes of deformation, which is usually present in laminated composite beams due to ply orientation. The partial differential equations of motion are derived applying the Hamilton’s principle and solved using DTM. Natural frequencies are calculated, related graphics and the mode shapes are plotted. Key words: composite Timoshenko beam, bending torsion coupling, differential transform method 1. Introduction Figure 1. Configuration of an axially loaded com- posite Timoshenko beam A composite thin-walled beam with length L, cross sectional di- mension B and wall thickness d is shown in Fig. 1. The dimensions are assumed to be d ≺≺ B so the terms related to the warping stiff- ness and the warping inertia are small enough to be neglected. The bending motion in the Z direction, torsional rotation about the X axis and rotation of the cross section due to bending alone are represented by w (x, t), ψ (x, t) and θ (x, t), respectively. A constant axial force, P, acts through the centroid of the cross section which coincides with the X axis. P, is positive when it is compressive as in Fig. 1. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 280 KAYA, ÖZDEMİR 2. Formulation The governing undamped partial differential equations of motion are derived for the free vibration analysis of the beam model represented by Fig. 1. Applying the Hamilton’s principle, the following equations of motion are obtained − ρIθ̈ + EIθ′′ + kAG (w′ − θ) + Kψ′′ = 0 (1) − µẅ − Pw′′ + kAG (w′′ − θ′) = 0 (2) − Isψ̈ − P (Is/µ)ψ′′ + Kθ′′ + GJψ′′ = 0 (3) Here, µ = ρA is the mass per unit length; Is is the polar mass moment of inertia; K and kAG are flexure-torsion coupling rigidity and shear rigidity of the beam, respectively. The boundary conditions at x = L and x = 0 for Eqs. (1-3) are as follows ( EIθ′ + kψ′ ) δθ = 0 (4) [−Pw′ + kAG (w′ − θ)] δw = 0 (5) [− (PIs/µ)ψ′ + Kθ′ + GJψ′ ] δψ = 0 (6) A sinusoidal variation of w (x, t) ,ψ (x, t) and θ (x, t) with a circular natural frequency ω is assumed and the functions are approximated as w (x, t) = W (x) eiωt, ψ (x, t) = ψ (x) eiωt, θ (x, t) = θ (x) eiωt (7) The following nondimensional parameters can be used for simplification. ξ = x/L, W (ξ) = W/L, r2 = I/AL2, ()∗ = d()/dξ, (́) = d()/Ldξ (8) Substituting Eqs.(7) and (8) into Eqs.(1-3), the dimensionless equations of motion are obtained as follows A1θ̄ ∗∗ + A2θ̄ + A3W ∗∗ + A4ψ̄ ∗∗ = 0 (9) B1W ∗∗ + B2W + B3θ̄ ∗ = 0 (10) C1ψ̄ ∗∗ + C2ψ̄ + C3θ̄ ∗∗ = 0 (11) where the nondimensional coefficients are given by A1 = 1, A2 = µL4r2ω2 EI − kAGL 2 EI , A3 = kAGL2 EI , A4 = K EI , (12) B1 = 1 − P kAG , B2 = L2µ kAG ω2, B3 = −1, C1 = 1 − PIs GJµ , C2 = IsL2 GJ ω2, C3 = K GJ COMPOSITE TIMOSHENKO BEAM 281 3. The differential transform method The differential transform method is a transformation technique based on the Tay- lor series expansion and is used to obtain analytical solutions of the differential equations. In this method, certain transformation rules are applied to the original functions. As a result, the differential equations and the boundary conditions of the system are transformed into a set of algebraic equations and the solution of these algebraic equations gives the desired solution of the problem. A function f (x) , which is analytic in a domain D, can be represented by a power series with a center at xo , any point in D. The differential transform of the function is given by F [k] = 1 k! ( dk f (x) dxk ) x=x0 (13) where f (x) is the original function and F [k] is the transformed function. The inverse transformation is defined as f (x) = ∞∑ k=0 (x − x0)kF [k] (14) Combining Eqs. (13) and (14) and expressing f (x) by a finite series, we get f (x) = m∑ k=0 (x − x0)k k! ( dk f (x) dxk ) x=x0 (15) Here, the value of m depends on the convergence of the natural frequencies (Ho and Chen, 1998). Theorems frequently used in the transformation procedure are introduced in Table I and Table II (Ozdemir and Kaya, 2005). 4. Formulation with DTM Applying DTM to Eqs.(9-11),the following transformed equations of motion are obtained. Here, we use ψ and θ instead of ψ and θ. A1 (k + 2) (k + 1) θ [k + 2] + A2θ [k] + A3 (k + 1) W [k + 1] + A4 (k + 2) (k + 1)ψ [k + 2] = 0 (16) B1 (k + 2) (k + 1) W [k + 2] + B2W [k] + B3 (k + 1) θ [k + 1] = 0 (17) C1 (k + 2) (k + 1)ψ [k + 2] + C2ψ [k] + C3 (k + 2) (k + 1) θ [k + 2] = 0 (18) 282 KAYA, ÖZDEMİR TABLE I. Basic theorems of DTM Original Function Transformed Function f (x) = g (x) + h (x) F [k] = G [k] + H [k] f (x) = λg (x) F [k] = λG [k] f (x) = g (x) h (x) F [k] = ∑k l=1 G [k − l] H [l] f (x) = g (x) /h (x) F [k] = (k+n)!k! G [k + n] f (x) = xn F [k] = δ (k − n) = 0 if k � n 1 if k = n TABLE II. Basic theorems of DTM x = 0 x = 1 f (0) = 0 F [0] = 0 f (1) = 0 ∑∞ k=1 F [k] = 0 d f (0)/dx = 0 F [1] = 0 d f (1)/dx = 0 ∑∞ k=1 kF [k] = 0 d2 f (0)/dx2 = 0 F [2] = 0 d2 f (1)/dx2 = 0 ∑∞ k=1 k (k − 1) F [k] = 0 d3 f (0)/dx3 = 0 F [3] = 0 d3 f (1)/dx3 = 0 ∑∞ k=1 k (k − 1) (k − 2) F [k] = 0 Applying DTM to Eqs.(4-6), the boundary conditions are obtained as follows ξ = 0 ⇒ θ [0] = W [0] = ψ [0] = 0 (19) ξ = 1 ⇒ (k + 1) θ [k + 1] + A4 (k + 1)ψ [k + 1] = 0 (20) 5. Results and discussion An illustrative example, taken from (Baner jee, 1998), is solved and the results are compared with the ones in literature and the mode shapes are plotted. Variation of the first five natural frequencies of the above example with respect to the axial force is introduced in Table III. Here, the superscript γ represents the natural frequencies calculated when coupling is present (K � 0). In this table, it is noticed that the natural frequencies decrease as the axial force varies from tension to compression. Furthermore, it is seen that the coupled natural frequencies are lower than the uncoupled ones. The effects of the axial force P and the rotary inertia parameter, r , on the first four natural frequencies are introduced in Fig. 2. COMPOSITE TIMOSHENKO BEAM 283 Here, the solid and the dashed lines represent the frequencies of Timoshenko and Euler beams, respectively. As expected,the natural frequencies decrease with the increasing rotary inertia parameter, because Timoshenko effect, which is more dominant on higher modes, decreases the natural frequencies. However, since the fourth mode is torsion, r makes only a slight change in the fourth mode. Mode shapes of the beam under the effect of the compressive axial force, (P = 7.5) and coupling effect are introduced in Fig. 3. As it is seen, the first three normal modes are bending modes while the fourth normal mode is the fundamental torsion mode. Figure 2. Effects of the axial force and the rotary inertia parameter on the natural frequencies Figure 3. Normal mode shapes of the composite beam with bending-torsion coupling 284 KAYA, ÖZDEMİR TABLE III. Natural frequency variation with the axial force NaturalFrequencies P = −7.5 P = 0 P = 7.5 Present Li et al. Presentγ Baner jeeγ Presentγ Baner jeeγ Present Li et al. 40.975 40.970 37.106 37.100 30.747 30.750 28.064 28.060 224.259 224.250 197.672 197.700 189.779 189.800 210.162 210.160 598.668 598.660 525.665 525.600 518.791 518.800 586.519 586.510 647.595 647.590 648.495 648.600 648.269 648.300 647.228 647.220 1125.711 1125.710 992.878 - 986.199 - 1113.950 - References Banerjee J. R. (1998) Free Vibration of Axially Loaded Composite Timoshenko Beams Using the Dynamic Stiffness Matrix Method, Computers and Structures 69 197-208. Ho S. H., Chen C. K. (1998) Analysis of General Elastically End Restrained Non-Uniform Beams Using Differential Transform, Applied Mathematical Modeling 22 219-234. Li J., Shen R., Hua H., Jin X. (2004) Bending-Torsional Coupled Vibration of Axially Loaded Composite Timoshenko Thin-Walled Beam With Closed Cross-Section, Composite Structures 64 23-35. Ozdemir O., Kaya M. O. (2005) Flapwise Bending Vibration Analysis of a Rotating Tapered Cantilevered Bernoulli-Euler Beam by Differential Transform Method, Journal of Sound and Vibration (in press). 285 285–290. ESTIMATION OF MICROSTRETCH ELASTIC MODULI BY THE USE OF VIBRATIONAL DATA Ahmet Kırış1 and Esin İnan2 1Faculty of Science and Letters, İstanbul Technical University, 34390, İstanbul, Turkey 2Faculty of Arts and Sciences, Işık University, 34398, İstanbul, Turkey Abstract. In the present work, a nonlinear wave theory is used for the estimation of the material properties of a “microstretch” medium. For this purpose a thin plate is considered and triplicate Chebyshev polynomial series are used as admissible functions to ensure the satisfaction of geomet- ric boundary conditions of the plate. The Ritz technique is applied to derive the frequency equation of the microstretch plate and an optimization procedure is performed by minimising the least square “distance” between computed natural frequencies from the energy method and measured natural frequencies. To realize the optimization procedure, a genetic algorithm is used to estimate the elastic moduli of microstretch medium. Key words: microstretch, plate vibration, Ritz technique, genetic algorithm, inverse problem 1. Introduction The linear theory of elasticity is adequate for the vibration analysis of plates made of hard materials like steel. But it is well known that material response to external stimuli depends on the motions of its inner structures, so the linear theory of elasticity is unable to explain the behavior of many materials having complex microstructure like composite materials, polymers and porous media, etc. The in- fluence of material microstructure results in the development of new type of waves which is not found in the classical theory of elasticity for the case of vibrations characterized by high frequencies and small wave-lengths. Eringen’s microstretch elasticity which is included microstructural rotations and expansions provides a good model for such materials and reveal some high frequency branches of wave spectrum which is not appears in the classical theory of elasticity (Eringen, 1999). The wave propagation in materials modeled with microstretch theory is investigated by (Singh and Kumar, 1998), (Tomar and Garg, 2005) and (Inan, 1990). In this study, first we investigated the vibration of microstretch plates to esti- mate the overall material properties of microstretch plates. Here we used Cheby- shev polynomials as admissible functions which satisfy the geometric boundary Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 286 KIRIŞ, İNAN conditions of plate and have been used successfully by (Zhou et al., 2002) for vibration analysis of three dimensional rectangular plates in the linear theory of elasticity. The Ritz technique which is most commonly used method has some special advantages such as high accuracy. Second, an optimization procedure is performed by minimising the least square “distance” between computed natural frequencies from the energy method and measured natural frequencies obtained from experiments, and finally, we used a genetic algorithm to solve the inverse problem for identification of the over- all properties of microstretch plate. Genetic Algorithms (GAs) are evolutionary programs that manipulate a population of individuals and based on the evolution theory of Darwin to explain creation of life. He proposed that the nature of living creatures changed over the years to result in stronger specimens. The stronger specimens would then dominate. An initial population of individuals (solutions) is generated randomly and these then undergo evolution by means of reproduc- tion, crossover and mutation of individuals until a suitable solution is found. GAs have acquired great success in inverse problems and GAs are widely adopted as optimization techniques (Mota Soares et al., 1993) and (Chiroiu and Munteanu, 2002). In the present work, we had to limit Chebyshev polynomials with 5x5x1 for microstretch case while we could use 10x10x2 polynomials for classical case because of the computational difficulties and incapacity of the computer mem- ory. In spite of this, it is shown that the applied GA has great success for the determination of the overall properties of the microstretch plate. 2. Kinetic and potential energies of a microstretch plate The linearized constitutive equations for a microstretch elastic solid are given by (Eringen, 1999) as tkl = λ εmm δkl + (µ + κ) εkl + µ εlk + λ0 ψ δkl, mkl = α γmm δkl + β γkl + γ γlk, mk = a0 ψ,k (1) where tkl is the stress tensor, mkl is the couple stress tensor and mk is the mi- crostretch vector, λ and µ are Lamé constants, κ, α, β, and γ are the micropolar constants and λ0, λ1 and a0 are the microstretch constants, and ρ is the mass density. �kl, γkl and γk are the linear strain measures of microstretch elasticity with the following geometric relations εkl = ul,k+ ∈lkm φm, γkl = φl,k, γk = 3ψ,k. (2) The linear elastic strain energy density of an isotropic microstretch medium is given by W = 1 2 [ tkl εkl + mkl γlk + mk γk + (s − t)ψ ] (3) ESTIMATION OF MICROSTRETCH ELASTIC MODULI 287 where s = mkk and t = tkk and the kinetic energy per unit mass for an isotropic microstretch medium is given by K = 1 2 ρ u̇k u̇k + 1 2 ρ j φ̇k φ̇k + 3 2 ρ j ψ̇ ψ̇. (4) We consider a rectangular microstretch plate of thickness h, length a and width b. The Cartesian coordinate system is located at the middle of the plane denoted by Ω and the z axis normal to the plane. Substituting the constitutive equations (1) and the geometric relations (2) into (3-4), we can write the linear strain energy V and the kinetic energy T of the rectangular microstretch plate in integral form as V = a/2∫ −a/2 b/2∫ −b/2 h/2∫ −h/2 W dz dy dx = 12 a/2∫ −a/2 b/2∫ −b/2 h/2∫ −h/2 [ λΛ21 + (2µ + κ)Λ2 + 2µΛ3 + (µ + κ) Λ4 + αΛ 2 5 +(β + γ)Λ6 + 2βΛ7 + γΛ8 + a0 Λ9 + 2 λ0 ψΛ1 + λ1 ψ2 ] dz dy dx (5) and T = ρ2 a/2∫ −a/2 b/2∫ −b/2 h/2∫ −h/2 {[( ∂u1 ∂t )2 + ( ∂u2 ∂t )2 + ( ∂u3 ∂t )2] + j [( ∂φ1 ∂t )2 + ( ∂φ2 ∂t )2 + ( ∂φ3 ∂t )2] + 3 j ( ∂ψ ∂t )2} dz dy dx (6) where Λ1 = ε11 + ε22 + ε33, Λ2 = ε 2 11 + ε 2 22 + ε 2 33, Λ3 = ε12 ε21 + ε13 ε31 + ε23 ε32, Λ4 = ε 2 12 + ε 2 13 + ε 2 21 + ε 2 23 + ε 2 31 + ε 2 32, Λ5 = γ11 + γ22 + γ33, Λ6 = γ 2 11 + γ 2 22 + γ 2 33, Λ7 = γ12 γ21 + γ13 γ31 + γ23 γ32, Λ8 = γ 2 12 + γ 2 13 + γ 2 21 + γ 2 23 + γ 2 31 + γ 2 32, Λ9 = ψ 2 ,1 + ψ 2 ,2 + ψ 2 ,3. (7) 3. Ritz Method with Chebyshev polynomials Assuming a harmonic-time dependence, the periodic displacement, microrotation and microstretch components of the microstretch plate undergoing free vibration can be written in terms of the amplitude functions as follows {u(x, y, z, t), φ(x, y, z, t), ψ(x, y, z, t)} = {U(x, y, z, t),Φ(x, y, z, t),Ψ(x, y, z, t)} eiω t (8) 288 KIRIŞ, İNAN where ω denotes the natural frequency of the microstretch plate. For convenience and simplicity, we introduce the following non-dimensional parameters; ξ = 2x a , η = 2y b , ζ = 2z h (9) into the potential and kinetic energies (5) and (6), but we do not give their nondi- mensional forms in here for the shake of brevity. Each amplitude functions of the equation (8) can be written in the form of triplicate series of Chebyshev polynomials which ensure the satisfaction of the geometric boundary conditions as {U1(ξ, η, ζ), U2(ξ, η, ζ), U3(ξ, η, ζ), Φ1(ξ, η, ζ), Φ2(ξ, η, ζ), Φ3(ξ, η, ζ), Ψ(ξ, η, ζ)} = ∞∑ i=1 ∞∑ j=1 ∞∑ k=1 { Ai jk, Bi jk, Ci jk, Āi jk, B̄i jk, C̄i jk, ¯̄Ai jk } Pi(ξ) P j(η) Pk(ζ) (10) where Pi(χ) is the one-dimensional i th Chebyshev polynomial which can be writ- ten in terms of cosine functions as follows, i [ ] (11) Substituting the series expansions (10) into the nondimensional potential and ki- netic energy expressions and using the results in the maximum energy functional of the microstretch plate defined as Π = Vmax−Tmax and minimizing the functional Π with respect to the coefficients of the Chebyshev polynomials, i.e. ∂Π ∂Ai jk = 0, ∂Π∂Bi jk = 0, ∂Π ∂Ci jk = 0, ∂Π ∂Āi jk = 0, ∂Π ∂B̄i jk = 0, ∂Π ∂C̄i jk = 0, ∂Π ¯̄Ai jk = 0, (i, j, k = 1, . . . ,∞) (12) can be written as the following governing eigenvalue equation in matrix form [KU1U1 ] [KU1U2] [KU1U3 ] 0 [KU1Φ2 ] [KU1Φ3 ] [KU1Ψ] [KU1U2 ] T [KU2U2] [KU2U3 ] −[KU1Φ2 ] 0 [KU2Φ3 ] [KU2Ψ] [KU1U3 ] T [KU2U3 ] T [KU3U3 ] −[KU1Φ3 ] −[KU2Φ3 ] 0 [KU3Ψ] 0 −[KU1Φ2 ]T −[KU1Φ3 ]T [KΦ1Φ1 ] [KΦ1Φ2 ] [KΦ1Φ3 ] 0 [KU1Φ2 ] T 0 −[KU2 Φ3 ]T [KΦ1Φ2 ]T [KΦ2Φ2 ] [KΦ2Φ3 ] 0 [KU1Φ3 ] T [KU2Φ3 ] T 0 [KΦ1Φ3 ] T [KΦ2Φ3 ] T [KΦ3Φ3 ] 0 [KU1Ψ] T [KU2Ψ] T [KU3Ψ] T 0 0 0 [KΨΨ] P (χ) = cos (i − 1)arccos(χ) . ESTIMATION OF MICROSTRETCH ELASTIC MODULI 289 −Ω2 [M1] 0 0 0 0 0 0 0 [M1] 0 0 0 0 0 0 0 [M1] 0 0 0 0 0 0 0 [M2] 0 0 0 0 0 0 0 [M2] 0 0 0 0 0 0 0 [M2] 0 0 0 0 0 0 0 [M3] { Ai jk } { Bi jk } { Ci jk } { Āi jk } { B̄i jk } { C̄i jk } { ¯̄Ai jk } = 0 0 0 0 0 0 0 (13) where M1 = 14 D 0,0 i i G 0,0 j j H 0,0 k k , M2 = jM1, M3 = 3 jM1 and Ω = aω √ ρ. The elements of the stiffness sub-matrices and mass sub-matrices are given by [KU1U1] = (λ + 2µ + κ)D 1,1 i i G 0,0 j j H 0,0 k k + (µ + κ)α 2 1D 0,0 i i G 1,1 j j H 0,0 k k + 1 α21 D0,0i i G 0,0 j j H 1,1 k k , KU1U2 = α1(λD 1,0 i i G 0,1 j j H 0,0 k k + µD 0,1 i i G 1,0 j j H 0,0 k k ), [KU1φ2 ] = − a 2 κ 2 α1 α2 D0,0i i G 0,0 j j H 1,0 k k , [KU1U3] = α1 α2 (λD1,0i i G 0,0 j j H 0,1 k k + µD 0,1 i i G 0,0 j j H 1,0 k k ), [KU1ψ] = a 2λ0D 1,0 i i G 0,0 j j H 0,0 k k , [KU1φ3 ] = a 2 κ 2α1D 0,0 i i G 1,0 j j H 0,0 k k , [KU2U3] = α21 α2 (λD0,0i i G 1,0 j j H 0,1 k k + µD 0,0 i i G 0,1 j j H 1,0 k k ), [KU2U2] = α 2 1 ( (λ + 2µ + κ)D0,0i i G 1,1 j j H 0,0 k k + (µ + κ) 1 α21 D0,0i i G 0,0 j j H 1,1 k k ) +(µ + κ)D1,1i i G 0,0 j j H 0,0 k k , [KU2φ3 ] = − a 2 κ 2 D 1,0 i i G 0,0 j j H 0,0 k k , [KU2ψ] = a 2 λ0 α1D 0,0 i i G 1,0 j j H 0,0 k k , [KU3ψ] = a 2 λ0 α1 α2 D0,0i i G 0,0 j j H 1,0 k k [Kφ1φ3 ] = α1 α2 (αD1,0i i G 0,0 j j H 0,1 k k + βD 0,1 i i G 0,0 j j H 1,0 k k ), [KU3U3] = α 2 1 ( 1 α22 (λ + 2µ + κ)D0,0i i G 0,0 j j H 1,1 k k + (µ + κ)D 0,0 i i G 1,1 j j H 0,0 k k ) +(µ + κ)D1,1i i G 0,0 j j H 0,0 k k , [Kφ1φ1 ] = (α + β + γ)D 1,1 i i G 0,0 j j H 0,0 k k + a2 4 κ D 0,0 i i G 0,0 j j H 0,0 k k +γ α21(D 0,0 i i G 1,1 j j H 0,0 k k + 1 α21 D0,0i i G 0,0 j j H 1,1 k k ), [Kφ1φ2 ] = α1(αD 1,0 i i G 0,1 j j H 0,0 k k + βD 0,1 i i G 1,0 j j H 0,0 k k ), [Kφ2φ2 ] = α 2 1 ( (α + β + γ)D0,0i i G 1,1 j j H 0,0 k k + γ 1 α21 D0,0i i G 0,0 j j H 1,1 k k ) + γD1,1i i G 0,0 j j H 0,0 k k + a 2 4 κD 0,0 i i G 0,0 j j H 0,0 k k , [Kφ2φ3 ] = α21 α2 (αD0,0i i G 1,0 j j H 0,1 k k + βD 0,0 i i G 0,1 j j H 1,0 k k , [Kφ3φ3 ] = α 2 1 ( 1 α22 (α + β + γ)D0,0i i G 0,0 j j H 1,1 k k + γD 0,0 i i G 1,1 j j H 0,0 k k ) + γD1,1i i G 0,0 j j H 0,0 k k + a 2 4 κD 0,0 i i G 0,0 j j H 0,0 k k , [Kψψ] = a0 ( D1,1i i G 0,0 j j H 0,0 k k + α 2 1D 0,0 i i G 1,1 j j H 0,0 k k + α21 α22 D0,0i i G 0,0 j j H 1,1 k k ) + a 2 4 λ1D 0,0 ii G 0,0 j j H 0,0 kk , here the summation convention is not valid for indices with bar and where, Ds,s̄i i = 1∫ −1 { dsPi(ξ) dξs ds̄Pi(ξ) dξ s̄ } dξ, Gs,s̄j j = 1∫ −1 { dsP j(η) dηs ds̄P j(η) dηs̄ } dη, Hs,s̄k k= 1∫ −1 { dsPk(ζ) dζs ds̄Pk(ζ) dζ s̄ } dζ, (s,s̄=0, 1, i,i, j, j, k,k=1,...,∞). (14) 290 KIRIŞ, İNAN 4. Genetic algorithm and results To obtain the material properties of the microstretch plate, we constructed an optimization problem which minimizes the least square distance between the ex- perimental natural frequencies and computed natural frequencies calculated from the eigenvalue problem. We assume that the first 20 experimental frequencies are (Chiroiu and Munteanu, 2002): {96.54, 115.33, 202.67, 365.65, 432.97, 489.11, 543.3, 774.92, 995.87, 1076.44, 1320.49, 1341.81, 1456.47, 1559.23, 1678, 1805.4, 1988.3, 2290.45, 2654, 2876} (Hz). The constructed optimization problem is solved by using a GA. We assumed that micro inertia and the density of the microstretch plate is given as j = 6.25 10−7 m2, ρ = 1189 kg/m3 and the plate dimensions are 0.160×0.240×0.076m (length, width, height). We have nine material constants shown by {λ, µ, κ, α, β, γ, a0, λ0, λ1} for a microstretch body and each of them can be considered as a property of an individual. So, we can consider that i. individual corresponds to a solution like {λ, µ, κ, α, β, γ, a0, λ0, λ1}i. A population of individuals are called as “Gen- eration”. We assume the population of each generation is 300. After the first 516 generations with the crossover probability % 75 and the mutation proba- bility % 25, the material properties of the microstretch plate are found as λ = 7459MPa, µ = 6175MPa, κ = 34.4MPa, α = 2.3GN, β = 4.3GN γ = 6.1GN, λ0 = 3.1GN, λ1 = 4GN, a0 = 0.2GN. During computations, a P4 1.4GHz computer with 512MB RAM was used and the elapsed time was 453.42 minutes. References Chiroiu V. and Munteanu L. (2002) Estimation of micropolar elastic moduli by inversion of vibrational data, Complexity International 9, 1-10. Eringen, A. C. (1990) Microcontinuum Field Theories: Foundations and Solids, Springer-Verlag. Inan, E. (1990) Elastic waves in damaged material, In Proc. of Fourth ICOVP-99. A, 149-160. Calcutta, India. Mota Soares, C. M., Moreira F. M., Araujo A., and Pedersen P. (1993) Identification of material properties of composite plate specimens, Composite Structures 25, 277-285. Singh, B. and Kumar R. (1998) Wave propagation in a generalized thermo-microstretch elastic solid, Int. J. of Engineering Science 36, 891-912. Tomar, S. K. and Garg M. (2005) Reflection and transmission of waves from a plane interface between two microstretch solid half-spaces, Int. J. of Engineering Science 43, 139-169. Zhou, D., Cheung, Y. K., and Au F. T. K. (2002) 3-D vibration analysis of thick rectangular plates using Chebyshev polynomial and Ritz method, Int. J. of Solids and Structures 39, 6339-6353. 291 291–296. VIBRATIONS OF A CIRCULAR MEMBRANE SUBJECTED TO A PULSE Önder Korfalı and İsmail Burak Parlak Galatasaray University, stanbul, Turkey Abstract. In this paper we have studied the vibrations of a circular membrane subjected to a pressure pulse. The exact solution of the problem is given in terms of Bessel functions using a Laplace transformation. Key words: circular membrane, pressure pulse 1. Introduction For the vibrations of a circular membrane of radius a we must solve the two- dimensional wave equation (Dettman, 1969) ∇2ξ = 1 c2 ∂2ξ ∂t2 (1) in the interior of the membrane; that is, 0 ≤ r < a, 0 ≤ θ ≤ 2π, subject to the boundary condition ξ = 0 on the edge, r = a. We shall specify zero initial displacement, ξ(r, θ, 0) = 0 and zero initial velocity, that is, ξ̇(r, θ, 0) = 0. With the introduction of polar coordinates, the wave equation becomes ∂2ξ ∂r2 + 1 r ∂ξ ∂r + 1 r2 ∂2ξ ∂θ2 = 1 c2 ∂2ξ ∂t2 (2) For a circular membrane ∂ξ ∂θ = 0 , ∂2ξ ∂θ2 = 0 (3) Consequently we have ξ = ξ(r, t) together with ξ(a, t) = 0. Using Eqs. 3 the Eq. 2 reduces to ∂2ξ ∂t2 = c2( ∂2ξ ∂r2 + 1 r ∂ξ ∂r ) + Z ρ (4) where Z is force per unit area and Vibration Problems ICOVP 2005, c© 2007 Springer. İ (eds.), . şE. Inan and A. Kırı 292 KORFALI, PARLAK c2 = T ρ (5) T being the tension in the membrane and ρ, the density of the membrane material. With the introduction of Z = P0δ(t) (6) where δ(t) is impulse function we have ∂2ξ ∂t2 − c2(∂ 2ξ ∂r2 + 1 r ∂ξ ∂r ) = P0δ(t) ρ (7) with initial conditions ξ(r, 0) = 0 and ( ∂ξ ∂t )t=0 = 0 (8) and B.C. ξ(a, t) = 0 (9) 2. Laplace transformation With L{ξ(r, t)} = ξ̄(r, s) and using Eqs. 8 we have L{∂ 2ξ ∂t2 } = s2ξ̄(r, s) − sξ(r, 0) − (∂ξ ∂t )t=0 and L{ ∂2ξ ∂t2 } = s2ξ̄(r, s) (10) Using Eq. 10 together with L{δ(t)} = 1, Laplace transformation of Eq. 7 is ∂2ξ̄(r, s) ∂r2 + 1 r ∂ξ̄(r, s) ∂r − s 2 c2 [ξ̄(r, s) − P0 ρs2 ] = 0 (11) Now using the transformation ȳ(r, s) = ξ̄(r, s) − P0 ρs2 (12) we obtain ∂2ȳ(r, s) ∂r2 + 1 r ∂ȳ(r, s) ∂r − s 2 c2 ȳ(r, s) = 0 (13) which is Bessel’s equation of zeroth order with argument ir( sc ). We know that ξ(0, t) is finite, hence ξ̄(0, s) is finite, consequently ȳ(0, s) must be finite too. The solution of differential Eq. 13 which is finite at r = 0 is ȳ(r, s) = AJ0(ir s c ) (14) VIBRATIONS OF A CIRCULAR MEMBRANE SUBJECTED TO A PULSE 293 To determine the constant A we have ξ(a, t) = 0, consequently ξ̄(a, s) = 0. Using Eq. 12 ȳ(a, s) = ξ̄(a, s) − P0 ρs2 = − P0 ρs2 (15) Using Eq. 15 together with Eq. 14 ȳ(a, s) = AJ0(ia s c ) = − P0 ρs2 or A = − P0 ρs2J0(ia sc ) (16) Substituting this into Eq. 14 we have ȳ(r, s) = − P0 ρs2 J0(ir sc ) J0(ia sc ) (17) Now using Eq. 12 together with Eq. 17 we obtain ξ̄(r, s) = P0J0(ia sc ) − P0J0(ir s c ) ρs2J0(ia sc ) (18) 3. Inverse Laplace transformation Now we will find the inverse transformation (Spiegel, 1965) of ξ̄(r, s). The in- verse transformation of ξ̄(r, s) is determined as the sum of residues of estξ̄(r, s). Therefore we have to calculate the residues of estξ̄(r, s) = [ P0J0(ia sc ) − P0J0(ir s c ) ρs2J0(ia sc ) ]est (19) This has a second order pole at s = 0 and simple poles at the roots of equation J0(ia sn c ) = 0 (20) Putting iasn c = λn where n = 1, 2, . . . ,∞ (21) we have simple poles sn = λnc ia = − iλnc a (22) where λn’s are determined from J0(λn) = 0. The residue R0 at the double pole s = 0 is: 294 KORFALI, PARLAK R0 = lim s→0 d ds {s2[ P0J0(ia sc ) − P0J0(ir s c ) ρs2J0(ia sc ) ]est} (23) Now taking the derivative with respect to s and using the well-known formules (Mathews and Walker, 1965; Selby, 1974) J′0(x) = −J1(x) , J0(0) = 1 and J1(0) = 0 (24) we find that each of the terms within brackets below vanishes. R0 = {[ P0J0(ia sc ) − P0J0(ir s c ) ρJ0(ia sc ) ]test}s=0 + { [− iac P0J1(ia s c ) + P0 ir c J1(ir s c )]ρJ0(ia s c )e st ρ2J20(ia s c ) }s=0 + { [P0J0(ia sc ) − P0J0(ir s c )]ρ ia c J1(ia s c )e st ρ2J20(ia s c ) }s=0 (25) Consequently R0 = 0. Now we will calculate the residues at simple poles sn. Residue at s = sn is lim s→sn (s − sn)[ P0J0(ia sc ) − P0J0(ir s c ) ρs2J0(ia sc ) ]est (26) Using Eq. 20 this reduces to = lim s→sn −(s − sn)[ P0J0(ir sc ) ρs2J0(ia sc ) ]est (27) = [ lim s→sn −(s − sn) J0(ia sc ) ] ︸��������������︷︷��������������︸ = 00 [ lim s→sn P0est J0(ir sc ) ρs2 ] (28) and applying Hospital’s rule = [ 1 ia c J1(ia sn c ) ][ P0J0(ir sn c ) ρs2n esnt] = P0J0(ir sn c )e snt iaρ c s 2 nJ1(ia sn c ) (29) Using ia sn c = λn and ir sn c = λn( r a ) (30) this reduces to VIBRATIONS OF A CIRCULAR MEMBRANE SUBJECTED TO A PULSE 295 = P0J0(λn ra )e − iλnca t iaρ c i2λ2nc2 a2 J1(λn) = iP0J0(λn ra )e − iλnca t ρ caλ 2 nJ1(λn) (31) and passing to summation ξ(r, t) = ∞∑ n=1 iP0J0(λn ra )e − iλnca t ρ caλ 2 nJ1(λn) (32) or in trigonometric form ξ(r, t) = ∞∑ n=1 iP0J0(λn ra )[cos(− λnct a ) + i sin(− λnct a )] ρ caλ 2 nJ1(λn) (33) 4. Conclusion Taking finally the real part of Eq. 33 we have the series solution of the problem ξ(r, t) = ∞∑ n=1 P0J0(λn ra ) sin( λnct a ) ρ caλ 2 nJ1(λn) (34) which satisfies Eq. 8 and Eq. 9, the initial conditions and boundary condition respectively. Using the initial condition, (∂ξ∂t )t=0 = 0 we have from Eq. 34 ( ∂ξ ∂t )t=0 = P0 ρ ∞∑ n=1 J0(λn ra ) λnJ1(λn) = 0 (35) and consequently the important result ∞∑ n=1 J0(λn ra ) λnJ1(λn) ≡ 0 for 0 ≤ r ≤ a (36) where λn’s (n = 1, 2, . . . ,∞) are the roots of J0(λ) = 0. For r = 0 using J0(0) = 1 this simplifies to ∞∑ n=1 1 λnJ1(λn) = 0 (37) another property of Bessel functions. Taking the first term of rapidly converging series of Eq. 34 we have the approximate solution 296 KORFALI, PARLAK ξ(r, t) � P0J0(λ1 ra ) sin( λ1ct a ) ρ caλ 2 1J1(λ1) (38) which satisfies the initial condition, ξ(r, 0) = 0 and the boundary condition, ξ(a, t) = 0 but not the other initial condition, (∂ξ∂t )t=0 = 0. Now inserting λ1 � 2.40 and J1(λ1) � 0.5191 from (Selby, 1974) into Eq. 38 and rearranging we have ξ(r, t) � P0J0(2.40 ra ) sin( 2.4ct a ) ρ ca (2.40) 2(0.5191) (39) and finally ξ(r, t) � P0J0(2.4 ra ) sin( 2.4ct a ) 3ρ ca (40) as the approximate solution of the problem. 5. Symbols J0 :Zeroth order Bessel function. J1 :First order Bessel function. P0 :Magnitude of the pressure pulse. ξ :Vertical displacement of the membrane. References Dettman, J. W. (1969) Mathematical Methods in Physics and Engineering, McGraw-Hill Book Company. Mathews, J. and Walker, R. L. (1965) Mathematical Methods of Physics, W. A. Benjamin, Inc. Selby, S. M. (1974) Standard Mathematical Tables, CRC Press. Spiegel, M. R. (1965) Theory and Problems of Laplace Transforms, Schaum’s Outline Series, McGraw-Hill Book Company. 297 297–303. A METHOD OF DISCRETE TIME INTEGRATION USING BETTI’S RECIPROCAL THEOREM Nahit Kumbasar Faculty of Civil Engineering, İstanbul Technical University, Maslak 34469, İstanbul, Turkey Abstract. There are several known algorithms for the numerical integration of the equation of motions in structural dynamics. However, efforts have been made recently to obtain more efficient and accurate methods of numerical integration. This paper aims to obtain a simple and efficient algorithm, based on Betti’s reciprocal theorem. Key words: Betti’s reciprocal theorem 1. Introduction The differential equation of motion for a multi-degree of freedom system is of the form MẌ + CẊ + KX = F (1) in which M,C,K and F represent mass, damping, stiffness and force in matrix form. M,C and K are constant for linear systems. Superscript dots denote differ- entiation with respect to time. This equation may be uncoupled for some types of damping, using modal analysis, into a series of differential equations that are similar to the equation for single degree of freedom system: mẍ + cẋ + kx = f (2) Several algorithms are known for the numerical integration of equation (1). A sur- vey of these direct time integration methods is given by (Dokanish and Subbaraj, 1989a,b). More elaborate methods have also been published such as those by (Chang, 1997) and (Golly and Amer, 1999). In general, these algorithms consider displacement, velocity (and some of them acceleration) as specified at time t and calculate their values at time t + ∆t. Most of these methods estimate the varia- tion of the displacement in cubic form, such as the well known Newmark-β and Wilson-θ methods (Wilson et al., 1973). Some of these methods use the weighted residual method instead of the equation of motion itself (1). The requirements of an algorithm for discrete time integration are stated by (Hilbert et al., 1977) as: Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 298 KUMBASAR 1. To be unconditionally stable when applied to linear problems. 2. To possess a numerical dissipation that can be controlled by a parameter other than the time step. In particular no numerical dissipation should be possible. 3. To possess a numerical dissipation that does not affect lower modes too strongly. These requirements are essentially needed to get rid of spurious effects for structural systems analysed using finite elements. However, they decrease the accuracy of the integration procedures. The parameters β and θ in Newmark and Wilson methods may be chosen to satisfy the above conditions. In this paper, a simple and effective procedure, obtained using Betti’s reciprocal theorem, is presented. Since the velocity and the acceleration are not included in this proce- dure, the memory size needed is less than those of the known methods. Numerical accuracy and computing time are also better, but the proposed method is only conditionally stable. It is shown that an unconditionally stable procedure may be found by including velocity and acceleration, which gives similar results as the Newmark method. 2. General description Numerical integration of differential equation of motion is mostly performed by the use of algorithms based on the finite difference methods. (Cakiroglu, 1979) proposed an improved finite difference method for structural systems, making use of the Betti’s reciprocal theorem. Differential equation of motion for a sin- gle degree of freedom system may be interpreted as the differential equation of equilibrium for a shear beam on elastic foundation. Displacements of a shear beam are governed by the equation GAu′′ = −p(x) (3) where GA is the shear rigidity of the beam, u is the deflection, p(x) represents the external load and (′′) represents differentiation with respect to the x coordinate. Consider a shear beam, with GA = m, on an elastic foundation producing a dis- tributed reaction proportional to the deflection and another reaction proportional to the slope, with spring constants k and c reciprocally. The differential equation of equilibrium for this beam is mu′′ + cu′ + ku = −p(x) (4) This equation is similar to equation (2). The main difference between this equa- tion and the ordinary shear beam equation is the type of the foundation spring constants, which are not physically meaningful compared to the accustomed ones. However, one may prefer to solve the equation (4) that represents the equilibrium A METHOD OF DISCRETE TIME INTEGRATION 299 of an elastic system, instead of equation (2), since one may use Betti’s reciprocal theorem to obtain a solution in finite difference form. Considering three consecutive points of the shear beam with equal distances h, a displacement pattern, triangular at these points, is assumed as the first system (Figure 1). The actual displacement, which is the solution of the differential equa- tion, is the second system. Corresponding loads are shown at the bottom of each deflection form. Application of Betti’s reciprocal theorem on these systems yields a relation among the three consecutive displacements, similar to a finite difference equation. Figure 1. Figure 2. −mh ui−1 + 2m h ui − m h ui+1 − k.h 12 (ui−1 + 10ui + ui+1) + c(ui−1 − ui+1)/2 = − h12 (pi−1 + 10pi + pi+1) (5) A parabolic variation is assumed for the external load and the actual displacement to evaluate the integrals for this equation. Making use of the two previously com- puted displacement values, one can have an approximate value for the third point’s displacement. Each time, the relation among the three displacements contains only one unknown. To start the algorithm, the first two displacements can be obtained using initial conditions. The integration algorithm is thus formulated. ui+1 = (−kh2 − 12m + 6ch)ui−1 + (−10kh2 + 24m)ui + h2(pi−1 + 10pi + pi+1) kh2 + 12m + 6ch (6) Numerical examples, based on this simple procedure, yields the results, with less computing time, better than those of the classical methods such as linear acceleration or Newmark-β methods. It is clear that one can assume different type of deflections instead of the single triangle in Figure 1. An asymmetrical triangle may be chosen for variable mesh size. Some interesting alternatives will be shortly presented. One of the alternatives is to consider four consecutive points and assume an antisymmetrical variation as shown in Figure 2. It yields ui+1 = 1kh2+12m+6ch ( (−kh2 − 12m + 6ch)ui−2 + (−9kh2 + 36m − 6ch)ui−1 +(9kh2 − 36m − 6ch)ui + h2(pi−2 + 9pi−1 − 9pi − pi+1) ) (7) 300 KUMBASAR which is quite different from (6), but gives exactly the same results. Since it necessitates first three ordinates at the outset it is not preferable. Another alternative is to include velocities at each point. Considering four consecutive points and an antisymmetrical variation for the first system as before, we may use a fourth degree polynomial for the actual displacement in terms of ui−2, vi−2, ui−1, vi−1, and vi. Application of Betti’s theorem through Figure 2 yields ui−2(−12 mh − 2c) + ui−1(12 m h + 2c) + vi−2h(−4 m h − 5 6 c) + vi−1h(−10 m h + kh 2 − 7 3 c) +vih(2 mh + kh 2 + 7 6 c) + h 12 (pi−2 + 9pi−2 − 9pi − pi+2) = 0 (8) from which the velocity vi maybe computed. The displacement of the point i may be found as ui = ui−2 + (vi−2 + 4vi−1 + vi) h 3 This procedure is found to be unconditionally stable, but not so efficient, as will be seen in numerical examples. Yet another alternative is to find an unconditionally stable procedure that pos- sesses numerical dissipation. To this end, the scheme of variation in Figure 1 is used including the velocity v and the acceleration a of the first mesh point. The displacement and the velocity of midpoint are expressed in terms of parameters of previous point, similar to Newmark method: vi+1 = vi + (1 − α)aih + αai+1h ui+1 = ui + vih + (1/2 − β)aih + βai+1h (9) A third degree polynomial for the actual displacement is formed as: u = ui+1 + vi+1t + ai+1 t2 2 + (ai+1 − ai) t3 6h (10) Application of Betti’s theorem through Figure 1 yields −uikh − vih(kh + c) + aih2[(α − β − 13 ) m h − ( 1 2 − β)kh − ( 23 24 − α)c] +ai+1h2[(β − α − 23 ) m h − (β + 1 12 )kh − (α + 1 24 )c] + h 12 (pi−1 + 10pi + pi+1) = 0 (11) from which a2 and through (9) u2 and v2 may be obtained. Similar to Newmark-β method, this procedure is unconditionally stable for (α = 0.5, β = 0.25), but not much improved. 3. Multi-degree of freedom systems Extension of the procedure obtained for single degree of freedom systems to multi-degree of freedom systems is straightforward. One may consider a different A METHOD OF DISCRETE TIME INTEGRATION 301 shear beam for each degree of freedom displacement, connected to 2n springs, to take into account the internal and damping forces, n being degree of freedom. Thus one may obtain each of the equations of motion. This will correspond to replace the scalar quantities m, c and k, with structural mass, damping and stiffness matrices M,C and, K. 4. Numerical examples Numerical examples are arranged to check and compare the efficiency of the proposed method with some well known methods. Some simple problems are considered in order to obtain the theoretical solution easily. 4.1. SINGLE DEGREE OF FREEDOM SYSTEM SUBJECTED TO HARMONIC LOAD The theoretical solution of an undamped single degree of freedom system sub- jected to Cosine load: mü + ku = p0Cos�t is known as u = p0 k(1 − �2 ω2 ) (Cos�t − Cosωt) where ω2 = k m Three different numerical integration procedures are applied to solve the given differential equation. Time step length h is chosen as 0.01sec, one tenth of the natural period which is generally suggested for numerical integrations. The last 20 steps of the 100 step numerical results are presented in Figure 3, for finite difference method, Newmark-β method, and the first alternative (Eq.6) of the proposed method as well as the theoretical solution. In Figure 4, numerical results of the second (Eq.8) and third alternative (Eq.11) are displayed with those of Newmark-β method and the theoretical solution. The difference between third alternative and Newmark-β method is imperceptible. All are quite different from the theoretical solution, especially in the last steps. 302 KUMBASAR Figure 3. Figure 4. Figure 5. A METHOD OF DISCRETE TIME INTEGRATION 303 4.2. THREE DEGREES OF FREEDOM SYSTEM SUBJECTED TO HARMONIC LOADS The theoretical solution of an undamped three degrees of freedom system sub- jected to harmonic loads at each mass, may also be formed. Since the fundamental period of the system is found to be 0.5sec, the time step length is chosen as 0.05sec for this case. In Figure 5, numerical results of the first alternative are displayed with those of Newmark-β method and the theoretical solution as the last 20 steps of 100 time steps. The presented diagrams show that for the suggested time step size, while the results of the suggested method is quite close, those of Newmark method tends to deviate from the theoretical solution. References Chang, S. Y. (1997) Improved numerical dissipation for explicit methods in pseudo dynamic tests, Earthquake Engineering and Structural Dynamics 26, 917-929. Çakıroğlu (1979) A computation method of discrete system for structures (in Turkish), Proceedings of First National Congress of Mechanics. Dokanish, M. A, Subbaraj, K. A. (1989) Survey of direct time integration methods in computational structural dynamics. I. Explicit Methods, Int. Computers and Structures 32, 1371-1386. Dokanish, M. A, Subbaraj, K. A. (1989) Survey of direct time integration methods in computational structural dynamics. II. Implicit Methods, Int. Computers and Structures 32, 1387-1401. Golly, B. W. and Amer, M. (1999) An unconditionally stable time stepping procedure with algo- rithmic damping: a weighted integral approach using two general weight functions, Earthquake Engineering and Structural Dynamics 28, 1345-1360. Hilbert H. M., Hughes T. J. R., Taylor R. L. (1977) Improved numerical dissipation for time inte- gration algorithms in structural dynamics, Earthquake Engineering and Structural Dynamics 5, 283-292. Wilson, E. L., Farhoomand, I., Bathe, K. J. (1973) Nonlinear dynamic analysis of complex structures, Earthquake Engineering and Structural Dynamics 1, 241-252. Zienkiewitz, O. C, Wood, W. L., Taylor, R. L. (1980) An alternative single step algorithm for dynamic problems, Earthquake Engineering and Structural Dynamics 8, 31-40. 305 305–316. BRIDGES IN VIBRATED GRANULAR MEDIA Anita Mehta S N Bose National Centre for Basic Sciences, Block JD, Sector 3, Salt Lake, Calcutta 700098, India Abstract. We study a particular consequence of the dynamics of vibrated granular media, which is the spontaneous formation of stable bridges. Here we examine their geometrical characteristics, and compare the results of a simple theory with those of independent simulations of three-dimensional hard spheres. Our conclusion is that bridges are the signatures of spatiotemporal inhomogeneities, the carriers of the so-called ‘force chains’. Key words: granular media, bridges 1. Introduction Granular media are often referred to as a special state of matter; they flow like liquids, pack like solids, and in the fluidised state, can behave like gases (Mehta, 2006). This results predominantly from the fact that granular media are ather- mal, i.e., grains of sand do not move spontaneously in response to the kinetic energy provided by room temperature. For sand to flow, it needs to be externally perturbed; a particularly convenient form of external perturbation is vibration. We have studied structural properties of vibrated sand extensively in the past, details of which can be found in (Mehta, 1994; Mehta and Halsey, 2003; Mehta and Barker, 1991): our studies concern, among other things, the dependence of volume fraction, coordination number and correlation functions of displacement and position, on the intensity of vibration. These and other structural details, as well as their dependence on dynamics, are of course of crucial importance to engineering applications, which is indeed where the study of granular materials first began (Brown and Richards, 1966). In this article, we focus on recent work on the formation and dynamics of bridges, which are cooperative structures that are ubiquitous in granular media. 2. Definitions Under imposed vibrations, grains can fall independently under gravity to a point of stability; when instead, two or more grains fall together to a position of mutual Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 306 MEHTA support, bridges (Mehta et al., 2004) are formed. These can be stable for arbitrarily long times, since the Brownian motion that would dissolve them away in a liquid is absent in granular systems (which for simplicity and variety, we will often to as ‘sandpiles’ in what follows) – as mentioned above, grains are simply too large for the ambient temperature to have any effect. As a result, bridges can affect sandpile dynamics: our studies (Mehta and Barker, 1991) have shown that in the last stages of compaction in densely packed sand, voids can only be flushed out of the system by the gradual collapse of long-lived bridges. Bridges are also responsible for jamming in granular processes, for example, as grains flow out of a hopper (Brown and Richards, 1966). We first define a bridge in more quantitative terms. Consider a stable packing of hard spheres under gravity, in three dimensions. Each particle typically rests on three others which stabilise it, in the sense that downward motion is impeded. A bridge is a configuration of particles in which the three-point stability conditions of two or more particles are linked; that is, two or more particles are mutually stabilised. Bridges thus cannot be formed sequentially. While it is impossible to determine bridge distributions uniquely from a distribution of particle positions, we are able via our algorithm to obtain the most likely positions of bridges in a given scenario (Mehta et al., 2004). We now distinguish between linear and complex bridges via a comparison of Figures 1 and 2. Figure 1 illustrates a complex bridge, i.e., a mutually stabilised cluster of five particles (shown in green), where the stability is provided by six stable base particles (shown in blue). Of course the whole is embedded in a stable network of grains within the sandpile. Also shown is the network of contacts for the particles in the bridge: we see clearly that three of the particles each have two mutual stabilisations. Figure 2 illustrates a seven particle linear bridge with nine base particles. This is an example of a linear bridge. The contact network shows that this bridge has a simpler topology than that in Figure 1. Here, all of the mutually stabilised particles are in sequence, as in a string. A linear bridge made of n particles therefore always rests on nb = n + 2 base particles. For a complex bridge of size n, the number of base particles is reduced (nb < n + 2), because of the presence of loops in their contact networks. An important point to note is that bridges can only be formed sustainably in the presence of friction; the mutual stabilisations needed would be unstable otherwise! Although our Monte Carlo simulations (described below) do not con- tain friction explicitly, our configurations indirectly include this: in particular, the coordination numbers lie in a range consistent with the presence of friction (Mehta and Barker, 1991; Edwards, 1998; Silbert et al., 2002). VIBRATED SAND 307 Figure 1. A five particle complex bridge, with six base particles (left), and the corresponding contact network (right). Thus n = 5 and nb = 6 < 5 + 2 Figure 2. A seven particle linear bridge with nine base particles (left), and the corresponding contact network (right). Thus n = 7 and nb = 9 = 7 + 2 3. Simulation details In the following, we give details of the simulation algorithm which enables us to model vibrated sand (Mehta and Barker, 1991): its main modelling ingredients involve stochastic grain displacements and collective relaxation from them. This algorithm has three distinct stages. (1) The granular system is dilated in a vertical direction (with free volume be- ing introduced homogeneously throughout the system), and each particle is given a random horizontal displacement; this models the dilation phase of a vibrated sandpile. 308 MEHTA (2) The assembly is compressed in a uniaxial external field representing gravity, using a low-temperature Monte Carlo process. (3) Individual spheres in the assembly are stabilised using a steepest descent ‘drop and roll’ dynamics to find local potential energy minima. Steps (2) and (3) model the quench phase of the vibration, where particles relax to locally stable positions in the presence of gravity. Crucially, during the third phase, the spheres are able to roll in contact with others; mutual stabilisations are thus allowed to arise, mimicking collective effects. The final configuration has a well-defined contact network (Mehta and Barker, 1991) where each sphere is supported by a uniquely defined set of three other spheres. The simulation method recalled above builds a sequence of static packings. Each new packing is built from its predecessor by a random process and the sequence achieves a steady state, where structural descriptors such as the mean packing fraction and the mean coordination number fluctuate about well-defined mean values. The steady-state mean volume fraction Φ typically evolves to values in the range Φ ∼ 0.55 – 0.61, depending on the shaking amplitude; the mean coordination number is always Z ≈ 4.6 ± 0.1. It is known that for frictionless packings in d dimensions, Z = 2d (= 6 for d = 3) (Donev et al., 2004), while for frictional packings the minimal coordination number is Z = d + 1 (= 4 for d = 3) (Edwards, 1998); our configurations thus clearly correspond to those generated in the presence of friction. This is confirmed by the results of molecular dynamics simulations of sphere packings in the limit of high friction, which yield a mean coordination number slightly above 4.5 (Silbert et al., 2002). Each of our configurations includes Ntot ≈ 2200 particles. Segregation is avoided by choosing monodisperse particles: a rough base prevents ordering. A large number of restructuring cycles is needed to reach the steady state for a given shaking amplitude: about 100 stable configurations (picked every 100 cycles in order to avoid correlation effects) are analysed, corresponding to Φ = 0.56 and Φ = 0.58. From these configurations, and following specific prescriptions, our algorithm identifies bridges as clusters of mutually stabilised particles (Mehta et al., 2004). Figure 3 illustrates two characteristic descriptors of bridges used in this work. The main axis of a bridge is defined using triangulation of its base particles as follows: triangles are constructed by choosing all possible connected triplets of base particles, and the vector sum of their normals is defined to be the direction of the main axis of the bridge. The orientation angle Θ is defined as the angle between the main axis and the z-axis. The base extension b is defined as the radius of gyration of the base particles about the z-axis; note that this is distinct from the radius of gyration about the main axis of the bridge. VIBRATED SAND 309 Figure 3. Definition of the angle Θ and the base extension b of a bridge. The main axis makes an angle Θ with the z-axis; the base extension b is the projection of the radius of gyration of the bridge on the x-y plane 4. Bridge sizes and diameters: When does a bridge span a hole? In the following, we present statistics for both linear and complex bridges. While we recognise that bridge formation is a collective dynamical process, we adopt an ergodic viewpoint (Edwards, 1994) here. Inspired by polymer theory (Doi and Edwards, 1986), we visualise a linear bridge as a random chain which grows as a continuous curve, i.e. ‘sequentially’ in terms of its arc length s. (For complex bridges, this simplification is not possible in general – a direct consequence of their branched structure). This replacement of what is in reality a collective phe- nomenon in time by a random walk in space is somewhat analogous to the ‘tube model’ of linear polymers (Doi and Edwards, 1986): both are simple but efficient effective pictures of very complex problems. We first address the question of the length distribution of linear bridges. We define the length distribution fn as the probability that a linear bridge consists of exactly n spheres. We make the simplest and the most natural assumption that a bridge of size n remains linear with some probability p < 1 if an (n + 1)th sphere 310 MEHTA is ‘added’ to it: this leads to the exponential distribution fn = (1 − p)pn. (1) The exponential distribution above can also be derived by means of a contin- uum approach. Here, a linear bridge is viewed as a continuous random curve or ‘string’, parametrised by the arc length s from one of its endpoints. We assume also that such a bridge disappears at a constant rate α per unit length, either by changing from linear to complex or by collapsing. The survival probability S (s) of a linear bridge upto length s thus obeys the rate equation Ṡ = −αS and falls off exponentially, according to S (s) = exp(−αs). Consequently, the probability distribution of the length s of linear bridges reads f (s) = −Ṡ (s) = α exp(−αs), a continuum analogue of (1). This is in good accord with the results of independent simulations, which exhibit an exponential decay of linear bridges of the form (1), with α ≈ 0.99 (Mehta et al., 2004), which is clearly seen until n ≈ 12. Around n ≈ 8, com- plex bridges begin to predominate; these have size distributions which show a power-law decay: fn ∼ n−τ. (2) with τ ≈ 2 (Mehta et al., 2004). We have also measured the diameter Rn of linear and complex bridges of size n, which is such that R2n is the mean squared end-to-end distance. Our data on diameters and size distributions (Mehta et al., 2004) indicate that linear bridges in three dimensions start off as planar self-avoiding walks, which eventually collapse onto each other because of vibrational effects; on the other hand, complex bridges look like 3d percolation clusters. Another issue of interest to us is the jamming potential of a bridge. A measure of this, in the case of a linear bridge, is its the base extension b (see Figure 3); this is the horizontal projection of the ‘span’ of the bridge. Our simulation results (Mehta et al., 2004) indicate that three-dimensional bridges of a given length have a fairly characteristic horizontal extension, making it relatively easy to predict whether or not they would ‘jam’ a given hole. In order to compare our simulations with experiment (To et al., 2001; Liu et al., 1995), we plot in Figure 4 the logarithm of the probability distribution of base extensions p(b) against the (normalised) base extension b/〈b〉. This figure emphasises the exponential tail of the distribution function, and also shows that bridges with small base extensions are unfavoured. We note that this long tail is characteristic of three-dimensional experiments on force chains in granular media (Liu et al., 1995). The sharp drop at the origin as well as the long tail in Figure 4, are observed in normal force distributions obtained via molecular dynamics sim- ulations of particle packings (Erikson, 2002), in the limit of strong deformations. Realising that the measured forces propagate through chains of particles, we use VIBRATED SAND 311 this similarity to suggest that bridges are really just long-lived force chains, which have survived despite strong deformations. We suggest also that with the current availability of 3D visualisation techniques such as NMR (Fukushima and Seidler, 2003), bridge configurations might be an easily measurable and effective tool to probe inhomogeneities in shaken sand. -1.6 -1.2 -0.8 -0.4 0.0 0.4 0.0 0.5 1.0 1.5 2.0 2.5 b/ L og (p (b )) Figure 4. Distribution of base extensions of bridges, for Φ = 0.58. The logarithm of the normalised probability distribution is plotted as a function of the normalised variable b/〈b〉, where 〈b〉 is the mean extension of bridge bases 5. Turning over at the top; how linear bridges form domes Recall that a linear bridge is modelled as a continuous curve, parametrised by its arc length s. We here focus on its most important degree of freedom, the tilt with respect to the horizontal; the azimuthal degree of freedom is neglected. Accordingly, we define the local or link angle θ(s) between the direction of the tangent to the bridge at point s and the horizontal, and the mean angle made by the bridge from its origin up to point s, also with the horizontal: Θ(s) = 1 s ∫ s 0 θ(u) du. (3) The local angle θ(s) so defined may be either positive or negative; it can even change sign along the random curve which represents a linear bridge. Of course, 312 MEHTA the orientation angle Θ measured in our numerical simulations is positive by con- struction, being defined as the angle between the main bridge axis and the z-axis (see Figure 3). (Note that by simple geometry, this ‘zenith angle’ made by the bridge axis with the vertical, equals the mean angle made by the basal plane of the bridge with the horizontal). Our simulations show that the mean angle Θ(s) typically becomes smaller and smaller as the length s of the bridge increases. Small linear bridges are almost never flat (Mehta et al., 2004); as they get longer, assuming that they still stay linear, they get ‘weighed down’, arching over as at the mouth of a hopper (Brown and Richards, 1966). Thus, in addition to our earlier claim that long linear bridges are rare, we claim further here that (if and) when they exist, they typically have flat bases, becoming ‘domes’. We use these insights to write down equations to investigate the angular dis- tribution of linear bridges. These couple the evolution of the local angle θ(s) with local density fluctuations φ(s) at point s (with ′ denoting a derivative with respect to s): θ′ = −aθ − bφ2 + ∆1η1(s), (4) φ′ = −cφ + ∆2η2(s). (5) The effects of vibration on each of θ and φ are represented by two independent white noises η1(s), η2(s), such that 〈ηi(s)η j(s′)〉 = 2 δi j δ(s − s′), (6) whereas the parameters a, ..., ∆2 are assumed to be constant. The phenomenology behind the above equations is the following: the evo- lution of θ(s) is caused, in our effective picture, by the sequential addition of particles to the bridge at its ends. The fluctuations of local density φ at a point s are caused by collective particle motion (Mehta et al., 1996). The first terms on the right-hand side of (4), (5) say that neither θ nor φ is allowed to be arbitrarily large. Their coupling via the second term in (4) arises as follows: if there are density fluctuations φ2 of large magnitude at the tip of a bridge, these will, to a first approximation, ‘weigh the bridge down’, i.e., decrease the angle θ locally. Reasoning as above, we therefore anticipate that for low-intensity vibrations and stable bridges, both density fluctuations φ(s) and link angles θ(s) will be small. Accordingly, we linearise (4), obtaining thus an Ornstein-Uhlenbeck equation: θ′ = −aθ + ∆1η1(s). (7) Let us make the additional assumption that the initial angle θ0, i.e., that observed for very small bridges, is itself Gaussian with variance σ20 = 〈θ 2 0〉. The angle θ(s) is then a Gaussian process with zero mean for any value of the length s. Its VIBRATED SAND 313 correlation function can be easily evaluated to be (Uhlenbeck and Ornstein, 1930): 〈θ(s)θ(s′)〉 = σ2eq e−a|s−s ′| + (σ20 − σ 2 eq)e −a(s+s′). (8) It follows from this that the variance of the link angle is: 〈θ2〉(s) = σ2eq + (σ20 − σ 2 eq)e −2as, (9) We see from the above that orientation correlations decay with a characteristic length given by ξ = 1/a; also, in the limit of an infinite bridge, the variance 〈θ2〉 relaxes to σ2eq = ∆21 a (Mehta et al., 2004). Thus, as the chain gets longer, the variance of the link angle relaxes from its initial value of σ20 (i.e. that for the initial link) to σ2eq for infinitely long chains. Given the above, it can be shown that the mean angle Θ(s) will also have a Gaussian distribution. Its variance can be derived by inserting (8) into (3): 〈Θ2〉(s) = 2σ2eq as − 1 + e−as a2s2 + (σ20 − σ 2 eq) (1 − e−as)2 a2s2 . (10) The asymptotic result 〈Θ2〉(s) ≈ 2σ2eq as ≈ 2∆21 a2s , (11) confirms our earlier statement that the longest bridges form domes, i.e. they have bases that are almost flat. Each such bridge can be viewed as consisting of a large number as = s/ξ � 1 of independent ‘blobs’ of length ξ; this result suggests, yet again, strong analogies between linear bridges and linear polymers (Doi and Edwards, 1986). The result (11) has another interpretation. As Θ(s) is small with high proba- bility for a very long bridge, its extension in the vertical direction reads approxi- mately Z = z(s) − z(0) ≈ sΘ(s), (12) so that 〈Z2〉 ≈ s2〈Θ2〉(s) ≈ 2(∆1/a)2 s. Switching back to a discrete picture of an n-link chain, we have Zn ∼ n1/2. (13) The vertical extension of a linear bridge is thus found to grow with the usual random-walk exponent 1/2, in agreement with experiments on two-dimensional arches (To et al., 2001). Our observations on horizontal extensions of three- di- mensional bridges have yielded (Mehta et al., 2004) a non-trivial exponent νlin ≈ 0.66. Putting all of this together, our results predict that long linear bridges are domelike; also, they are vertically diffusive but horizontally superdiffusive. Ev- idently, jamming in a three-dimensional hopper would be caused by the planar projection of such a dome. 314 MEHTA We now compare the results of this simple theory with data on bridge struc- tures obtained from independent numerical simulations of shaken hard sphere packings (Mehta and Barker, 1991). Figure 5 confirms that the mean angle is Gaussian to a good approximation, while Figure 6 shows the measured size de- pendence of the variance 〈Θ2〉(s). The numerical data are found to agree well with a common fit to the first (stationary) term of (10) - the ‘transient’ effects of the second term of (10) are too small to be significant at our present accuracy. We thus conclude that our simple theory captures the principal structural features of linear bridges. Figure 5. Plot of the normalised distribution of the mean angle Θ (in radians) of linear bridges of size n = 4, for both volume fractions. The sinΘ Jacobian has been duly divided out, explaining thus the larger statistical errors at small angles. Full lines: common fit to (half) a Gaussian law 6. Discussion In the above, we have studied the kinetics of bridge formation in sandpiles, as well as examined their structure. Their classification into linear and complex types enables a detailed examination of their geometrical, as well as their stability and spanning, properties. Such information, while useful in a physics context, is no less useful in an engineering one, from the point of view of the design of real bridges: it would be interesting to see such a relationship concretised from an engineering viewpoint. VIBRATED SAND 315 Figure 6. Plot of the variance of the mean angle of a linear bridge, against size n, for both volume fractions. Full line: common fit to the first (stationary) term of (10), yielding σ2eq = 0.093 and a = 0.55. The ‘transient’ effects of the second term of (10) are invisible with the present accuracy Another issue that our study clarifies is that of inhomogeneities in granular systems. It is well known that forces are not transmitted homogeneously in sand- piles (Mehta and Halsey, 2003); they follow typically stringlike paths (Behringer, 2005) depending on the geometry of the underlying granular configuration. The configurational properties of these force chains have been found to be in close agreement with those of the linear bridges above, leading to the suggestion that the latter are the geometrical objects which sustain the major stresses in the sandpile. Our tentative conclusion is thus that long-lived bridges are natural indicators of sustained inhomogeneities in granular systems. References Barker G. C., Mehta A. 1992 Phys. Rev. A 45 3435. Barker G. C., Mehta A. 1993 Phys. Rev. E 47, 184. Behringer R. P. 2005 Nature, to appear. Brown R. L., Richards J. C. 1966 Principles of Powder Mechanics (Oxford: Pergamon). de Gennes P. G. 1999, Rev. Mod. Phys. 71 S374. Doi M., Edwards S. F. 1986 The Theory of Polymer Dynamics (Oxford: Clarendon). Donev A. et al. 2004 Science 303 990. 316 MEHTA Edwards S. F. 1994 in Granular Matter: An Interdisciplinary Approach, ed. Mehta, A. (Springer, New York). Edwards S. F. 1998 Physica 249 226. Erikson J. M. et al. 2002 Phys. Rev. E 66 040301. Fukushima E., Seider G. T. 2003 see chapters in Challenges in Granular Physics edited by Mehta A and Halsey T C (Singapore: World Scientific). Jaeger H. M., Nagel S. R., Behringer R. P. 1996 Rev. Mod. Phys. 68 1259. Liu C. H. et al. 1995 Science 269 513. Mehta A., Barker G. C. 1991 Phys. Rev. Lett. 67 394. Mehta A. 1994 in Granular Matter: An Interdisciplinary Approach, 1994 ed. Mehta A., (Springer, New York). Mehta A., Luck J. M., Needs R. J. 1996 Phys. Rev. E 53 92. Mehta A., Halsey T. C. 2003, see, e.g. Challenges in Granular Physics edited by Mehta A and Halsey T C, 2003, (Singapore: World Scientific). Mehta A., Barker G. C., Luck J. M. 2004 JSTAT P10014 Mehta A. 2006. The Physics of Granular Materials, 2006 (Cambridge University Press, Cam- bridge), to appear. Mueth D. M., Jaeger H. M., Nagel S. R. 1998 Phys. Rev. E 57 3164. O’Hern C. S. et al. 2002 Phys. Rev. Lett. 88 075507. Silbert L. E. et al. 2002 Phys. Rev. E 65 031304. To K., Lai P. Y., Pak H. K. 2001 Phys. Rev. Lett. 86 71. Uhlenbeck G. E. and Ornstein L. S. 1930 Phys. Rev. 36 823. 317 317–322. SOLUTIONS FOR DYNAMIC VARIANTS OF ESHELBY’S INCLUSION PROBLEM Thomas M. Michelitsch,1 Harm Askes,1 Jizeng Wang2 and Valery M. Levin3 1Department of Civil and Structural Engineering, The University of Sheffield, Mappin Street, Sheffield S1 3JD, United Kingdom 2Max-Planck Institute for Metals Research, Heisenbergtrasse 3, D-70569 Stuttgart, Germany 3Instituto Mexicano del Petroleo, C.P.07730, Mexico D.F., Mexico Abstract. The dynamic variant of Eshelby’s inclusion problem plays a crucial role in many areas of mechanics and theoretical physics. Because of its mathematical complexity, dynamic variants of the inclusion problems so far are only little touched. In this paper we derive solutions for dynamic variants of the Eshelby inclusion problem for arbitrary scalar source densities of the eigenstrain. We study a series of examples of Eshelby problems which are of interest for applications in materials sciences, such as for instance cubic and prismatic inclusions. The method which covers both the time and frequency domain is especially useful for dynamically transforming inclusions of any shape. Key words: dynamic variants of Eshelby inclusion problem, Helmholtz potentials, retarded poten- tials, wave equation, Helmholtz equation 1. Introduction Due to the mathematical complexity of the dynamic variant of Eshelby problem there are only a few cases for which closed form solutions exist, namely for homo- geneous spherical inclusions (Mikata and Nemat Nasser 1990). For homogeneous ellipsoidal inclusions the solution of the dynamic Eshelby problem was reduced recently to simple surface integrals (Michelitsch et al., 2003). 2. Dynamic Eshelby inclusion problem We consider a homogeneous material with elastic constants Ci jrs. The constitutive relations are σi j = Ci jrs(�rs − �∗rs) (1) Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 318 MICHELITSCH, ASKES, WANG, LEVIN where σ, �, �∗,C denote stress, strain, eigenstrain and the tensor of elastic con- stants, respectively, with �rs = 1 2 (∂rus + ∂sur) (2) The inclusion S is assumed to undergo a uniform time dependent transforma- tion with a spatially non-uniform eigenstrain �∗ which has the form �∗(r, t) = f (t)ρ(r)�0 (3) and with a ”density” function ρ given by ρ(r) = Θs(r)χ(r) (4) where r = (xi) = (x, y, z) denotes the space point and Θs(r) denotes the char- acteristic function of inclusion1 and �0 is a symmetric and constant tensor. In (4) we have introduced the scalar function χ(r) which characterizes the spatial variation of the eigenstrain. The function f (t) is an arbitrary function of time2 which describes the transformation evolution of the eigenstrain. We assume the absence of external body forces. The displacement field then can be written as (Wang et al., 2005) ul(r, ξ) = −Ck jrs�0rs∂ jGkl(r, ξ) (5) where ξ denotes a set of geometric characteristics of the inclusions (for instance semi axes in case of an ellipsoidal inclusion) and G(r, ξ, t) = 1 µ g2(r, ξ, t)1 + 1 ρm ∇ ⊗ ∇{h1(r, ξ, t) − h2(r, ξ, t)} (6) in the case of an elastic isotropic medium. The functions gi(r, ξ, t) and hi(r, ξ, t) are determined by (∆ − 1 c2i ∂2 ∂t2 )gi(r, ξ, t) + δ(t)ρS (r) = 0 (7) and ∂2 ∂t2 hi(r, ξ, t) = gi(r, ξ, t) (8) are causal functions (retarded potentials) being nonzero only for t > 0, i.e. after the δ(t) excitation. The strain becomes 1 Θs(r) = 1, r ∈ S and Θs(r) = 0, r � S 2 It is only assumed that ∫ ∞ −∞ f (t)dt < ∞. DYNAMIC ESHELBY INCLUSION PROBLEMS 319 �il(r, ξ) = −Ck jrs�0rs(Pi jkl(r, ξ))(il) (9) where (li) indicates symmetrization with respect to the subscripts il. Moreover it is denoted Pi jkl(r, ξ) = ∂i∂ jGkl(r, ξ) (10) Using (9) we define the dynamic Eshelby tensor S analogously to statics by �il(r, ξ) = S ilrs(r, ξ)�0rs (11) where the dynamic Eshelby tensor S is given by S ilrs(r, ξ) = −Ck jrs(Pi jkl(r, ξ))(il) (12) This relation covers both, the time and frequency domain. It holds for arbitrary density functions ρS (r) and hence for arbitrary shapes of the inclusion. Pi jkl is a spatially non-uniform tensor function inside an ellipsoidal inclusion with homo- geneous eigenstrain. When we expand Pi jkl in the frequency domain into a series with respect to frequency ω, the zero order in ω corresponds to the static Eshelby tensor. 3. Results and discussion The retarded potential defined in (7) is given by the convolution g(r) = ∫ ĝ(r − r′)ρS (r′)d3r′ (13) In the frequency domain ĝ(R, β) = e iβR 4πR denotes the Helmholtz Green’s func- tion3 and in the time domain ĝ(R, t) = δ(t− Rc ) 4πR denotes the retarded Green’s func- tion4. For the numerical evaluation of the δ-function in the time domain Green’s function we make use of δ(t − Rc ) 4πR = lim �→0+ e−(t−R/c) 2/(4�) 8πR √ π� (14) To compute integral (13) we use the Gauss-Chebyshev quadrature formula (e.g. Press et al. (1992)). 3 ĝ(R, β) is defined by (∆ + β2)ĝ(r, β) + δ3(r) = 0. 4 ĝ(R, t) is defined by (∆ − 1 c2 ∂2 ∂t2 )ĝ(r, t) + δ3(r)δ(t) = 0. 320 MICHELITSCH, ASKES, WANG, LEVIN In Figs. 1 the retarded potential of a homogeneous solid sphere is drawn ac- cording to a source density δ(t)Θ(a − r) (a = inclusion radius). It is available in closed form (Wang et al., 2005) g(r, a, t) = ginΘ(a − r) + goutΘ(r − a) gin(r, a, t) = c2tΘ(a − r − ct) + c4r ( a2 − (ct − r)2 ) Θ ( r2 − (ct − a)2 ) gout(r, a, t) = c4r ( a2 − (ct − r)2 ) Θ ( a2 − (ct − r)2 ) (15) where gin and gout denote the potentials of the internal space (r < a) and the external space (r > a), respectively. Figure. 1 show that the excitation δ(t)Θ(a− r) generates an outgoing spherical wave package of wavelength 2a which remains constant due to the absence of dispersion. Due to the finite propagation velocity c the wave package arrives at an external point r at (r − a)/c which is the runtime form the closest source point located on the boundary of the sphere. Moreover a remarkable superposition effect takes place in the internal space at spacepoint r for t < (a − r)/c: In this time range the potential increases linearly in time (for 0 < t < (a − r)/c is gin = c2t). In Fig. 2 the real part g (frequency domain) for an inhomogeneous ellipsoidal inclusion of density (ρS = Θ(1 − P)χ, χ = 1, x, x2, resp.) is drawn. The symmetry x ↔ −x of χ = 1, x2 is reflected by the potential. Moreover the effect of breaking this symmetry is shown for χ = x. In the frequency domain spatial oscillations occur due to a finite frequency (β = ωc ). The frequency domain representation (real part) of g for a homogeneous tri- angular prismatic inclusion is shown in Figs. 3. The prismatic inclusion covers the region |x| < 1, 0 ≤ y ≤ 1 − |x|, |z| < 1. 0 1 2 3 4 5 6 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 r g ct=0 ct=a ct=a/2 ct=2a ct=3a ct=4a ct=5a 0 1 2 3 4 5 6 0 1 2 3 4 0 0.5 1 r ct/a g a=1,c=1,γ=0 Figure 1. Time evolution of the retarded potential of a solid spherical source with radius a = 1 (Eq. (15)) for different times t (c = 1) DYNAMIC ESHELBY INCLUSION PROBLEMS 321 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −0.07 −0.06 −0.05 −0.04 −0.03 −0.02 −0.01 0 0.01 x g ρ=ex ρ=1 ρ=e−x 2 Figure 2. Helmholtz potentials (frequency domain β = ωc ) for ellipsoidal sources for differ- ent densities ρS = Θ(1 − P)χ(x), P2 = 3∑ i=1 x2i a2i , χ = {1, exp (x), exp (−x2)}, resp., vs. r=(x,0,0). (a1 = 1, a2 = 0.7, a3 = 0.4, β = 10) Figure 3. Cross section of the Helmholtz potential of a prismatic source of unit density vs. r=(x,y,0) (β = 10) 4. Conclusion The dynamic variant of the Eshelby tensor of a three-dimensional infinite isotropic medium with a spatially inhomogeneous, dynamically transforming inclusion with eigenstrain (3) was derived. In the time domain the problem was reduced to de- termine the retarded potentials defined by (7). Examples of inclusions of different 322 MICHELITSCH, ASKES, WANG, LEVIN shapes and densities were considered such as spheres, ellipsoids, cubes and trian- gular prisms. The dynamic Eshelby tensor is a main cornerstone for the solution of wide range of dynamical problems in theoretical physics. References Eshelby J. D. (1957) The determination of the elastic field of an ellipsoidal inclusion, and related problems, Proc. Roy. Soc. Lond. A 241 376-396. Levin V. M., Michelitsch T. M., Gao H. (2002) Propagation of electroacoustic waves in the transversely isotropic piezoelectric medium reinforced by randomly distributed cylindrical inhomogeneities, Int. J. Solids Structures 39 5013-5051. Michelitsch T. M., Levin V. M., Gao H. (2002c) Dynamic potentials and Green’s functions of a quasi-plane piezoelectric medium with inclusion, Proc. Roy. Soc. Lond. A 458 2393-2415. Michelitsch T. M., Levin V. M., Gao H. (2003) Dynamic Eshelby tensor and potentials for ellipsoidal inclusions, Proceedings of the Royal Society of London A 459 863-890. Mikata Y., Nemat-Nasser S. (1990) Elastic field due to a Dynamically Transforming Spherical Inclusion, J. Appl. Mech. ASME 57 845-849. Press W. H., Teukolsky S. A, Vetterling W. T., Flannery, B. P. (1992) Numerical Recipes in Fortran 77: the Art of Scientific Computing (Second Edition), Cambridge University Press. Pao Y.H. (Ed.) (1978) Elastic Waves and non-destructive testing of materials, AMD-Vol. ASME 29. Wang J., Michelitsch T. M., Gao H., Levin V. M. (2005) On the solution of the dynamic Eshelby problem for inclusions of various shapes, International Journal of Solids and Structures 42 353-363. 323 323–326. NOTE ON LARGE DEFLECTIONS OF CLAMPED ELLIPTIC PLATES UNDER UNIFORM LOAD Subrata Mondal Jalpaiguri Govt. Engineering College, Jalpaiguri, West Bengal, India Abstract. Large deflections of Clamped Elliptic Plates under Uniform Load have been investigated by many authors among which the work of Nash and (Nash and Cooley, 1959) needs special men- tion. As a result of bending the edge of the plate tends to decrease and if the boundary of the plate is immovable, the plate is subjected to tension along the edges. The question then arises as to whether the divergence of Nash’s results from the Linear Theory is mainly due to this induced radial tension or to using a non Linear Theory. The object of this paper is to investigate this point. Key words: eliptic plates, large deformations 1. Analysis Let us consider a clamped thin Elliptic Plate of major axis 2a, minor axis 2b and thickness h. The plate is also immovable, i.e., the radial movement of the plate is prevented. If T0 be the radial tension per unit circumferential length, the differential equation for the normal displacement W takes the following form ∇4W − T0 D ∇2W = q D (1) where q is the uniform load, D is the flexural rigidity of the plate and ∇2 = ∂ 2 ∂x2 + ∂2 ∂y2 A particular integral of (1) can be written as W0 = − q 4Dα2 (x2 + y2) (2) where α2 = T0D assumed constant. For the complementary function of (1) we rewrite (1) in the following form ∇2(∇2 − α2)W = 0 Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 324 MONDAL Let W = W1 + W2. Thus we have, ∇2W1 = 0 (3) and ∇2W2 − α2W2 = 0 (4) Let us now introduce elliptic co-ordinates (ξ, η) defined by x + iy = d cosh(ξ + iη) where 2d is the interfocal distance of the ellipse. Now we get, W1 = α∑ m=0 C2m cosh 2mξ cos 2m η (5) W2 = α∑ m=0 C2mCe2m(ξ,−q1)Ce2m(η,−q1) (6) where Ce2m(η,−q1) and Ce2m(ξ,−q1) are Mathieu function and Modified Mathieu function of the first kind of order 2m and q 1 4 = α 2d2 Thus the general solution determining the normal displacement W can be put as W = W1 + W2 + W0 Using the approximation proposed by (Nagdi, 1955) we can write the final solu- tion as W = C1Ce0 (ξ,−q1)Ce0 (η,−q1) − qd2 8Dα2 (cosh 2ξ + cos 2η) + C2 (7) If the boundary of the plate be clamped at ξ = ξ0, we have W = 0 at ξ = ξ0 and ∂w ∂ξ = 0 at ξ = ξ0. Using the above boundary conditions, the equations to determine the two unknown constants C1 and C2, multiplying these equations by Ce0 (η j,−q1) and integrating with respect η from 0 to 2π and using the orthogonality relations and normalization (Mclachlan, 1947), we get C1 and C2 Here A0(0) and A2(0) are the first two Fourier Coefficient in the expression of Ce0 (η,−q1). Thus W is completely determined. = T0d2 Eh (1 − ν2) sinh 2ξ0 (8) It is to be noted that in evaluating the integrals in the left hand side of (8) we have used the results (Mclachlan, 1947) Ce0 (ξ,−q1) = α∑ r=0 (−1)rA2r(0) cosh 2rξ Ce0 (η,−q1) = α∑ r=0 (−1)rA2r(0) cosh 2rη LARGE DEFLECTIONS OF CLAMPED ELLIPTIC PLATES 325 Where, A2r(0) are the Fourier Coefficient. Finding ∂w∂ξ , ∂w ∂η term by term integra- tions have been carried out determining α2 = T0D . 2. Numerical results Numerical results have been carried for the ellipse a = 2b. To determine the deflection at any point of the ellipse one has to start with assumed values of αd (Here αd has been assumed to be αd = 1, 2, 3 . . .) in the equation (16) leading to the particular values of the load function qa4Eh4 . These values of αd and qa4Eh4 will determine finally central deflections W0 h from equation (11). If αd is known, α a is also known for the ellipse a = 2b. Load vs. central deflection is graphically shown. Figure 1. 326 MONDAL 3. Conclusion From the load deflection curve it is clear that the results of the present study are in very good agreement with those given by (Nash and Cooley, 1959). Thus the large deflection terms have little effect within the range considered. Acknowledgement The Author gratefully acknowledges the receipt of Financial Support from State Project Facilitation Unit, West Bengal to enable him to attend the ICOVP 7th Conference to be held at the ISIK University, Istanbul, Turkey during September 5 - 9, 2005. References Mclachlan, N. W. (1947) Theory and application of Mathieu Function. Nagdi, P. M. (1955) Journal of Applied Mechanics 22, 89-94. Nash, W. A., and Cooley, I. D. (1959) Large Deflection of elliptic plates, ASME 26-2. Timoshenko S. P., Woinowisky-Krieger, (1959) Theory of plates and shells, 2nd Edn. McGraw-Hiil, Newyork. 327 327–336. LOCALISED DEFECT MODES AND A MACRO-CELL ANALYSIS FOR DYNAMIC LATTICE STRUCTURES WITH DEFECTS A. B. Movchan, S. Haq and N. V. Movchan Department of Mathematical Sciences, University of Liverpool, Liverpool L69 3BX, UK Abstract. This paper presents an analysis of inertial lattice structures with periodically distributed defects. Such structures exhibit stop bands for certain ranges of frequencies, prohibiting propagation of waves of such frequencies through the structure. In the engineering literature, structures of this type are referred to as “passive mass dampers”. In particular, we consider a problem of accumulation of damage within the vibrating micro- structure. The eigensolutions of the corresponding spectral problem, set for a sufficiently large macro-cell, are sought in the form of Bloch-Floquet waves. A criterion is introduced for forma- tion of the damage pattern within the macro-cell. The damaged macro-cell shows new spectral properties, which are illustrated with the use of dispersion diagrams. A particular attention is given to highly localised vibration modes associated with the presence of defects. The other type of damaged structures discussed in the paper involves layered materials with small variation of density (due to change in porosity from layer to layer). We describe an analytical algorithm which is used to analyse the spectral properties of such structures and to determine the position of stop bands for Bloch-Floquet waves. 1. Introduction The paper considers periodic discrete structures with defects and analyses how the accumulation of damage within such structures influences their spectral prop- erties. In particular, we are interested in existence or otherwise of the so-called stop bands, intervals of frequencies for which no propagating modes are possible. This paper was motivated by earlier publications (Balk, 2001-Zalipaev, 2002). In particular, (Balk et al., 2001a, 2001b) present analysis of non-linear waves in lattice structures with non-monotonic stress-strain law. The paper (Martinsson and Movchan, 2003) and the book (Movchan et al., 2002) give the details of spec- tral analysis for inhomogeneous lattice structures and asymptotic models relating high-contrast continuum composites and lattice structures. The structure of the present paper is as follows. In Section 2 we employ a macro-cell analysis to study a two-dimensional bi-atomic lattice consisting of particles connected by pairs of springs. We assume that when the deformation of the link connecting the neighbouring particles reaches a certain level, one of Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 328 MOVCHAN, HAQ, MOVCHAN the springs breaks thus changing the stiffness of the link and creating a defect in the lattice. The eigensolutions of the corresponding spectral problem in a macro- cell are sought in the form of Bloch-Floquet waves. The corresponding dispersion diagrams show the existence of vibration modes localised in the vicinity of the region occupied by defects. In Section 3 we consider a periodic lattice structure modelling a porous lay- ered medium. The damage in the lattice is introduced by slightly varying the porosity of each layers with a period which differs from the period of the lattice, so that the damage is accumulated over a macro-cell whose size can be quite large compared to the size of the elementary cell in the original (undamaged) lattice. We analyse the spectral properties of the damaged lattice and obtain analytic estimates for the position and width of stop bands as well as frequencies of the localised modes. 2. Accumulation of defects in a two-dimensional bi-atomic lattice Consider a bi-atomic square lattice which consists of particles of mass m(1) and m(2) connected by pairs of parallel springs of equal stiffness. The position of each particle within the lattice is described by a multi-index n = (m, n) ∈ Z2, the stiffness coefficient of the elastic link between the neighbouring particles n and n′ is denoted by Cnn′ , and the distance between the particles n and n′ is l. We consider Bloch-Floquet waves in a square macro-cell generated by N2 unit cells, and denote the amplitude of the displacement by u(n). The equations of motion then have the form m(n)ω2u(n) + ∑ n′ Cnn′ l (u(n) − u(n′)) = 0, (1) where ω is the radian frequency of time-harmonic (out-of-plane) vibrations, and the summation is taken over all nearest neighbours n′ of the particle n. For any particle n within the macro-cell the Bloch-Floquet conditions have the form u(n + m) = u(n)eiK.m, (2) where K = (K1, K2) is the Bloch vector, m = p1a1 + p2a2. The multi-index p = (p1, p2) has integer components, and a1, a2 are the basis translation vectors that define the shape and the size of the macro-cell. The system (1) can be written in the matrix form ω2Mu = σ(K)u, (3) where σ(K) is the N2 ×N2 stiffness matrix, and M is the diagonal matrix of mass. This system has non-trivial solutions if and only if det{Mω2 − σ(K)} = 0. (4) LOCALISED DEFECT MODES AND A MACRO-CELL ANALYSIS FOR 329 Equation (4) is the dispersion equation which links the radian frequency ω and the Bloch parameter K. 2.1. DEFECTS IN THE MACRO-CELL Defects in the lattice are created by reducing the stiffness coefficients of elastic links between the particles. Since the stiffness matrix σ(K) of the macro-cell changes with the presence of defects, solutions of the dispersion equation (4) will also change. For the sake of illustration, we assume that the two springs connecting the neighbouring particles are of the same stiffness, but one of the springs is characterised by a lower value of critical stress compared to the other spring. When the strain reaches the critical value, this spring breaks and hence the elastic link between the two particles reduces its stiffness (see Fig. 1). 0 critical σ criticalε (Strain)ε (Stress)σ One spring is being broken Single spring connection Double spring connection Figure 1. Stress-strain relation We analyse evolution of damage within the lattice structure for the case of standing waves when dωd|K| = 0. For numerical simulations we adopt the following procedure: − We first solve the dispersion equation (4). − According to (3) we then find an eigenvector u corresponding to standing waves within the macro-cell, the Bloch vector is chosen as K = KA (see Fig. 2). − Then we analyse the strain matrices for the elastic links within the macro- cell, find the weakest link where the deformation attains its maximum value, break one spring in the corresponding double connection and thus create a defect. 330 MOVCHAN, HAQ, MOVCHAN B C − − x y A π d π d π d π d Figure 2. Irreducible Brillouin zone in the reciprocal lattice − Equations (3) and (4) are then analysed for the new discrete system with a defect and the steps 1, 2 and 3 are repeated again. − All single springs connections remain intact. The iterations stop when defor- mations within double spring links does not exceed those in the single spring connections. The next section presents the results of numerical simulations. 2.2. NUMERICAL SIMULATIONS FOR A LATTICE WITH DEFECTS In the numerical example, we consider a square macro-cell consisting of 36 atoms. The masses of particles within the bi-atomic structure are m(1) = 4, m(2) = 1. We choose the Bloch-Floquet eigensolutions that represent standing waves; in the present numerical simulation, the Bloch vector K is the position vector of the point A in Fig. 2. In Fig. 3 we show a localised eigenmode in the lattice structure with defects. The corresponding dispersion curve (marked by ×) is nearly flat, and it represents a standing wave. In Fig. 4 the dispersion diagrams and the structures with defects are shown for different stages of damage. The standing waves, studied in this example, are aligned along the vertical coordinate axis. For a damaged macro-cell, we observe a high frequency band gap associated with the presence of defects. After 18 itera- tions we arrive to the structure shown in Fig. 4d, which contains weak layers of single-spring connections along the vertical coordinate axis. In Fig. 5 we plot projections of the dispersion curves onto the ω-axis. Each ver- tical line corresponds to one iteration in the series, single dots represent LOCALISED DEFECT MODES AND A MACRO-CELL ANALYSIS FOR 331 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.5 1 1.5 2 2.5 A B C A w X (a) (b) 1 2 3 4 5 6 1 2 3 4 5 6 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 (c) Figure 3. Localised vibration mode is shown in diagram (c); the corresponding dispersion curve is marked by × in the diagram (a), the damaged links are circled in diagram (b) localised modes associated with the presence of defects. As expected, the ad- missible range of frequencies is reduced with the decrease of the overall stiffness within the macro-cell. 3. High-contrast stratified structure with defects For a high-contrast stratified medium, the weak layers are replaced by springs, and stiff porous layers are replaced by rigid solids. The porosity is assumed to be different from layer to layer, and a small variation of porosity results in the perturbation of the mass density. The corresponding structure is shown in Fig. 6. Here we consider a type of damage where defects are generated via the change of porosity of the layers within the array. In the unperturbed state, an elementary cell of periodicity contains two porous layers connected by elastic springs. The per- turbed periodic structure is characterised by a larger period, and the corresponding macro-cell contains more than two layers of different porosity. For such structures, we study the transmission properties. 332 MOVCHAN, HAQ, MOVCHAN −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.5 1 1.5 2 2.5 w A B C A (a) (b) −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.5 1 1.5 2 2.5 w A B C A (c) (d) Figure 4. Dispersion diagrams for different stages of the iterative procedure and the corresponding structures with defects. The parameters are m(1) = 4, m(2) = 1, l = 1 and d = 0.5. Diagrams (a), (b) correspond to the 10th iteration, and diagrams (c), (d) to the 18th iteration 3.1. MACRO-CELL OF THE PERTURBED STRUCTURE Let m(κ), κ = 1, 2, be mass densities of the layers in an elementary cell of the unperturbed structure, of period T (old) = 2. Let f (1) and f (2) be bounded periodic functions of period T ( f ); the perturbation of mass densities of the layers is defined by m( j)n = m ( j) + � f ( j)(n), j = 1, 2, (5) where � is a small non-dimensional parameter. For example, we can choose f (1)(x) = |cos(2x − 1)πα|, f (2)(x) = |cos2xπα|, where α = L/N is rational (L, N ∈ Z), and T ( f ) = 1/(2α). The period T (new) = NT (old) may be large, compared to the original size of the elementary cell. For time-harmonic vibrations of radian frequency ω, the equations of motion for the LOCALISED DEFECT MODES AND A MACRO-CELL ANALYSIS FOR 333 0 2 4 6 8 10 12 14 16 18 20 0 0.5 1 1.5 2 2.5 N (iterations) w Figure 5. The structure of the spectrum in various stages of two-dimensional square lattice with defects Figure 6. Porous layered structure with a periodic porosity function perturbed problem are (m(1) + � f (1)(n))ω2u(n) = c l ∑ n′ ( u(n) − v(n′)), (m(2) + � f (2)(n))ω2v(n) = c l ∑ n′ ( v(n) − u(n′)), (6) where u(n) and v(n) are amplitudes of vibration of layers of mass density m( j)n , n′ are coordinates of nearest neighbours to n, and c is the stiffness coefficient. At the boundary of such a macro-cell we set the Bloch-Floquet conditions u(n + N + 1) = u(n)e2iKNl, v(n − 1) = v(n + N)e−2iKNl, (7) where K is the Bloch parameter. In the matrix form, the equations of motion can be written as Mω2u = Σ(K)u, (8) 334 MOVCHAN, HAQ, MOVCHAN where Σ(K) is the stiffness matrix, and M is the diagonal matrix of mass densities. The corresponding dispersion equation, which relates ω and K, is given by det{Mω2 − Σ(K)} = 0. (9) 3.2. TRANSMISSION MATRICES Here we consider a macro-cell of the size T (new) containing 2N particles, as de- scribed above. Using the results of (Felbacq et al., 1998), (Lekner, 1994) we develop an analytical approach for analysis of transmission properties of such a macro-cell and link it with the structure of band gaps on the dispersion diagram associated with equation (9). 3.2.1. Unperturbed structure Let F(n) = ( v(n) u(n + 1) − v(n) ) , n = 1, 2, 3, . . . , N. Then, for a macro-cell consisting of N elementary cells, we can write F(N + 1) = T N F(1). (10) Here T is the transmission matrix for the elementary cell, T = 1 − m(1)ω2a 2 − m(1)ω2 a m(1)m(2)ω4 a2 − (m (1)+m(2))ω2 a (2 − m(1)ω2 a )(1 − m(2)ω2 a ) − 1 , where a = cl . The Bloch-Floquet conditions (7) imply (T N − βI)F(1) = 0, (11) where β = e2iNlK . A non-trivial solution exists if det{T N − βI} = 0, which is equivalent to β2 − tr(T N)β + 1 = 0. (12) When |tr(T N)| > 2, the waves do not propagate through the layered structure. 3.3. PERTURBED STRUCTURE Perturbation of porosity across the stratified structure leads to increase of the width of the macro-cell. Analysis of the trace of the transmission matrix of this macro-cell enables us to determine the frequencies of non-propagating waves. LOCALISED DEFECT MODES AND A MACRO-CELL ANALYSIS FOR 335 Let the transmission matrices τ j of the cells, included in the large macro-cell, be defined by τ j = T + A j, where the matrices A j have small entries. For the macro-cell, the transmission matrix is given by N∏ j=1 τ j = (T + A1)(T + A2) . . . (T + AN)) (13) = T N + DN + . . . , where DN = N−1∑ i=0 T N AN−iT N−i−1. Thus, tr( N∏ j=1 τ j) = tr(T N) + tr(DN) + . . . . (14) Even if the entries of the matrices A j are small, for sufficiently large number 2N of layers within the macro-cell, the trace of the transmission matrix may change by a finite increment, which inevitably brings changes in the transmission diagrams. This is illustrated in the next section which presents model numerical simulations. 3.4. NUMERICAL EXAMPLE In this example 16 layers are used for the numerical simulation, with � = 0.3. Fig. 7 includes the dispersion diagram for the perturbed structure and the transmission diagram for the macro-cell corresponding to this perturbation of porosity across the stratified medium. In Fig. 7a extra band gaps occur in the high frequency region. The transmission diagram in Fig. 7b shows that the values of ω, for which (tr(T N))2 > 4, correspond to the stop band shown in Fig. 7a. Comparing Fig. 7b with the dispersion diagram in Fig. 7a we can say that the solution of the transmission problem for a finite width macro-cell describes accurately the width and position of the band gaps associated with the spectral problem for an infinite periodic structure. 4. Concluding remarks We have studied the localisation of eigenfields within lattice structures with de- fects and the connection between transmission problems for finite stacks and spec- tral problems for infinite periodic structures. The analytical methods described in 336 MOVCHAN, HAQ, MOVCHAN −4 −3 −2 −1 0 1 2 3 4 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 K 0 w (tr(Π j=1 j N τ )) 2 0 1 2 3 4 0.5 1 1.5 2 ω (a) (b) Figure 7. (a) Dispersion diagram for the perturbed structure with defects; (b) transmission diagram for the macro-cell; here a=1, l=1, m(1) = 2, m(2) = 1, � = 0.3 the paper enabled us to obtain analytical estimates of position and width of stop bands as well as frequencies of localised modes. The ideas presented in this paper can be extended to other types of lattices in two and three dimensions. Acknowledgement S. Haq acknowledges the financial support of the Ministry of Education of the Government of Pakistan through the merit scholarship. References Balk A., Cherkaev A., Slepyan L. (2001a) Dynamics of chain with non-monotone stress-strain relation. 1. Model and numerical experiments, J. Phys. Mech. Solids 49 131-148. Balk A., Cherkaev A., Slepyan L. (2001b) Dynamics of solids with non-monotone stress-strain relation. 2. Nonlinear waves and waves of phase transition, J. Mech. Phys. Solids 49 149-171. Felbacq D., Guizal B., Zolla F. (1998) Limit analysis of the diffraction of a plane wave by a one- dimensional periodic medium, J. Math. Physics Vol. 39 4604-4607. Lekner J. (1994) Light in periodically stratified media, J. Opt. Soc. America. A 11 2892-2899. Martinsson P. G., Movchan A. B. (2003) Vibrations of lattice structures and phononic band gaps, Quart. J. Mech. Appl. Math. 56 45-64. Movchan A. B., Movchan N. V., Poulton C. G. (2002) Asymptotic Models of Fields in Dilute and Densely Packed Composites, Imperial College Press. Zalipaev V. V., Movchan A. B., Poulton C. G., McPhedran R. C. (2002) Elastic waves and homogenization in oblique periodic structures, Proc. R. Soc. Lond. A 458 1887-1912. 337 337–344. ACOUSTIC WAVE PROPAGATION THROUGH DISCRETELY GRADED MATERIALS Abhijit Mukherjee and Ritesh Samadhiya Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai 400076, India Abstract. Functionally graded materials (FGMs) have great potential as energy absorbing devices, electrical transducers, thermal barriers, etc. Study of the propagation of elastic waves is imperative for a number of such applications. In this paper, a simple one-dimensional model is proposed to study the stress waves in discretely graded media. The model uses spectral approach to determine the stresses due to the incident and reflected waves in FGMs. It has been observed that FGMs attenuate and delay the peak stresses considerably as compared to composites with sharp interfaces. Key words: functionally graded materials; wave mechanics; spectral method; peak stresses; time delay 1. Introduction Grading of constituent materials of a composite for developing a new material with contradicting performance requirements has shown considerable promise (Bruck, 2000) and (Joshi et al., 2003). Such composites are called functionally graded materials (FGM). Thermo-electro-mechanical characterization of these materials is imperative for proper exploitation of the superior properties offered by them. (Joshi et al., 2003) has proposed a simple one dimensional model for the analysis of stress waves in discretely layered FGMs. In this paper, a simple one dimensional model is formulated for analyzing the stresses due to reflected waves in discretely layered FGMs. The reflected stress peaks are determined using the spectral approach (Doyle, 1997). The spectral method is discussed later in the paper. 2. Modelling of discrete FGM FGMs can be modelled as shown in Fig. 1. It consists of discrete layers of gradu- ally varying material property. The left side of the FGM consists of base material-1 and on the right side of it is base material-2. In between these two materials, Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 338 MUKHERJEE, SAMADHIYA Figure 1. Configuration of discretely graded FGMs showing power law variation discrete layers of intermediate material properties are present. The physical prop- erties of these inner layers are assumed to follow linear rule of mixtures. Lin- ear, elastic, longitudinal waves that are propagating in one direction only are considered. 3. Spectral analysis In the spectral method, the excitation is transformed from the temporal form to the spectral form by Fast Fourier Transform (FFT) algorithm. The governing partial differential wave equation is reduced to a set of ordinary differential equations. Their solution is easier than the original differential equation. However, often approximate solutions are sought. These solutions of the governing equations are used as shape functions for spectral element formulation. The advantage of this method is that the inertial effects are exactly represented and hence often exact solutions are obtained. In addition, often the resulting element is super convergent and very few elements are required to model the system. 3.1. RECONSTRUCTION OF WAVES The significance of the spectral approach to waves along with differential equation is that once the signal is characterized at one position in space then it is known at all positions, and therefore, propagating it becomes a fairly simple matter. The solution obtained by the spectral method is in frequency domain and to get back to the time domain inverse FFT is done. It should be observed that while using FFT for inversion the transformed solution in frequency domain is calculated upto Nyquist frequency and the rest is obtained by imposing the condition that it must be the complex conjugate of the initial part. This is done to ensure that the reconstructed history is real. ACOUSTIC WAVE PROPAGATION 339 3.2. SPECTRAL APPROACH The governing wave motion equation is given by, ∂ ∂x { EA ∂u ∂x } = ρA ∂2u ∂t2 (1) where, E is the elastic modulus, ρ is the mass density and A is the cross-sectional area of the rod. Considering both the modulus and the area do not vary with posi- tion, the displacement solution to the above equation for rod of length L, according to spectral approach (Doyle, 1997), can be represented by, u(x, t) = ∑ n ûn(x, ωn)e iωnt (2) where ωn represents the frequency and the spectral displacements ûn has the following solution, ûn = Ae−ikn x + Beikn(L−x) (3) where k is the wave number and A and B are undetermined amplitudes at each fre- quency, determined by applying boundary conditions. Hence, the displacements in the two base materials for sharp interface (with no inner layer present between the two base materials) can be written as, û1 = A1e−ik1 x + B1eik1 x (4) û2 = A1e−ik2 x (5) Here, û1 and û2 are displacements in respective mediums, k is the correspond- ing wave number, and A1, B1 and A2 are parameters determined by applying appropriate boundary conditions. Applying the boundary conditions at the interfaces (x = 0 and x = d), the motions can be expressed in matrix form as, −1 1 1 0 p1 p2 −p2 0 0 e−ik2d/n eik2d/n −e−ik3d/n 0 −p2e−ik2d/n p2eik2d/n p3e−ik3d/n B1 B2 B3 B4 = 1 p1 0 0 (6) where, pu = (kEA)u. Hence, the transformed solution in frequency domain can now be obtained in terms of the incident wave. To get back in time domain inverse FFT is required. If the FGM consists of m inner layers between the two base materials then all the 340 MUKHERJEE, SAMADHIYA parameters can be obtained in a similar manner. The assembled matrix then will be a square matrix of size 2m + 2. Hence, it is possible to determine the reflected and transmitted stresses in each of the inner layers. Several cases that explain the utility of the proposed model are discussed in the next section. 4. Numerical results The proposed model is applied to a number of problems with variable material grading. For all the cases the same triangular sharp impulse is applied on base material 1. The thickness (d) of the graded layer is 2mm in all cases. The reflected stress is measured at the interface (x = 0) of base material 1 and the first graded layer (Fig. 1). The composite discussed here is Alumina-Aluminum FGM. This combination is of special interest due to its energy absorbing capability. The prop- erties of the FGM are in Table 1. The stress time history of reflected waves has been plotted. The absolute stresses have been presented to enable comparison of magnitude only. TABLE I. Physical properties for Alumina-Aluminum Property Alumina (material 1) Alumina (material 2) Young’s modulus, (GPa) 400 70 Density, ( kg/m3) 4000 2700 Acoustic impedance (Kg/m2/s) 1265 435 Discrete layered FGM consist of many layers, whose physical properties are intermediate of that of the base material properties. These inner layers graded according to the power law. For power law variation the volume fraction of the base materials 1 and 2 at x (Fig. 4) can be written as, V2i = ( xi D )n (7) V1i = 1 − V2i (8) where Vi is the volume fraction of the respective base materials in the ith layer, xi is the coordinate of the center of the layer and n is an exponent that can varied as required. Case I: Sharp Interface In sharp interface, there is no inner layer between the two base materials. Eq. (6) can be used for determining stresses in the reflected and the transmitted waves. ACOUSTIC WAVE PROPAGATION 341 In this case a single peak is observed due to the reflection from the interface of the two base materials. The magnitude of the reflected stress is about half of that of the incident stress. Rest of the energy is transmitted to base material 2. Case II: Power law gradation The sharp interface is eliminated by graded layer(s) between the two base materials. The variation of material properties in the inner layer(s) is assumed to follow the power law. Fig. 2(a) shows the reflected stress time history for one inner layer for linear variation. Only three significant peaks are observed since after four reflections the magnitude of the reflected stress is reduced to almost 1% of the incident stress. The peak reflected stress is also reduced to about 25% of the incident stress wave. The time taken to attain peak stress is delayed too. The peaks are labelled to show the number of reflections coming back to base material 1. Fig. 2(b) shows the stress time history for 2 inner layers. It is noticed that as the number of inner layers increases the peak reflected stress goes down. This is expected since the number of reflections increases with the increase in the number of inner layers. Figure 2a. Reflected stress time history for one inner layer (linear variation) Figure 2b. Reflected stress time history for two inner layer (linear variation) Other power law variations are examined for their energy absorbing potential. Fig. 3(a) depicts the peak reflected stresses for different power law variations with the varying number of inner layers. It can be observed that with the increase in the number of inner layers the peak stresses come down for all the cases. The acoustic impedance ratios for the material systems can be studied to understand the phenomenon. Fig. 3(b) shows the acoustic impedance ratios for varying num- ber of inner layers for n = 2. The impedance ratios increases, reaches near to 1 and slopes of the curves reduce with the increase in the number of inner layers. As the ratio comes near to 1, the reflected stress reduces as clear form eq. (6). Ratio near to 1 shows gradual changes in the material properties. The reflected 342 MUKHERJEE, SAMADHIYA stress increases in case the ratio vary away from 1. It may be either greater or lesser than 1. It can be explained from the material properties point of view. Ratio near to 1 depicts that the material properties changes very slowly and hence all the waves will transmit, since waves will experience almost the same material when travelling from one layer to another. However, manufacturing constraints may limit the number of inner layers. However, manufacturing constraints may limit the number of inner layers. Therefore, a judicious design decision can be arrived at by choosing the number of inner layers depending on the target reduction of peak stress. When we compare the peak stresses for different exponents (n) for the same number of inner layers it is clear that n = 0.5 gives the least peak stress. The reflected peak stress was only about 11% of the applied peak stress in case of four inner layers. We shall once again study the acoustic impedance ratios to analyze these results. Fig. 2(c) shows the variation of impedance ratios with varying n for four inner layers FGM. We observe that in case of n = 0.5 the slopes of the curve are the lowest. In other words, the impedances in the case of n = 0.5 have varied the least. The increased variations in the impact ratios also increase the variation in stress peaks. It may be noted that n = 0.5 results in a quadratic curve and its second derivative is constant (in the present case in a discretized sense). Therefore, the grading that has uniform, or most uniform, rate of change of properties will result in the lowest peak stresses. Figure 3a. Peak reflected stress for power law variation Figure 3b. Impedance ratios for n = 2 ACOUSTIC WAVE PROPAGATION 343 Figure 3c. Impedance ratio for four inner layers Figure 4. Time delay for power law varia- tions The time delay histograms are plotted for these gradations in Fig. 4. Intuition suggests that the grading that has the maximum slope away from the point of measurement will have the highest delay times. That means the delay time should increase with the exponent n. However, the results do not follow this conjecture. For higher values of n the reflected waves travel through a stiffer medium resulting in higher velocities. As a result, the delay times do not follow the expected path and for exponents higher than 2 the delay times reduce marginally. Exponent of 2 has given the highest delay time. However, the delay time for n = 0.5 was not much lower. Therefore, considering the peak stress benefit, n = 0.5 seems to be the ideal grading for this system. 5. Concluding remarks A simple one-dimensional model based on the spectral approach has been pro- posed for material characterization and managing stress waves in discrete FGMs. The model is accurate and does not lack generality. Following conclusions can be drawn from the present investigation: 1. Stress wave magnitudes attenuate very fast in the graded media as compared to the sharp interface structure. 2. Graded media also delays the peak stress arrival, which should delay the time of damage initiation. 3. The stress time histories are highly sensitive to the type of grading and base material properties. 4. An FGM with power law grading exponent of 0.5 is best suited for the reduc- tion of peak reflected stress. 344 MUKHERJEE, SAMADHIYA 5. The best time delay is achieved for the exponent of 2.0. 6. Overall the performance of n = 0.5 was found to be most satisfactory. References Bruck, H. A. (2000) A one-dimensional model for designing functionally graded materials to manage stress waves, International Journal of solids and Structures 37, 6383-6395. Doyle, J. F. (1997) Wave propagation in structures, 2nd ed., Springer, New York. Joshi, S., Mukherjee, A., Schmaudar, S. (2003) Numerical characterization of functionally graded active materials under electrical and thermal fields, Smart Materials and Structures 12 571-579. 345 345–356. USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY Rainer Nordmann TU-Darmstadt Mechatronik und Maschinenakustik, Germany Abstract. Mechatronic components are getting more and more common in mechanical systems. As an example Active Magnetic Bearings (AMB) are often used in Rotating Machinery. Besides their function of an oil-, contact- and frictionless levitation of the rotor, they are best suited to be used as an exciter and measurement instrument to extract more information from the system under observation. In this paper it is shown, how Active Magnetic Bearings can be used for identification, diagnosis and optimization purposes. Key words: mechatronics, rotating machinery, diagnosis, optimization 1. Introduction In the field of rotating machinery the number of applications using mechatronic components is increasing. In comparison to conventional systems such mecha- tronic products consisting of mechanical, electrical and electronical components have the ability to pick up changes in their environment by sensors and react to the system or process by means of actuators after an appropriate information processing, carried out in a microprocessor (Aenis, 2002). Nowadays, rotors running with active magnetic bearings or with other mecha- tronic bearings or components already offer a variety of advantages. Some of them are the tuning possibilities for stiffness and damping, the absence of friction and wear, the high running speeds, the vibration isolation, the active vibration damp- ing and possible unbalance compensation. However there is much more potential in such systems with respect to a smart behaviour. In rotating machinery with mechatronic components, consisting of built in control, sensors, microproces- sors, actuators and last but not least built in software different novel features like identification, diagnosis and correction can be realized. In this way it is possible to design new machines with higher performance, higher reliability and longer lifetime. This paper particularly describes rotating machines with mechatronic compo- nents and concentrates on the mentioned smart features of identification, process diagnosis and correction and process optimization as well. After a short introduc- tion to Mechatronics, technical applications of rotating systems with mechatronic components will be presented. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 346 NORDMANN Figure 1. Block diagram of a Mechatronic system 2. What is mechatronics Mechatronics is a multidisciplinary area of engineering including mechanical and electrical engineering together with information technology. In mechatronic sys- tems signals are measured by sensors and processed in embedded microcontrollers. Actuator forces and/or moments then act on the process, controlled by the micro- processor functions. In Figure 1 the block diagram shows the different components of a mecha- tronic system (microprocessor, actuator, process and sensor) and their connections for the flow of energy and information. However, a diagram like this can also be applied for classical mechanical-electrical control systems. In comparison modern mechatronic systems have much more potential. They are characterized by two types of integration: the hardware integration and the integration of functions. The hardware integration or integration of components results from designing the mechatronic system as an overall system embedding the sensors, actuators and microcontrollers into the mechanical process. This spatial integration may be limited to the process and sensor or the process and actuator. The microcomputers can be integrated with the actuator, the process or sensor, or be arranged at several places. Integration by information processing (software integration) is mostly based on advanced control functions. Besides a basic feedforward and feedback con- trol, an additional influence may take place through the process knowledge and corresponding on-line information processing. This includes the solution of tasks like supervision with fault diagnosis, identification, correction and process opti- mization. In this paper applications of such integrated functions will be presented especially for rotating machinery. 3. Rotating machinery: Configuration and technical applications In various technical areas rotating machinery systems are in operation, like tur- bines, pumps, compressors, motors and generators etc. An example of a steam USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY 347 Figure 2. Configuration of Mechanical Components in Rotating Machinery turbine rotor is shown in Figure 2 with its main mechanical components. The ro- tating shaft with the blading system (energy transfer from thermal to mechanical) is supported in oil film bearings. The rotating parts are arranged in the turbine housing and sealed to the environment by means of labyrinths. Users expect, that their machines are running safe and reliable and have a high efficiency and availability as well. In order to satisfy these requirements mechanical problems especially vibrations have to be considered very careful. Particularly the dynamic behaviour of the rotating components and their different interactions with the stator are of great importance for the durability and lifetime of a machine. Therefore the machine designer already starts in an early stage of the development to investigate and to predict the machines dynamic behaviour and the corresponding internal forces and stresses, respectively. Figure 3 shows as another example a blower system consisting of the rotating shaft supported in two roller bearings. The impeller in the center of the shaft 348 NORDMANN Figure 3. Blower System with model delivers a specified gas flow for the process in the plant. In order to investigate the vibration problem of forced vibrations including the resonance problem a relatively simple model can be used with a flexible shaft, rigid bearings and the impeller with its inertia characteristics. The natural frequency f (Hz) of the first bending of the rotor system can easily be calculated by f = 1 2π ω = 1 2π √ c m (1) With this natural frequency f or the circular natural frequency ω = 2π f the critical angular of velocity is known. c is the shaft stiffness in N/m and m is the mass of the impeller in kg. Due to the fact, that the center of rotation is not equal to the center of gravity a mass eccentricity e has to be considered. This leads to an excitation by unbalance forces,with the rotational angular of velocity Ω. The resulting forced unbalance vibration x̂(Ω) e = mΩ2√ (c − mΩ2)2 + (dΩ)2 (2) is shown in Figure 4. USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY 349 Figure 4. Run up curve for blower The diagram shows the relative amplitude of vibration x̂(related to the mass eccentricity) versus the running speed (related to the natural frequency ω). The different curves belong to different damping values. It can clearly be seen, that the resonance effect occurs (critical speed), when the rotational frequency is equal to the natural frequency (Ω = ω). Damping (passive or active) helps to decrease the vibrations in the resonance. The example of Figure 3 is relatively easy to solve. More general rotordynamic tools for computer simulations are available nowadays, usually based on the Finite Element method. These routines allow to include all important components like shafts, impellers, bearings, seals etc. and take into consideration the corresponding effects like inertia, damping, stiffness, gyroscopics, unbalance and fluid structure interaction forces. They predict modal parameters like natural frequencies, damping values, mode shapes and unbal- ance and transient vibrations as well. While these powerful tools itself usually work without difficulties, problems more often occur in finding the correct input data. Particularly not all of the physical parameters are available from theoretical derivations. This is especially true for the rotordynamic coefficients, describing various fluid structure interactions. In such cases the required data have to be taken 350 NORDMANN from former experience or have to be determined experimentally via identification procedures. The identification of the dynamic characteristics of a rotating system by means of mechatronic components (Active Magnetic Bearings) offers further possibilities besides identification. The developed procedures are an important base for failure detection and failure diagnosis during operation. By using the ac- tuators as elements to introduce static and dynamic forces into the system process optimization and correction can also be performed. This will be shown in the next section. 4. Use of mechatronic components in rotating machinery Identification techniques have already been used in different applications in or- der to find rotordynamic coefficients (stiffness, damping, inertia) e.g. in bearings and seals. One of the main problems to work with identification techniques in rotordynamics is the excitation of a rotating structure during operation. On the one side it is not easy to have access to the rotor and on the other side the force measurement is difficult, especially when a machine is running with full power and speed and the signal to noise ratio is bad. In some recent investigations Active Magnetic Bearings (AMB’s) have been used in order to solve this difficult task. These new techniques seem to be very promising, because AMB’s do not only support the rotor, but act as excitation and force measurement equipment as well. In cases, where active magnetic bearings arc designed as bearing elements for turbomachinery systems, it seems helpful to use them also as excitation and force measurement tool. In such applications identification of the dynamic behaviour of the rotating machinery system would be possible during normal operation. 4.1. PRINCIPLE OF ACTIVE MAGNETIC BEARINGS Unlike conventional bearing systems, a rotor in magnetic bearings is carried by a magnetic field. This means, that sensors and controllers are necessary to stabilize the unstable suspense state of the rotor. Therefore, essential dynamic characteris- tics like stiffness and damping properties of the whole system can be influenced by the controller (DSP-Board). For the following applications the rotor can be moved on almost arbitrarily chosen trajectories independently of the rotation, e.g. harmonic motions in one plane, forward or backward whirls. Additionally, an im- balance compensation can be performed. A magnetic bearing system consists of four basic components: magnetic actuator, controller, power amplifier, and shaft position sensor. To keep the rotor in the bearing center, the position sensor signal is used as input for a control circuit to adequately adjust the coil currents. The bearing configuration is composed of four horseshoe- shaped magnets (8 magnetic poles) and is operated in the so-called differential driving mode, where one magnet USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY 351 Figure 5. Principle of an Magnetic Bearing is driven with the sum of bias current and control current, and the other one with the difference (Figure 5). 4.2. FAILURE DIAGNOSIS OF A CENTRIFUGAL PUMP Today, monitoring systems are normally not an integral component of turboma- chines. With these failure detection systems, the relative and/or absolute motions of the rotor are measured as output signals. After signal processing, certain fea- tures (threshold values, orbits, frequency spectra etc.) are created from the mea- sured data. With the deviations of these features from a faultless initial state, the diagnosis attempts to recognize possible faults. The difficulty with these proce- dures is that the causes of the modifications of the output signals can not be detected clearly. The reason might either be a change of the process or a modifica- tion of the system itself. An improvement of the existing diagnostic techniques can 352 NORDMANN Figure 6. Centrifugal Pump with Magnetic Bearings be achieved by using AMBs. They are well suited to operate contactless as actua- tor and sensor elements in rotating machinery. Consequently, frequency response functions (stiffness or compliance frequency responses) can be determined from the measured input and output signals, of which the physical parameters or modal parameters (natural frequencies, eigenmodes, modal damping) of the system can be identified. A magnetically suspended centrifugal pump is used to validate and to demonstrate the performance of the developed model based diagnosis methods. Figure 6 shows the single stage pump with the pipe system and the driving mo- tor. Two radial and axial magnetic bearings support the pump shaft and serve in addition as actuators and sensors for the determination of the frequency response functions (compliance functions). As an example, Figure 7 represents a measured frequency response function (reference) without any failure. If a failure occurs, e.g. dry run of the pump system, the frequency response function will change its behaviour, e.g. to the curve Dry Run (measured). Due to the dry run the fluid structure interactions in narrow seals are no longer working and this changes the pump dynamic behaviour completely (stiffness and damping coefficients of the fluid). From the frequency response USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY 353 Figure 7. Frequency Response functions for the centrifugal pump change it can be concluded that the failure dry run has occurred. The task is to find out, when a fault occurs and to determine the faults type, location and extend. A model of the pump system can help to solve this problem (model based diagnosis). In Figure 8 the simulation for the dry run system leads to another curve, which fits very well to the measurements of the faulty rotor. The corresponding pump parameters for this simulation belong to the dry run situation. 4.3. DIAGNOSIS AND PROCESS OPTIMIZATION OF A HIGH SPEED GRINDING PROCESS Internal grinding is applied on products such as outer rings of ball bearings or injection parts of combustion engines. The requirements for this process are very high and contradictory. On the one hand, very high shape and size accuracies as well as surface quality of the work-pieces are demanded. On the other hand short process cycles are wanted due to the mass production. The mean measures which can be taken to meet these requirements are the improvement of the bearing system, the process monitoring and diagnosis as well as the process correction with regard of the process optimization. High speed grinding spindles in AMBs are best suited in order to carry out these measures (Figure 8). 354 NORDMANN Figure 8. Grinding Spindle with Magnetic Bearings Figure 9. Process diagnosis with normal grinding force Regarding the magnetic bearing system, they allow high rotational speeds up to 180.000 rpm leading to a decreasing normal process force. The not existing mechanical friction within the bearings permits a long bearing life. Furthermore a high static spindle stiffness can be reached by an appropriate controller design. Using the measured displacement signals and the given control currents supplied to the AMB system the AMB-forces are obtained and out of it the normal, tan- gential and axial process forces respectively. The normal process force is the essential quantity for the process diagnosis since it permits a direct conclusion to the process state (Figure 9). Based on the measured and calculated quantities the diagnosis algorithms are carried out to evaluate the process state, for instants with regard of chatter or a USE OF MECHATRONIC COMPONENTS IN ROTATING MACHINERY 355 Figure 10. Static tool bending compensation broken grinding wheel. Out of the diagnosis procedure or directly out of the sig- nals correction procedures take place in order to optimize the grinding process. In this phase the AMBs are used as actuators to move the spindle to a required position. One correction procedure is shown in Figure 10. Due to its flexible property the work- tool is bent during grinding leading to an undesirable conical bore shape. By tilting the spindle within the bearing this effect does not take place. The described concept is developed and tested on a AMB high speed grinding spindle test rig. Another process optimization is possible by oscillation of the spindle in axial direction with a frequency of about 30 Hz and small amplitudes of about 20 µm. This motion, generated by the axial magnetic bearing, leads to an improvement of the workpiece surface. 5. Conclusions The use of mechatronic components in rotating machinery may contribute to a better performance of this type of machines. As an example Active Magnetic Bearings can also be used as sensor, actuator and exciter. Accurate measurements of forces and displacements and the generation of axial and radial shaft motions are possible. In this way AMBs offer new applications in turbomachinery, e.g. 356 NORDMANN improved identification, diagnosis and optimization techniques. This technology supports the development of new products in the field of rotating systems. Acknowledgements The author wishes to thank Mrs. Birgit Lampert and Mr. Alexander Spiess for their support to complete this paper. References Aenis, M. (2002) Einsatz aktiver Magnetlager zur modellbasierten Fehlerdiagnose in einer Kreiselpumpe, Ph.D. thesis, TU-Darmstadt. 357 357–369. COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS ∗ Miloslav Okrouhlı́k Institute of Thermomechanics, Prague, The Czech Republic Abstract. The contribution deals with artifacts that might occur when the finite element (FE) analysis is applied to transient problems of continuum solid mechanics. The side-effect phenomena – in many cases unimportant and easily neglected might play overwhelming role in others – will be described and explained with intention to show beneficial effects of analytical, experimental and computational synergy. When trying to ascertain the precision, threshold limits, reliability of modeling approaches and the extent of their validity, one has to realize that the models we are using have almost no self-correcting features. That’s why we have to let the models check themselves, be checked by independent models and let the systematic doubt be our everyday companion. Key words: finite element method, continuum vs. discretized medium, dispersion, threshold, model 1. Introduction It is all about proper modeling. Generally, a model tries to predict nature, it describes what would happen if the studied phenomena were governed by cer- tain simplifying assumptions with a pious hope that the neglected features in the model would not lead to results being too different from those observed in nature. The accepted assumptions, however, determine the validity range of the model. Using the model outside the scope of validity will necessarily lead to wrong or questionable results. This, however, is not the deficiency of the model but the lack of user’s judgment. Every model has to be tested by an experiment or by another model to deter- mine the scope of its validity. The theory (model) is considered to be ’correct’ if it agrees with results of experimental observation. If the scope of validity of one theory is greater than that of another, people have tendency to say that the former is better than the latter. The more sophisticated theory, however, usually requires more input data, and greater effort. Using it in cases where the simple theory would give acceptable results is just an unjustifiable luxury. The paper ponders about cases on the borders of validity of models and in the vicinity of observational and computational thresholds. ∗ with thanks to my teachers R. Brepta, C. Höschl and credits to my colleagues S. Pták, F. Valeš and J. Trnka. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 358 OKROUHLIK 2. Problem dependent benchmarks Benchmarks are useful tools for assessment of validity, precision and robustness of tested methods. For FE analysis either analytical or experimental data could serve as useful means for comparison. We will show how carefully the benchmark results have to be evaluated. 2.1. LAMB’S PROBLEM Horace Lamb (1849 - 1934) derived analytical solution for a semi-infinite half- space loaded by a point force applied perpendicularly to its free surface. The time distribution of the loading was assumed in the form of Heaviside function. See (Love, 1944). For computational purposes the half-space was modeled by a cylinder, whose symmetrical part is sketched in Fig. 1a. The time distribution of the loading point force - applied at point A - is prescribed by a rectangular pulse, whose duration corresponds to time when the primary longitudinal wave reaches half the length of the specimen. Until the primary wave reaches the surface of the cylinder and/or the face opposite to the loaded one, the considered cylinder is a proper representation of the half-space and the rectangular pulse represents the Heaviside function. The original Lamb’s analysis employs Fourier superposition of harmonic waves for the transient normal loading on a half-space. Later Cagniard presented a nu- merical method for the inversion of the Fourier’s transform, and Brepta, Valeš (Vale¢s and Brepta, 1972) and others were able to carry out the inversion ana- lytically even for cases with more complicated boundaries. Solving the Lamb’s problem on a free boundary is relatively simple and is available in a closed form giving space-time distributions of radial and axial displacements. It is attributed to Pekeris, whose analytical formulae, also presented in (Graff, 1975), are easy to evaluate. Computed results, depicted in space-time coordinates in Fig. 1b, show two significant singularities as well as arrivals of longitudinal, shear and Rayleigh waves. The available analytical solutions are as a rule based on a point force loading whose time dependencies are usually prescribed by Heaviside or Dirac functions. This fact partially simplifies the dreary workload needed for carrying out the an- alytical solutions. The finite element results of the Lamb’s problem, demonstrate ’divergence’ behaviour as a function of increasing mesh density. This is docu- mented in Fig. 1c, where the time dependence of displacements directly under the loading is depicted. The finer meshes yield gradually greater displacements under the loading force. One can conclude that the finite element method, being an orthodox daughter of continuum mechanics, would give infinite displacements for infinitely fine mesh. COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 359 Figure 1c also shows the influence of the element type used for finite element computation. One can also see that the double mesh density of bilinear elements (L) produces roughly the same results as the reference density with biquadratic elements (Q). The considered meshes are 20 by 20, 40 by 40 and 80 by 80, respectively for the considered space domain. Bilinear and biquadratic square axisymmetric elements, full quadrature, consistent mass matrix formulation and Newmark time step operator without algorithmic damping were systematically used. In continuum, the effects of different localized equivalent loads cannot be dis- tinguished in areas located sufficiently far from their application. This is what St. Venant’s principle states. In discretized continuum, however, the effects of point and equivalent distributed loadings are different and furthermore mesh dependent. This is manifested by a paradoxical increase of stored potential and kinetic ener- gies with the increasing mesh density, which are plotted in Fig. 2 as a function of time. One is almost tempted to employ this phenomenon for a pollution-free en- ergy production. The pollution-free energy production problem disappears when an equivalent distributed (non-point) loading is applied. See Fig. 1d showing that the response to an equivalent pressure loading is almost independent of meshsize and element type. 2.2. THICK OR THIN ROD In Fig. 3a there a cylindrical rod with a spiral groove modelled by 3D elements. Fig. 3b shows the energy balance in the rod being subjected to the axial impact loading by a ‘short’ (denoted shopu) and ‘long’ pulse (denoted lopu) respectively. Kinetic, potential and total energies (denoted Ekin, Epot, Esum) are plotted in dimensionless form as functions of time. The ratio of total energies is 4:1. This corresponds to the subsequent loading of the rod by two strikers (the length of the long striker is four times that of a short one, of the same material and diameter as the rod) with identical initial velocities. The old wisdom is well documented here. Namely the fact that the notion of thinness or thickness of the rod (in other words its 1D or 3D behaviour) does not depend on the ratio of the rod’s length to its diameter, but on the frequency spectrum of the loading. One can see that the rod, being loaded by a short pulse (having higher contribution of high frequency components), behaves according to 3D theory while the same rod loaded more gently by a longer pulse manifests its 1D nature. See (Brepta et al., 1996). Details could be found in (Okrouhl¢k and Pt¢ak, 2003). 2.3. THICK OR THIN SHELL The Fig. 4 shows comparison of FE displacements, velocities and accelerations as functions of time at the location mp1 as obtained by different elements (shell and 360 OKROUHLIK Figure 1. Impact loading of elastic halfspace Figure 2. Pollution-free energy production a) Lamb problem – axisymmetric approach b) Pekeris solution for free surface c) Axial displacement at A – point loading d) Axial displacement at A – pressure loading COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 361 b) Energy comparison for short and long pulses Figure 3. Impact loading of elastic rod 362 OKROUHLIK solid) applied to two similar but slightly different shells - denoted A and D respec- tively. The shells are loaded by an impact force, perpendicularly to the surface; the impact is applied from inside. The location mp1 lies on the outside surface, on the line parallel to axial axis, not far from the loading. The left column shows results obtained for the A shell computed by coarse and fine shell (A1 and A2) and coarse solid (A1 hex) elements. The right column shows results obtained for the D shell computed by coarse shell (D1) and coarse and fine solid (D1 hex and D2 hex) elements. The results for both geometries are similar, since the location mp1 is close to the loading point and the differences in solutions for different geometries did not have time to evolve. There is not a visible difference between A1 and A2 shell results - the finer mesh does not ‘improve’ the solution. This is not a proof of correctness – it only means that for a given geometry and loading spectrum the meshsize and the timestep are set up correctly. In Fig. 4 there is a visible time shift in results obtained by shell and solid elements. This is attributed to the fact that solid and shell elements represent different models of continuum. The shell element knows nothing about wave processes within the shell thickness since the thickness is just a computational constant and the mechanical phenomena occurring there are a priori assumed. The solid element has, however, the ‘active’ dimension in the normal direction of the thickness and is able to capture the wave and energy processes in this direction. One should realize that i) a too small number of elements in this direction partially disqualifies the results (notice the high frequency contents in D2hex accelerations) due to dispersion side effects and that ii) stress waves carry a negli- gibly small amount of energy within the thickness. Thus in this case simple shell elements offer a seemingly more reasonable solution devoid of false high fre- quency components. On the other hand the finer and coarse mesh results for solid elements agree remarkably well confirming thus the consistency of the approach. As far as the time shift in results due to different element modelling is con- cerned, we believe that the solid model result is more reliable, since - regardless of false high frequency contents of the response - it better represents the more flexible and more complex nature of the shell. The visible time shift is attributed to the fact that the loading effect, applied from inside of the shell, needs of about 1 µs to pass through the shell thickness. Computational and other details can be found in (Okrouhl¢k and Pt¢ak, 2005). 2.4. FE SELF-CHECK CONSIDERATIONS The limits of FE model have to be carefully considered when FE approach is employed for the assessment of other methods. Due to the spatial dicretization process the FE model exhibits a dispersive character (Okrouhl¢k, 1994) and has only a countable number of degrees of freedom and eigenfrequencies. The COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 363 Figure 4. Shell and solid element compared frequency spectrum has the cut-off limit, which means that the FE model is not able to transmit frequencies above a certain limit. Internal forces acting between finite elements of FE model are limited to ad- jacent neighbours only. That’s why the stiffness and mass matrices are sparse in FE analysis. They could also be relatively narrow-banded if an efficient node or element numbering is secured. To find the response of a body in time due to the impact loading requires solving the system of differential equations in which the inversion of stiffness or mass matrices (depending on the mass matrix formulation and on the time integration method being employed) is needed. See (Bathe, 1996). This way, the sparse structure of relevant matrices might vanish, and a sort of unlimited range of influence (both in space and time) could prevail. If we worked with infinite number of significant digits we would get a nonzero response over the whole domain for any time greater than zero. Working, however, with standard precision arithmetics we will get no ‘measurable’ response in domains where the computed values are, in absolute value, less than the computational threshold - the smallest positive floating point number. Besides, the limited range of influence is secured by the fact that a substantial part of the energy is always carried by low frequency components. 364 OKROUHLIK Figure 5. Shell response to impact loading. Rendering threshold adjusted This is illustrated in Fig. 5 where a shell response in normalized normal dis- placements (normalized with respect to their maximum value) to impact loading, perpendicular to the surface - for a given time - is presented in the form of contour lines for a quarter of the shell loaded at the lower left-hand corner. Instead of observing computational threshold we are just adjusting the rendering one to show that besides of the actual, physically sound data, we get the area of nice looking garbage pattern, followed by another area where the response cannot be seen at all, since it is smaller that the rendering threshold. The previous discussion might appear rather academic. The threshold issue, however, is really important when the speed of propagation is to be determined by experimental means. Observing the first measurable response at a certain time in a given distance from the loading point one can estimate the propagation speed. This way the estimated value depends on the threshold value of an apparatus being used. The dependence of detected velocity of propagation on the threshold value could be simulated analyzing the FE data. In left-hand side of Fig. 6 there are plotted distributions of normal displace- ments (normalized with respect to maximum value) in axial and circumferential directions for a shell modelled by a relatively coarse-mesh shell elements, regis- tered at a certain time. The four smaller upper right-hand side subfigures depict the same situation, but showing more details by subsequently refining the scales of plotting. The speed of propagation can be evaluated by registering the most distant response, which - in absolute value - is greater than the simulated threshold. The detected values of velocity of propagation as a function of threshold are plotted in the lower right-hand side subfigure. One can see that for correct captur- ing of the longitudinal velocity the relative precision of at least of the order of 10−6 COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 365 Figure 6. Simulated threshold value and detected speed of propagation is required. This is a tough demand. The threshold of the order of 10−3 is more common in experimental practice, which might explain why the measured time of the arrival of the first measurable response leads to detected speed of propagation of the order of 3000 m/s. All this fuzz is about margins of our ability to distinguish something against nothing. This is, however, crucial for any meaningful human activity. There are two nice books about this subject by J.D. Barrow, namely The Book of Nothing and The Limits of Science and the Science of Limits. 3. Sensitivity thresholds Error assessment depends, among other things, on threshold values. Computa- tional threshold is related to the computer representation of real numbers (in purely mathematical sense) while that of experiment is substantially more com- plex depending on methods and instruments being employed for the experimental task. 366 OKROUHLIK 3.1. COMPUTATIONAL All computers designed from 1985 use so called IEEE floating point arithmetics which means that there is a machine independent standard for the treatment of floating point numbers. This means that the floating point numbers are expressed in the form x = ±(1 + f ) 2e , where f is normalized integer mantisa repre- sented by 52 bits and e is another integer within the interval −1022 ≤ e ≤ 1023 related to the number of bits reserved for exponents representation. It is the finiteness of exponent which limits the interval of real numbers that can be rep- resented by floating point numbers. The smallest floating-point number 2−1022 = 2.225073858507201 × 10−308 is the underflow limit and can be viewed as the computational threshold. The maximum floating point number, pointing to the overflow limit, is 21023 = 1.797693134862316 × 10308. These two limits should be distinguished from another important quantity as- sociated with representation of floating point numbers, namely the unity round-off error, also called machine epsilon, corresponding to the distance from 1.0 to the next larger floating point number. Its value is 2−52 = 2.220446049250313×10−16. From it follows that the number of decimal digits corresponding to 52 binary digits is approximately 16 – which is good old Fortran double precision format. 3.2. EXPERIMENTAL The experimental threshold for Double Pulse Holographic Interferometry (DPHI) method, (Trnka and Landa, 2002), – being used for evaluation of displacement field on the surface of the impacted shell – depends on the wavelength of ruby light being employed. Its determinations can be shown on the comparison of dis- placements which is in Fig. 3 where displacements for a shell geometry obtained by the experimental and FE analysis with a coarse mesh, corresponding to time 44.2µs after the beginning of impact, are presented. Experimental displacements are normal components of displacements as seen by a camera in central projec- tion. They are represented by black and white fringe pattern. The step of contour lines corresponds to the quarter-wave length of the ruby laser light used. The FE displacements, plotted as contour fields - with the same step, are those in normal direction (see Fig. 7) obtained by coarse-mesh shell elements. For more details see (Trnka and Dvo¢r¢akova, 2004). Comparing the maximum displacement for this case with the the quarter-wave length of the ruby laser light, i.e. 1.735e-7 m, of the Double Pulse Holographic Interferometry (DPHI) method used, gives the relative threshold is of the order of 2 percents. COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 367 Figure 7. Comparison of displacement fields 3.3. FE – EXPERIMENT SYNERGY OR HOW TO MAKE A DISAGREEMENT SMALLER This study was invoked by a disagreement of FE and experimental displacement fields which appeared due to practical impossibility to generate identical loading pulse by experiment. Observations of experimental and FE data comparisons evoked a more de- tailed numerical study of similarities and differences of the experimental loading pulses with the intention to find suitable measures allowing to judge the ‘mag- nitude’ and ‘quality’ of the loading pulse - at the moment it is created - and eventually to drop those that are ‘unsatisfactory’. It was found that the “quality” and “reproducibility” of loading pulses can be assessed by means of simple computational procedures applied to these pulses immediately after their application. It was shown that the first term of FFT, usually called c0, and the mean value of the pulse are in excellent correlation – see Fig. 8 – and proportional to the momentum of the loading. Furthermore, the Euclidean norm of the time history of the loading pulse is proportional to the square root of the total input energy. For more details see (Okrouhl¢k and Pt¢ak, 2005). 4. Conclusion It was shown how numerical methods could help ascertaining the reproducibil- ity and precision of experiment and how the experiment could provide data for 368 OKROUHLIK Figure 8. Mean value, c0 value, norm and square root of input energy making computational procedures reliable, efficient and robust. When trying to ascertain the reliability of modelling approaches and the extent of their validity one has to realize that the models as a rule do not have self-correction features. That’s why we have to let the models to check themselves, be checked by indepen- dent models and let the systematic doubt be our everyday companion. Conditions of consistency, Fourier’s analysis and observation of a few global measures - all of them available during the experiment - could be used for deleting certain exper- imental data making thus the loading process more reliable. We believe that the experimental and numerical synergy that has been pursued is needed for a proper establishment of precision of applicability limits of methods used for modelling nature. Acknowledgements Financial support from the Grant 1ET400760509 of Czech Academy of Sciences is greatly appreciated. COMPUTATIONAL LIMITS OF FE TRANSIENT ANALYSIS 369 References Bathe, K. J. (1996) Finite Element Procedures, Prentice-Hall, Englewood Cliffs. Brepta, R., Okrouhl¢k, M., and Vale¢s, F. (1985) Wave and Impact Processes in Solids and the Methods of Solution, Academia, Praha. Graff, K. (1975) Wave Motion in Elastic Solids, Clarendon Press, Cambridge. Love, A. (1944) A treatise on the Mathematical Theory of Elasticity, Dover Publications, New York. Okrouhl¢k, M. (1994) Mechanics of Contact Impact, Applied Mechanics Reviews 47, 34-99. Okrouhl¢k, M., and Pt¢ak, S. (2003) Numerical Modelling of Axially Impacted Rod with a Spiral Groove, Engineering Mechanics 10, 359-374. Okrouhl¢k, M., and Pt¢ak, S. (2005) Assessment of Experiment by Finite Element Analysis: Comparison, Self-check and Remedy, Strojnicky Casopis 56, 18-39. Trnka, J. and Landa, M. (2002) Double pulse holounterferometry and PZP transducers in the study of ultrasonic guided wave propagation in a thin czlindricel shell, Acta Technica CSAV 47, 87-97. Trnka, J., and Dvo¢r¢akova, P. (2004) A Contribution to Quantitative Interpretation of Double Pulse Holointerferograms of Thin Walled cylindrical Shells, In Proceedings of the Colloquium Dynamics of Machines, Institute of Thermomechanics, Prague. Vale¢s, F., and Brepta, R. (1972) State of Stress of Two-dimensional Configurations due to Nonsteady-state Shock Loading, In Proceedings of the 13th IUTAM Congress, Moscow. 371 371–376. THE EFFECT OF NON-LINEAR BEHAVIOR OF CONCRETE ON THE SEISMIC RESPONSE OF CONCRETE GRAVITY DAMS A. Vatani Oskouei1 and Orhan Aksoğan2 1Buildings and Housing Research Center , Islamic Azad University, P.O Box 13145- 1696, Tehran, Iran 2Department of Civil Engineering, Çukurova University, 01330 Adana, Turkey Abstract. During earthquakes tensile stresses in concrete dams may exceed the tensile strength of concrete. This study examines the earthquake response of gravity dams due to the non-linear behavior of concrete. For this purpose, Drucker-Prager criterion is used. The linear interaction between the dam and compressible water is included. In the finite element analysis of the two media the Lagrangian formulation was used. For numerical analysis, the earthquake response of Pine Flat gravity dam monolith has been studied. Linear and non-linear responses have been compared. A rigid foundation has been assumed for the dam and the effect of alluvium and sediments in the bottom reservoir on the response has been neglected. Key words: gravity dams, concrete non-linear behavior, dam-water interaction, earthquake analy- sis, Drucker-Prager criterion, Lagrangian approach 1. Introduction Many dams have failed throughout history. Although very few of these failures re- sulted from earthquakes, and none involved major concrete dams (Chopra, 1988). The real behavior of these important structures has not been evaluated completely yet. In some concrete dams subjected to strong ground motions, high tensile stresses have caused cracking, for example Koyna gravity dam in India in 1967, Hsing Feng Kiang buttress dam in China in 1962 and SefidRud buttress dam in Iran in 1990 (Hinks and Gosschalk, 1993). Although field evidence of damage is limited, Hall (1988) emphasizes that no large dam with full reservoir has been subjected to a large, long duration earthquake at a small epicenter distance. Earthquake-induced damage of concrete dams has been examined experimen- tally by small-scale models on model earthquake simulators. Although small-scale model tests present difficult issues of simulation, they have shown that a discrete cracking forms near the upper portion of upstream face and propagates to the downstream face, but that the dam remains stable after the earthquake (Hall, 1988 Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 372 OSKOUEI, AKSOĞAN and Niwa and Clough 1980). For treating the real behavior of concrete dams, a nonlinear analysis should be used. U.S. National Research Council report identi- fied the non-linear modeling of concrete as an important research need to assess the earthquake safety of dams. Vargas-Loli and Fenves (1989) have analyzed dy- namic response of Pine Flat dam, using a fracture mechanics method, namely crack band theory. 2. Non-linear model of concrete As mentioned above, in the present study the Drucker-Prager criterion, which is one of the plasticity methods, has been used to model concrete non-linear behavior. The Drucker-Prager (1952) yield criterion takes the form: f (I1, J2) = αI1 + √ J2 − κ = 0 (1) where α and κ are positive material constants, related to the internal angle of friction and cohesion of the material. I1 is the invariable of stress tensor and J2 is the invariable of deviator stress tensor. As described in equation 1, the yield surface, f = 0, in principal stress space is a right-circular cone with its axis equally inclined with respect to each of the coordinate axes and its apex in the tension octant. If it is assumed that during an earthquake vertical joints between forming blocks of dams, open and operate separately, it is better to use plane stress analysis, but if it is assumed that these joints do not open and dam remains solid, using plane strain analysis is more appropriate. In this study plane strain analysis has been chosen. 3. Dam-reservoir interaction modeling In the present study for the dynamic interaction of dam-reservoir, the Lagrangian method has been used. In this method, in both structure and fluid environment, the variable is the displacement, and the coefficient matrix of the system of equations is symmetric. As in the Lagrangian method the main goal is to simulate the fluid environ- ment relationships like the solid environment. Water is assumed as a non-viscous material with zero shear coefficients, zero values appear on the diagonal of the elasticity matrix. So during the calculation of the inverse matrix, division by zero terminates the computer program. To overcome this problem, non-rotation coef- ficient is used instead of shear coefficient. Many researchers have considered this coefficient to be 100 times the Bulk modulus of the fluid (Oskouri and Dumanoğlu, 1998, Wilson and Khalvati, 1983, Oskouri and Dumanoğlu, 2001). Wilson and Khalvati (1983) have assumed the rotation mode of fluid as a spring whose one end is rigid and the other end is connected to the Gaussian points of the element. NON-LINEAR BEHAVIOR OF CONCRETE 373 They also defined the non-rotation coefficient by spring stiffness. Then, since the fluid environment is non-rotational, rotation can be limited by a coefficient α. For two dimensional stress-strain of fluid environment can be written as follows: { P P̄x } = [ K 0 0 αx ] { �v �̄x } (2) where P is the pressure, �v is the volumetric strain, P̄ is the rotational stress, K is the Bulk modulus of fluid and �̄ is the rotational strain. Reduced integration method, offered by Wilson and Khalvati (1983), has been used for the numerical calculation of fluid element. 4. Dynamic analysis of pine flat dam 4.1. SPECIFICATIONS OF DAM AND FINITE ELEMENTS MESH The earthquake response of the tallest, non-overflow monolith of Pine Flat dam has been chosen for the analysis. The finite element model of the Pine Flat dam monolith and impounded water is shown in Figure 1. The finite element model of dam consists of 40 quadrilaterals, 8 node elements with 2 × 2 Gaussian points. The water consists of 45 quadrilateral, 9 nodes elements with 2 × 2 Gaussian points, and it extends upstream a distance of 348.36 m, which is three times the dam height. The horizontal (S00E) component of the 1940 El Centro earthquake acceleration has been used in the analysis. The first 6.5 seconds of these records has been considered. Figure 1. Specifications and finite elements mesh of Pine Flat dam 374 OSKOUEI, AKSOĞAN Figure 2. Time history of horizontal displacement 4.2. MATERIAL PROPERTIES The concrete has a unit weight of 2482.1 kg/m3, Poisson’s ratio of 0.2, and the modulus of elasticity is taken as 34475 MPa. Compressive strength of the concrete has been assumed as 26.6 MPa. The concrete tensile strength is assumed to be 15% of compressive strength (3.9 MPa). The water has the following properties: the unit weight, 1000 kg/m3, the Bulk modulus, 2070 MPa, and the rotational stiffness coefficient, 207 GPa (100 times the Bulk modulus). In the analysis, the damping ratio is assumed to be 5% of the fundamental frequency of system. 4.3. ANALYSIS PROCEDURES Linear and non-linear seismic responses of the dam are analyzed with full reser- voir. In the time-integration procedures, time steps of 0.002 and 0.001 seconds are used for the case of linear and non-linear analysis procedures in the dam, respectively. The Wilson- θ method is applied for integration and the modified Newton method is chosen in the non-linear analysis. Horizontal components of displacement crest, obtained from the linear and non-linear analysis, are com- pared (Fig. 2). Cracking pattern of the dam when maximum displacement at the crest go on to upstream and downstream occurs are obtained (Fig. 3). Maximum hydrodynamic pressures, which applied from reservoir to the upstream face of the dam, are plotted (Fig. 4). 5. Conclusions The results obtained from the seismic analysis of the Pine Flat concrete gravity dam, can be summarized as follows: (1) Concrete tensile cracking is an important non-linear factor in the behavior of dams. The responses obtained from the linear and non-linear analyses very similar before cracking, but afterwards they differ NON-LINEAR BEHAVIOR OF CONCRETE 375 Figure 3. Cracked element a-at time 2.498 when max. disp. Dam crest occurred at moving downstream b- at time 2.676 when max. disp. Dam occurred at moving upstream Figure 4. Maximum hydrodynamic pressures completely. (2) Reduction of dynamic non-linear response frequencies of the dam after cracking indicates that as a result of cracking in some elements, the dam stiffness decreases and dam displacement increases. (3) The critical section of the dam is the level of discontinuity in downstream face, because cracking initiates at this level from upstream and downstream faces, and these cracks join through the dam and lead to the separation of the upper portion. (a) (b) 499000 49891.8 376 OSKOUEI, AKSOĞAN References Chopra A. K. (1988) Earthquake Response Analysis of Concrete Dams, (Ed: Jansen R. B) Advanced Dam Engineering for Design, Construction, and Rehabilitation. Vol Van Nostrand Reinhold, New York. Drucker D. C., Prager W. (1952) Soil Mechanics and Plastic Analysis On Limit Design, Quart.Appl. Math 10. Hall J. F. (1988) The Dynamic and earthquake behavior of concrete dams, Soil Dyn. Earthquake Engng. 7 58-121. Hinks J. L., Gosschalk E. M. (1993) Dam and Earthquake-A Review, Dam Engineering IV(1). National Research Council (1990) Earthquake Engineering for Concrete Dams: Design, Perfor- mance, and Research Needs, National Academy Press. Niwa A., Clough R. W. (1980) Shaking table research on concrete dam models., UCB/EERC 80-5. Earthquake Engineering Research Center. University of California, Berkeley, CA. Oskouei V., Dumanoglu A. A. (1998) Effect of Cavitation on Dynamic Behavior of Concrete Gravity Dams, ICOLD, Dam Safety Editor, Berga, Spain. Oskouei V., Dumanoglu A. (2001) Nonlinear Dynamic Response of Concrete Gravity Dams: The Cavitation, Soil Dynamics and Earthquake Engineering 21. Vargas-Loli L. M., Fenves G. L. (1989) Effects of Concrete Cracking on the Earthquake Response of Gravity Dams, Earthquake Engineering and Structural Dynamics 18 575-592. Wilson E. L., Khalvati M. (1983) Finite Element for the Dynamic Analysis of Fluid-Solid System, International Journal for Numerical Methods in Engineering 19. 377 377–382. CALCULATION OF MICROWAVE PLASMA OSCILLATIONS IN HIGH TEMPERATURE SUPERCONDUCTORS Zeynep Güven Özdemir,1 Özden Aslan,2 and Ülker Onbaşlı2 1 ˙ 2Physics, Marmara University, İstanbul, Turkey Abstract. It has been determined that the superconducting system, that consists of copper oxide based mercury oxide, behaves as if an oscillating cavity with the plasma frequency of 1012 Hz under particular amount of magnetic field at space temperature of 100 K. Moreover, plasma frequency of superconducting system (mercury based superconductors) were found to be varying from infrared to microwave region by increasing the temperature from liquid helium of 4.2 K to space temperature of about 100 K. Furthermore, by dealing with calculations of Lawrance-Doniach Model it was found that the plasma frequency versus temperature relation satisfies the sigmoidal distribution whereas the anisotropy factor coincides with the polynomial fitting function. Key words: resonance, microwave plasma oscillation, 3-D Bose-Einstein condensation 1. Introduction Many metals, alloys, and ceramic compounds exhibit a phase transition to a state having zero electrical resistance at Meissner critical temperature, Tc. This phe- nomenon is known as superconductivity. Mercury based copper oxide layered ceramic cuprate superconductors have the highest Meissner transition temperature of 138 K (Onbasli et al., 1996) at normal atmospheric pressure among the other high Tc superconductors, such as La2CuO4, YBa2Cu3O7, Bi2S r2Ca2Cu3O10, etc (Chen et al., 2004; Gonzalez-Jorge et al., 2004). In this point of view, the deter- mination of some electrodynamical and thermodynamical parameters of mercury cuprates becomes very important for advanced technological applications, such as magnetic resonance spectroscopy (MRS), bolometric measurements made in space, etc. In this study, it will be focused on the investigation and calculation of some electrodynamical (coupling energy, plasma frequency, anisotropy factor...) and thermodynamical (superconducting condensation energy) parameters of optimally oxygen annealed quench melt growth HgBa2Ca2Cu3O8+x (Hg-1223) supercon- ductor that has a magnetic onset temperature of 136 K (Gonzalez-Jorge et al., Vibration Problems ICOVP 2005, c© 2007 Springer. Physics, Y ld z Technical University, Istanbul, Turkeyı (eds.), . şE. Inan and A. Kırı ı 378 ÖZDEMİR, ASLAN AND ONBAŞLI 2004). These parameters have been calculated in the temperature range of 4.2- 77 K. In the light of theoretical calculations, we have determined that the plasma frequencies, which have a great role in forming superconductivity, were found to be shifting from infrared to microwave region for the temperature range of 4.2-77 K. The main factor in forming superconductivity is the harmony among phonon wave function, quantized material waves, and Cooper pairs wave function in the ab plane of the sample. Moreover, it has been realized that this harmony is extended to all Copper oxide layers i.e. along the c-axis by Josephson plasma res- onance frequency, ωp in the superconducting state. Therefore, the mercury based superconducting system becomes an oscillating cavity with the plasma frequency varying to from infrared to microwave region for 4.2 K and 77K, respectively. Clearly, these plasma frequencies are neither equal to the Cooper pair wave nor the phonon wave function frequency. In cuprate superconductors, the system has a repeated structure consisting of two superconducting CuO2 planes separated by insulating layer, namely the Josephson junction (Figure 1). In this point of view, the superconducting system can be considered as a system that many intrinsic Josephson junction are parallel to each other and coupled together by Josephson tunnelling between adjacent lay- ers. Inevidently, this coupling requires some energy and it is known as Josephson coupling energy, �J . On the other hand, the electrons in the copper oxide ab plane have a minimum or threshold thermodynamical energy which is known as super- conducting condensation energy, Eb, that spontaneously form a spin zero bosonic state. In this study, it was found that the Josephson coupling energy, �J ,which is directly related to the electrodynamical coupling between CuO2 layers along the c-axis, and superconducting condensation energy, Eb have exactly the same value for mercury based cuprate superconductors at low temperature of 4.2K. This result is also confirmed by the interlayer theory of superconductivity (Anderson, 1998). According to the interlayer theory, Josephson coupling energy must be equal to superconducting condensation energy of perfectly coupled superconducting mate- rials along the c-axis. In other words, a perfectly coupled superconducting system along the c-axis behaves like an oscillating cavity with plasma frequency. 2. Theoretical In order to calculate the c-axis electrodynamical parameters of highly anisotropic mercury cuprates, Lawrance-Doniach Model (Tinkham, 1996) has been used. According to the model, the system can be viewed as a stack of nearly ideal, intrinsic Josephson junctions and that the supercurrent flows via the Josephson tunnelling mechanism. Calculations start with the determination of Josephson CALCULATION OF MICROWAVE PLASMA OSCILLATIONS 379 Figure 1. The structural intrinsic properties of copper oxide layered mercury cuprate supercon- ductors coupling energy �J , �J = Jc Φ0 2π (1) where Jc and Φ0 are the critical current density and magnetic flux quantum, respectively. Critical current density has a key role in calculation of Josephson penetration depth, λJ , λJ = √ Φ0 2πJcµ0s (2) where µ0 represents the permeability of free space and “s” is the average spac- ing of copper oxide adjacent layers (Gough, 1998). According to the literature, Josephson penetration depth is considered as a measure of magnetic penetration depth of the magnetic field, induced by super current. This penetration depth also determines both anisotropy factor, γ , and plasma frequency, ωP. A dimensionless parameter is given by, γ = λJ s (3) In this study, anisotropy factor represents the variation of the electrodynamical properties along the c-axis of the sample. The other crucial electrodynamical quantity is c-axis Josephson plasma fre- quency, ωP, that is equivalent measure of the electromagnetic coupling. 380 ÖZDEMİR, ASLAN AND ONBAŞLI ωP = c λJ (4) where c is the velocity of light. c-axis Josephson plasma frequency is generated by the Josephson supercurrents due to quantum phase difference between CuO2 layers (Matsuda et al., ). In the superconducting state (resonance state), these Josephson plasma oscillations extend over almost all insulating layers due to the collective quantum phase oscillation of all superconducting CuO2 layers along the c-axis. In this context, superconductor that is locked to the same phase difference behaves like a quantum mechanical macro system. This is an intrinsic nature of the superconductivity for the copper oxide layered superconducting systems. A thermodynamical quantity, superconducting condensation energy, Eb, is defined by the Eb = Bc2 2µ0 (5) where Hc is the thermodynamical critical magnetic field and is deduced from the measurements of magnetization versus applied magnetic field (M-H). Thermody- namical critical magnetic field is related thermodynamically to the free energy difference between normal and superconducting states per unit volume in zero field, so called the condensation energy of the superconducting state. 3. Results According to the data from dynamic hysteresis measurements (M-H) of Hg-1223 sample at the temperatures of 4.2, 27, and 77 K (Ozdemir et al., 2004), critical current density and thermodynamic critical magnetic field values are summarized in Table 1. Referring to the measurements in Superconducting Quantum Interference De- vice (SQUID), external field was applied parallel to the c-axis of the sample, so that critical currents flow in the ab plane. It is seen that as the temperature increases from 4.2 K to 77 K, the critical current density decreases two order of magnitude. By using Jc data in Table 1, one can calculate Josephson coupling energy per unit area by Eq.(1). Superconducting condensation energy per unit volume (joule/m3), Eb, was determined by Eq. (5). In order to equalize the unit of �J and Eb ,Eb must be mul- tiplied by the average of spacing between adjacent layers, “s” that, was extracted from X-ray diffraction data as 7.887x10−10m (Onbasli et al., 1996). As is shown in Table 1, both coupling and condensation energies decrease with temperature. Moreover, it was observed that Josephson coupling and superconducting con- densation energies have the same value for liquid helium temperature of 4.2K. CALCULATION OF MICROWAVE PLASMA OSCILLATIONS 381 TABLE I. Critical current density, critical magnetic field values, Josephson cou- pling and superconducting condensation energy per unit area at different temperature of optimally oxygen annealed mercury cuprates ( Tc = 136K) T(K) Jc(A/m2) at Hc1 Hc(T) Eb(Joulem2) �J(Joulem2) 4.2 1x1012 1.024 3.285x10−4 3.29x10−4 27 1.62x1011 0.724 1.64x10−4 5.33x10−5 77 1x1010 0.612 1.173x10−4 3.29x10−6 TABLE II. Josephson penetration depth, λJ , anisotropy factor, γ , plasma frequency ωP, and fp. T(K) λJ(m) γ ωP (rad/s) fP (Hz) 4.2 0,575x10−6 729 5,217x1014 8,3x1013 27 1,43x10−6 1812 2,09x1014 3.33x1013 77 5,75x10−6 7290 5,217x1013 8,3x1012 This result shows that, mercury cuprate superconductors, which display electro- magnetic coupling in 3D, confirm the interlayer theory of superconductivity at low temperature. Josephson penetration depth, anisotropy factor, and plasma frequency values were determined by Eq. (2), (3), and (5), respectively and are given in Table 2. According to Table 2, anisotropy factor and λJ increase with the temperature, whereas plasma frequency decreases. Due to the high anisotropy factor along the c-axis, the plasma frequencies occur in the far-infrared region and extremely low compared to the plasma frequencies along the other directions of the same material and that in the conventional low temperature superconductors (Shibata and Yamada, 1996). The result that the plasma frequency of mercury cuprates varies from infrared to microwave region with temperature is in good agreement with the previous statement. An estimation of the plasma frequency, fp, at 100 K was obtained by sigmoidal fitting function is given in Fig. 2. 4. Discussion and conclusion Despite of the fact that the high Tc cuprate superconductors have high anisotropy factors along the c-axis, the system becomes a huge quantum mechanical macro molecule due to the plasma resonance phenomenon. In this point of view, the har- mony of bosonic Cooper pairs in the ab-plane extends through the all copper oxide 382 ÖZDEMİR, ASLAN AND ONBAŞLI Figure 2. Plasma frequency, fp , versus temperature relation satisfies sigmoidal fitting and fp has a value of 7.971x1012 Hz for 100 K layers with plasma resonance frequency. In other words, perfectly coupled layered superconducting materials like mercury cuprates can be served as an example of 3-dimensional Bose-Einstein condensate in nature. Precise and reliable detection of the magnetic fluctuations in the space, on the earth and that of the brain waves of human being can only be achieved by the devices that have been produced by the mentioned perfectly coupled superconducting materials. References Anderson, P. (1998) c-Axis Electrodynamics as Evidence for the Interlayer Theory of High- Temperature Superconductivity, Science 279, 1196–1208. Chen, M., Donzel, L., Lakner, M., and Paul, W. (2004) High temperature superconductors for power applications, Journal of European Ceramic Society 24, 1815–1821. Gonzalez-Jorge, H., Gonzalez-Salgado, D., Peleteiro, J., and Domarco, G. (2004) Ability of a con- tactless inductive device for the characterization of the critical current versus temperature in superconducting rings at temperatures close to Tc, Cryogenics 44, 115–122. Gough, C. (1998) The Gap Symmetry and Fluctuations in High-Tc Superconductors, New York, Nato ASI Series B Physics Plenum Press. Matsuda, Y., Gaifullin, M., Kumagai, K., Kadowaki, K., and Mochiku, T., Collective Josephson Plasma Resonance in the Vortex State of Bi2Sr2CaCu2O8, Phys. Rev. Lett. 75. Onbasli, U. (2000) Effect of oxygen post-annealing at the critical parameters of mercury cuprates, Physica C 332, 333–339. Onbasli, U., Wang, Y., Naziripour, A., Tello, R., Kiehl, W., and A.M. Hermann (1996) Transport Properties of High-Tc Mercury Cuprates, Phys. Status Solidi B 194, 371–380. Ozdemir, Z. G., Aslan, O., and Onbasli, U. (2004) Determination of c-Axis Electrodynamics Parameters of Mercury Cuprates, In Spectroscopies of Novel Superconductors, Sitges, Spain. Shibata, H. and Yamada, T. (1996) Superconducting-plasma resonance along the c axis in various copper oxide superconductors, Phys. Rev. B , 1996 54, 7500–7512. Tinkham, M. (1996) Introduction to Superconductivity, New York, McGraw-Hill Inc. 383 383–388. ONE - DIMENSIONAL WAVE PROPAGATION PROBLEM IN A NONLOCAL FINITE MEDIUM WITH FINITE DIFFERENCE METHOD Ahmet Özkan Özer1 and E. İnan2 1Faculty of Science and Letters, İstanbul Technical University, 34469, Maslak, İstanbul, Turkey 2Faculty of Arts and Sciences, Işık University, 34398, Maslak, İstanbul, Turkey Abstract. In this work, the wave propagation problem in one - dimensional finite medium is investigated in the context of nonlocal continuum mechanics. The main purpose of the paper is to demonstrate the end effects. Numerical solutions are obtained by the finite difference method. Furthermore, the differences between the nonlocal infinite and nonlocal finite cases are shown. It is observed that the velocity of the propagating waves is slowing down and the amplitude is decreasing due to the end effects. method 1. Formation of the Problem For nonlocal, linear and isotropic media, fundamental equations are given as fol- lows (Eringen, 1974, 1984) : tkl,k + ρ( fl − al) = 0. (1) tkl = ∫ V σ′kl(x ′)dv(x′). (2) σ́kl = λ ′(|x′ − x|)e′rr(x′)δkl + 2µ′(|x′ − x|)e′kl(x ′). (3) e′kl = 1 2 (u′k,l − u ′ l,k), u ′ k ≡ uk(x ′). (4) Here tkl is the stress tensor, fl is the body forces, al is the acceleration vector, λ′ and µ′ are the nonlocal moduli and λ and µ are Lame constants and uk is the dis- placement vector. For homogenous materials nonlocal moduli may be expressed as in the following form (λ′, µ′) = (λ, µ)α(|x′ − x|) Vibration Problems ICOVP 2005, c© 2007 Springer. Key words: nonlocal theory, wave propagation, difference differential equation, finite difference (eds.), . şE. Inan and A. Kırı 384 ÖZER, İNAN Here α is the nonlocal kernel function. In the present work it is taken as linearly decreasing, α(|x|) = {1 a (1 − |x| a ), |x| < a. Here a is the lattice distance. Substituting equation (4) into equations (2) and (3), and ignoring the body forces, the equation of motion takes the following form, ∂ ∂xk ∫ V λ′(|x′ − x|)σkl(x′)dv(x′) = ρül(x). (5) And for one - dimensional finite problem we write ∂ ∂x ∫ 2h 0 λ′(|x′ − x|) ∂u ∂x′ dx′ − ρü = 0. (6) Here, l = 2h is the length of the one-dimensional medium. Effect function van- ishes for |x′ − x| > a. So, it is appropriate to analyze the above integral in subin- tervals. Every subinterval has a length a, the lattice distance. In this case kernel function α(|x′ − x|) become discontinuous in the first and the last intervals. Thus, the first and the last intervals should be considered seperately. By then, ui will be used to denote the displacement of ith interval; x ∈ I ⇒ u0, x ∈ II ⇒ u1, . . . , x ∈ N ⇒ uN−1 (7) a system of difference - differential equations; ü0 − c21 c2a [u1(x + a, t) − 2u0(x, t) + u0(0, t)] = 0; x ∈ I ü1 − c21 c2a [u2(x + a, t) − 2u1(x, t) + u0(x − a, t)] = 0; x ∈ II ... üN−1 − c21 c2a [uN−1(Na, t) − 2uN−1(x, t) + uN−2(x − a, t)] = 0; x ∈ N. (8) This system is solved by use of finite difference method. 2. Numerical solutions 2.1. FINITE DIFFERENCE METHOD In finite difference method, the solution domain {(x, t) : x ∈ [0, l = 2h], t ∈ [0, t f ]} (9) After some mathematical manupilations, the equation of motion (6) is reduced to ONE - DIMENSIONAL WAVE PROPAGATION PROBLEM 385 is discretized into cells. Here t f is the time where the approximation procedure ends. The space step should be smaller than lattice parameter a to understand the end effects. The spatial mesh size and the time step is expressed as shown by ∆t = t f J , J : number o f the points, 0 ≤ t ≤ t f ∆x = a4 , 0 � x � l (10) Each approximate solution for each subintervals are defined as u0(xn, t j) ≈ U 0 (xn, t j), . . . , uN−1(xn, t j) ≈ U N−1 (xn, t j) ui(xn, t j) ≈ U i (xn, t j) ≈ U i j n (11) In the first step, we have used forward difference for the first equation of the system, and backward difference for the last equation. The system of equations U 0 j+1 1 = 2U 0 j n − U 0 j−1 n + ν2[U 1 j n+4 − 2U 0 j n + U 0 j 0 ] ... U i j+1 n = 2U i j n − U i j−1 n + ν2[U i+1 j n+4 − 2U i j n + U i−1 j n−4 ] ... U N−1 j+1 n = 2 U N−1 j n − U N−1 j−1 n + ν2[ U N−1 j 4N − 2 U N−1 j n + U N−2 j n−4 ] (12) where υ2 = c21 a2 (∆t)2. (13) Same transformation procedure can be applied to the initial conditions ui(xn, 0) = U i 0 n , u̇i(xn, 0) = U̇ i 0 n . (14) and also to the boundary condition u(0, t j) = U 0 j 0 , u(l, t j) = U N−1 j (4N) . (15) 2.2. AN EXPLICIT SCHEME OF WAVE EQUATION An explicit sheme of the classical wave equation is given in the followings which we need for the comparison. In classical continuum mechanics, wave equation for one-dimension is ∂2u ∂t2 = c2 ∂2u ∂x2 , c2 = λ + 2µ ρ . (16) expressions (10), (Morton and Mayer) (8) for the following steps ( j = 2, 3, ...) are 386 ÖZER, İNAN We use central difference operator on both derivative for j = 1 V1n = ∆t.( ∂ ∂t V0n ) + V 0 n + ν 2 h[V 0 (n+1) − 2V 0 n + V 0 (n−1)], (17) and for j = 2, 3, ..., we write V j+1n = 2V j n − V j−1 n + ν 2 h[V j n+1 − 2V j n + V j n−1], (18) where u(xn, t j) ≈ V(xn, t j) ≈ V jn, u̇(xn, t j) ≈ ( ∂∂t V j n) (19) and υ2h = c21 (∆x)2 (∆t)2. (20) 3. Stability and consistency CFL (Courant, Friedrichs, Lewy) condition is a necessary condition to provide the stability of a scheme of partial differential equations. One can easily get this condition as |νh| ≤ 1. (21) The stability of the scheme can be determined by Fourier modes of the scheme of the infinite problem and we write Ū jn = (λ) jeikn∆x (22) Here, λ = λk is a complex valued amplification factor which its magnitude show the stability and dissipation properties of the difference scheme, and k is the real wave number of the arbitrary harmonic wave component. After some mathemati- cal manipulations, one may get λ1,2 = −υ2 sin2 ka 2 + 1 ± 2iυ sin ka 2 √ (1 − υ2 sin2 ka 2 ). (23) and find |λ±| = 1. Here, the difference between the equations (13) and (20) appears only in ∆x. ONE - DIMENSIONAL WAVE PROPAGATION PROBLEM 387 3.1. CONSISTENCY To show that the numerical scheme is consistent, the following expressions should be satisfied: T (x, t) = O((∆t)p), ∆t → 0 |T (x, t)| ≤ C[(∆t)p], ∆t → 0. (24) Here T (x, t) is the truncation error and C is a real constant. Truncation error of the system may be given as T1(x, t) = δt2 u0(x,t) (∆t)2 − c21 a2 [u1(x + a, t) − 2u0(x, t) + u0(0, t)] T2(x, t) = δt2 u1(x,t) (∆t)2 − c21 a2 [u2(x + a, t) − 2u1(x, t) + u0(x − a, t)] ... TN(x, t) = δt2 uN−1(x,t) (∆t)2 − c21 a2 [uN−1(4N, t) − 2uN−1(x, t) + uN−2(x − a, t)]. and T1(x, t) = T2(x, t) = T3(x, t) = . . . = TN(x, t) = O((∆t) 2). As it is seen above, the truncation error of the scheme is of order (∆t)2. 4. Numerical results In this section, the numerical results are given to illustrate the end effects in nonlocal theory. All programs of the following graphs given here are coded in Mathematica. We consider a finite medium made of aluminium and take l = 100 × a = 4 × 10−6cm , a = 4 × 10−8, ρ = 0.002699kg/cm3, λ + 2µ = 1.12 × 106N/(cm)2. (25) Several comparisons are made in this work, to see the end effects like nonlocal infinite-nonlocal finite, nonlocal infinite-local finite, nonlocal finite-local finite cases. Due to the page limitation, we gave only one numarical example here to show the differences in both waves. 5. Conclusions As it is mentioned in the abstract the main purpose of this work is to show the boundary effects on finite non-local problems. Because of the volume integrals in the constitutive equations in non-local theory, most of the problems are considered in infinite media to avoid the diffuculties occur in the final equations due to the finite boundaries. The results of this work show that the end effects on the wave 388 ÖZER, İNAN Figure 1. propagation problem in one dimensional finite medium in non local theory are not very dominant. It causes some decrease in the wave velocity and the waves slow down. Second effect is the decrease in the amplitude. But the magnitude of this decrease is very small. Similar results are also observed by the comparison of nonlocal finite and nonlocal infinite cases. References Eringen, A.C. (1974) Nonlocal elasticity and waves,Continuum Mechanics Aspect of Geodynamics and Rock Fracture Mechanics , 81–105. Eringen, A.C. (1984) On continous distributions of dislocations in nonlocal elasticity,J. Appl. Phys. 56(10), 2675–2680. Morton, K.W., Mayers, D.F. (1993) Numerical solution of partial differential equations, Cambridge University Press. 389 389–394. DYNAMIC RESPONSE ANALYSIS OF ROCKING RIGID BLOCKS SUBJECTED TO HALF-SINE PULSE TYPE BASE EXCITATIONS Mutlu Özer1 and G. Füsun Alışverişci2 1 1600 Holloway Ave. San Francisco State University, School of Engineering, San Francisco, 94132, CA, USA 2 Yıldız Technical University, Barbaros Bulvarı, 34349, İstanbul, Turkey Abstract. The motion of a rocking block subjected to half-sine pulse base excitation is analyzed by the mathematical model of “lumped-mass approximation”. The governing equation of the motion of the rocking block is derived as a second order ODE. The solutions of the derived equation for different size of rocking blocks are illustrated by mat lab figures in terms of displacement, velocity and acceleration responses. The coefficients of the equation of the minimum required overturning acceleration of rocking blocks are obtained, and the overturning instant of the rocking block is investigated through force vibration and free vibration. Parameter studies are also performed to verify the results accordingly to the Housner’s (1963) postulation and to the report of N. Makris (1998) about dynamic response of rocking blocks. Key words: lumped-mass approximation, phase, oscillation frequency, acceleration 1. Introduction The motion illustrated in Figure. 1(a), is called a rocking motion of rigid bodies. In the case of strong ground motion heavy industrial equipment such as heavy electrical transformers and large oil or chemical storage tanks undergo rocking motion that may overturn or slide from their foundation that can cause severe damage to industrial facilities and human lives. In such case, pumps and other devices required for rescue efforts would also not be in service because of delays to restoring power. It is experienced from recent earthquakes that overturning of industrial equipment are very costly to human lives and physical environment. Therefore proper assessment for dynamic response properties of rocking rigid body became a great interest for the engineers. The minimum acceleration amplitude of a half-sine pulse that will overturn the rigid block is controversial topic among researchers and further study shall be required. The minimum acceleration amplitude of a half-sine pulse that will overturn the rigid block was postulated by Housner, 1963 and later by N. Makris, 1998. Although both researchers had used the same model and governing equa- tions, they have reached different conclusion in terms of the amplitude of the Vibration Problems ICOVP 2005, c© 2007 Springer. ¸ (eds.), . şE. Inan and A. Kırı 390 ÖZER, ALIŞVERİŞÇİ minimum acceleration that will overturn rigid body and the overturning instant of the rocking rigid block. Figure 1. (a) Schematic of a rocking rigid block. (b) Lumped-Mass Model The controversy between both researchers would not be easily clarified since the governing equations of the rocking block they used are non-linear and ex- tremely difficult to solve it. In this report the governing equation of the rock- ing block is derived as a second order ODE by a mathematical model of the lumped-mass approximation. 2. Dynamic response analysis of a rocking rigid blocks by the model of lumped-mass approximation Within the limit of the linear approximation, the conditions for a block to overturn are also studied in this report by the proposed lumped-mass approximation model as illustrated in Figure. 1(b). By this model the dynamic equation for the minimum acceleration that block will overturn is derived as follows. Assumptions: • Rectangular shape rigid body is modeled to a dimensionless mass that moves forward- and-back through a path of c.g. of rectangular rigid body. • When t = 0, θ = 0,U = −b, U̇ = 0, and the ground acceleration is to left. • There is no energy losses through the path. DYNAMIC RESPONSE ANALYSIS OF ROCKING RIGID BLOCKS 391 Definitions : m Mass of rigid body x, y Moving axis attached to rigid block X,Y Imaginary fix axis 2h The height of rigid block 2b The width of rigid block R Position vector of lump mass of rigid block, R = √ b2 + h2 θ Rotation angle of lump mass α Slenderness angle U The amplitude of displacement of lumped-mass (when θ = 0,U = −b or θ = α,U = 0) Üg(t) Harmonic ground acceleration Üg(t) = ap0 sin(ω̄pt + φ) where, ap0 The amplitude of peak ground acceleration, ω̄p Excitation frequency p oscillation frequency of a rigid block, (p = √ 3g/4R, rectangular block) φ phase when rocking initiates, φ = ∗sin−1(αg/ap0), (Housner-1963) 3. Dynamic equilibrium of motion: m(−Ü + Üg) − mg sin(α − θ) = 0 sin(α − θ) = UR Ü + gR U = ap0 sin(ω̄pt + ∗phi) Ü + 43 p 2U = ap0 sin(ω̄pt + ∗phi) (1) 4. Matlab figures of the solution of the proposed equation (1) The following Mat Lab figures are the solution of equation (1). The figures il- lustrate dynamic responses of rocking rectangular bodies subjected to harmonic ground excitation with various input parameters. The results are also tabulated in Table 1. 5. Discussion and conclusion • This analytical and numerical study by lumped-mass model confirms Housners’ postulation about the overturning condition of a rocking rigid blocks. A rigid block will overturn (θ = α ) at the time (t = (π − φ)/ω̄p ) when half-sine pulse expires as postulated by Housner. However, the minimum acceleration amplitudes 392 ÖZER, ALIŞVERİŞÇİ Figure 2. Dynamic response of a rocking rigid block : b = 0.2m, h = 0.6m, (◦ap0 = 0.6288g), (�ap0 = 0.6738g, Housner) that rocking rigid blocks will overturn are less what Housner approximated by the expression of (ap0 ≈ αg √ 1 + (ω̄/p)2). The outcomes of lumped-mass ap- proximation model confirms that the equation, ap0 ≈ αg1 + β(ω̄/p) in the report of PEER-1998/05 by N. Makris and Y. Rouses, which is the best formulation to define the required minimum acceleration amplitude that rocking rigid blocks will overturn. The coefficient β of the required minimum acceleration that rocking rigid blocks will overturn is approximated by N. Makris as β ≈ 0.5. It is indeed a very accurate approximation for small size rocking blocks with b ≤ 0.3 when α = 18.4◦. However, since β varies by the size of rocking blocks and excita- tion frequency, the amplitude of the required minimum acceleration that rocking rigid block will overturn are not accurate when slender angels are different than α = 18.4◦ and also for the larger size rocking blocks with b > 0.3. Therefore the coefficients of β’s for different size of blocks are obtained at two different DYNAMIC RESPONSE ANALYSIS OF ROCKING RIGID BLOCKS 393 TABLE I. Comparison of the minimum required peak ground acceleration that rigid block will overturn Studies on min. req. peak acc Housner N. Makris Lumped-Mass that rigid block will overturn Approximation Model Parameters of Minimum the rigid Required ap0 ≈ αg √ 1 + (ω̄/p)2 ap0 ≈ αg(1 + β(ω̄/p)) ap0 ≈ αg(1 + β(ω̄/p)) rectangular Peak Ground β ≈ 0.5 β, See Fig. 3 block: Acceleration b = 0.2m, (ap0) that ω̄ = 2 � π, rocking block 0.6736 � g 0.6165 � g 0.6202 � g α = 18.4◦, will overturn Overturning condition of the rocking rigid block: Rigid block will overturn at Yes No, Yes time t = (π − φ)/ω̄, which is later, During free the time that half-sine pulse vibration regime. expires (Housner, 1963) Figure 3. β obtained by lumped-mass model at α = 18.4 and varies width of b(m) excitation frequency of ω = π and ω = 2π. As an example, the corresponding result plotted in the Fig. 3 at α = 18.4◦. • It is also confirmed that larger block will overturn with longer period of pulse while small block is more sensitive to higher peak ground acceleration as indicated in the report of N. Makris. If we consider two different half-sine accel- eration pulses with the same product as (ap0Tn)s = (ap0Tn)l , then larger block will overturn at the condition of (ap0)l ≤ (ap0)s/2. 394 ÖZER, ALIŞVERİŞÇİ • By the model of lumped-mass approximation, the dynamic analysis of the rocking blocks subjected to half-sine pulse base excitation is pretty simple, and proposed governing equation (1) would help engineers to obtain quick and reliable dynamic properties of rocking rigid bodies. References Housner, G. W. (1963) The behavior of inverted pendulum structures during Earthquakes, Bull. Seism. Soc. Am. 53, 404-417. Makris, N. R. (1998) Rocking Response and Overturning of Equipment Under Horizontal Pulse- Type Motions, Report PEER-1998/05 Pacific Earthquake Research Center, University of California ,Berkeley, USA. 395 395–401. A NEW SPECTRAL ALGORITHM FOR 3-D WAVE FIELD IN DEEP WATER ¨ 1 Nazmi Postacıoğlu2 and Bahar Zora1 1 Turkey 2Faculty of Science and Letters, Technical University, 34469, Turkey Abstract. A new spectral technique to calculate the dispersive gravity wave field created by small- scale disturbances is introduced. The technique is based on the use of sources and sinks and an integral equation formalism to calculate the potential field that satisfies the kinematic condition on the bottom. Comparisons with the laboratory experiments show that the technique is capable of representing the fundamental features of the dispersive wave field to a satisfactory extent. Key words: Tsunamis, deep water, dispersive wave 1. Introduction In this work we develop a spectral scheme to calculate 3-D wave field created by small-wavelength disturbances in deep water. The most immediate application of the method is on the landslide Tsunamis, yet the technique will also find possible applications in naval and offshore engineering. When a 3-D domain of incom- pressible fluid with a free surface is disturbed by a small-scale solid displacement on the bottom (that might be flat or non-flat) or within the fluid, the gravity wave field that would be generated on the free surface is dispersive, therefore it is not possible to employ the usual shallow water equations. There are existing schemes developed for the deep water (Ward 2000) and fully discretized numerical wave- tank techniques (Grilli 1997) however a spectral technique that can both account to the solid motion and the wave scattering from the uneven bottom is still lacking. Our aim in this work is to fill this gap. The difficulties in the modelling of the gravity waves in deep water mainly arises from the mathematical complications that relate to the free surface conditions, in this work we aim to find an appropriate source-sink distribution that would satisfy the non-flux condition on an uneven bottom, we then calculate the free surface response to the sources and sinks in such a way that both the kinematic and dynmic conditions on the free surface are satisfied. Vibration Problems ICOVP 2005, c© 2007 Springer. M. Sinan Ozeren, ˙ ˙Faculty of Mines, Istanbul Technical University, 34469, Istanbul, İstanbul İstanbul, (eds.), . şE. Inan and A. Kırı 396 ÖZEREN, POSTACIOĞLU, ZORA 2. The formulation of the problem and the source-sink distributions The velocity potential Φ for an inviscid and incompressible fluid is governed by the Laplace’s equation ∇2Φ = 0 (1) We take a case in which a block of thickness much smaller than the water depth slides on a profile depicted in the Figure 1. Our aim is to calculate the surface gravity wave field generated as a result of the sliding motion. We have to satisfy the non-flux condition everywhere the fluid is in contact with a solid surface, this condition can be written as ∂nΦ = 0 (2) where n stands for the local normal. Since H x f where the depth is constant. Apart from the prescribed sources χ at the edges of the block (to describe the motion of the block), a continuous source (sink) distribution σ over the entire topography will be inferred from the Neumann condition at the bottom. Here a necessity arises to make use of an integral equation, because the kinematic (Neumann) condition on the bottom couples the induced sources with each other. Before proceeding to the formulation of the integral equation, as an intermediate 3-D WAVE FIELD IN DEEP WATER 397 step, the linear response of the fluid of uniform depth to an impulsive source must be evaluated. We base our calculations on the generic Green’s function G(1)(t, r, t0, r0) = −Q 4π ( 1 |r − r0| + 1 |r − r1| ) δ(t − t0) (3) which represents the 3-D fluid response to an impulsive volume injection of the strength Q. Here r0 is the position of the explosion. r1 is the image of r0 with respect to the plane z = −h1 i.e. r1 = r0 + 2(−h1 − z0)k̂ , k̂ is the unit vector along z axis, h1 is the depth of the sea, note that z0 ≥ −h1. The use of the image in (3) insures that ∂zG(1) vanishes for z = −h1. Fourier transforming this Green’s function gives G(1) = −Q16π3 ∫ dω ∫ dkxdky ( exp(−k |z − z0| + exp(−k(z − z1) ) × exp(i(ω(t − t0) − k · (ρ − ρ0))/k (4) where ρ = xî+ yĵ, z1 = z0 +2(−h1 − z0) and kx,ky are the horizontal wavenumbers. For the pressure at the free surface the linear theory dictates p(1) = ρD − ∂G(1) ∂t ∣∣∣∣∣∣ z=0 − gη(1) (5) where ρ is the density of the fluid, g is acceleration due to gravity. The linearized kinematic condition at the free surface is ∂G(1) ∂z ∣∣∣∣∣∣ z=0 = ∂η(1) ∂t . (6) This, in turn means that the pressure at the free surface can be written as −ρDQ 16π3 ∫ dω ∫ dkxdky ( exp(−k |z0| + exp(−k ∣∣∣z′0 ∣∣∣ ) exp(iω(t−t0)−ik·(ρ−ρ0)) ( iω k − g iω ) (7) Using the linearity of the formulation, we can now generate another Green’s func- tion (say G(2) ) in such a way that when the pressure corresponding to the latter is added to the pressure that corresponds to the first Green’s function we will obtain zero pressure at the free surface (ignoring the atmospheric pressure), it reads G(2) = Q8π2 ∫ dkxdky ( ω20(k) gk + 1 ) [ exp (−k |z0|) + exp (−k |z1|) ] cosh(k(z+h1)) sinh(kh1) × exp (ik · (ρ − ρ0)) ( δ(t−t0) k − ω0 k sin (ω0(t − t0) ) . (8) Here we have to organize the images in such a way that they are positioned along the vertical axis in an equispaced manner but their signes alternate with respect 398 ÖZEREN, POSTACIOĞLU, ZORA to the free surface. Our final aim is to have the impulsive part of G(1) + G(2) to amount to zero at the free surface. The signs of the images do not alternate with respect to the bottom, therefore the normal derivative of G(1)+G(2) vanishes at the bottom. A short hand notation needs to be introduced the for image series C(rA, rB) = 1 4π +∞∑ n=0 −εn+1√ |ρA − ρB|2 + (zA − zB,n)2 + +∞∑ n=1 εn√ |ρA − ρB|2 + (zA − z′B,n)2 (9) here C(rA, rB) represents the value of the potential at rA , produced by a unit source at rB. 3. Scattering and the integral equation approach There are numerous contributions to the normal flow on the bottom, prohibited by the kinematic condition. These are namely the flows associated with the al- ready existing gravity waves, the induced and prescribed sources. If any point P is selected on the bottom, then the total normal flow at P at time step N is ∂nΦg,N−1(tN , rP)+(σN(rP) + χN(rP)) /2+lim ε→0 ∫ T−d(ε) ds(σN(rs) + χN(rs))∂nC(rP, rs) (10) where ∂n denotes differentiation in the inward (towards the fluid) direction, eval- uated at point P. rP is the position of point P. N is time step. σN(rs) denotes the induced source distribution on the sea bed while is used for any prescribed source. As rS scans the non-uniform sea bed, the integral gives the contribution to the normal flow at P created by the sources elsewhere. Φg,N−1(tN,rP) is the scalar potential of the gravity waves excited at the time steps t0, t1....tN−1. It has been shown that the velocity field associated with the gravity waves excited at t = tN vanishes for t → t+N . In equation (10) the integrand is not well defined if the point P and the variable of integration coincide. Therefore the integration is carried over the whole topography except for a small disc d of radius ε, the center of the disc being P. χN is a line source but we may assume it has a small spread β in the direction of the motion of the block with β > ε. According to the Gauss’ law the contribution from the disc to the normal flow is (σN(P) + χN(P)) /2. The factor 1/2 is due to the fact that only half of the “flow” from the disc goes into the liquid region. If we equate G(1) + G(2) to χn(P)/2 we obtain ∂nΦG,N−1(tN , rP) + σN(rP)/2 + 1 4π lim ε→0 ∫ T−d(ε) ds(σN(rs) + χN(rs))∂nC(rP, rs) = 0 (11) 3-D WAVE FIELD IN DEEP WATER 399 If the gravity waves excited at time steps t1, t2...tN−1 are known, the induced sources at time step N will be obtained solving numerically integral equation (11). Non-impulsive part of G(2) i.e. the gravity waves excited at time step tN , are related to the sources through equation 8. Equation 8 can be rewritten in the complex form as G(2)non−impulsive = Q 8π2� ∫ dkxdky ( ω20(k) gk + 1 ) [ exp (−k |z0|) + exp (−k |z1|) ] cosh(k(z+h1)) sinh(kh1) × exp (ik · (ρ − ρ0)) iω0(k) exp(iω0(t−tN ))k (12) where � stands for the real part. The gravity waves excited by the sliding block can be represented by the potential Φg = � (∫ dkxdky ak(t) exp(iω0(k)t − ik · ρ) cosh(k(z + h1)) sinh(kh1) ) (13) Complex amplitudes ak(t) vary with time because of the sources. These sources correspond to the scattering of the waves by the topography and to the prescribed motion of the block. At each time step during the tsunami evolution, we must add the waves excited from the impulsive volume injections to the already existing wave field. If only the sources on the surface element ds during time span dt are taken into account, with the help of equation (10) and (11) one finds that the resulting increment of the complex amplitudes at time step tN is dak = dQdt 8π2 exp(−iω0(k)tN + ik · ρ0) iω k ω20(k) gk + 1 [ exp (−k |z0|) + exp (−k |z1|) ] (14) where ρ0 is the horizontal coordinate of the surface element ds, dQ = (σ + χ) ds, χ = −2v dHds is the prescribed source distribution, the induced source distribu- tion is obtained solving the integral equation numerically, hence σ ≡ σ(t, s,Φg). Integrating equation (14) on the topography, one gets dak dt ∣∣∣ t=tN = 18π2 ∫ ds ( σ(tN , s,Φg) + χ(tN , s) ) × exp(−iω0(k)t0 + ik · ρ(s)) iω0k ( ω20(k) gk + 1 ) [ exp (−k |z0(s)|) + exp (−k |z1(s)|) ] (15) In our algorithm the wave vector k is a continuous parameter but for the pur- pose of numerical integration it will be discretized. By discretization it is meant that a finite number of wave vectors will be used i.e. k1,k2, ...,km. Accordingly the velocity potential Φg ≡ Φg(t, r, ak1 , ak2 , ..., akm) is inherently a function of m complex amplitudes. Equation (15) describes m coupled ordinary differential equations that govern the complex amplitudes. 400 ÖZEREN, POSTACIOĞLU, ZORA Figure 1. The bottom relief for the moving block 4. The spectral approach The spectral approach is based on a Legendre expansion of the induced sources and sinks of the following form σN = ∑ k 3-D WAVE FIELD IN DEEP WATER 401 Figure 2. The intermediate state of the Tsunami propagation calculated for a moving disturbance with the trajectory given in Figure 1 5. Conclusions The spectral approach given in this article proved to be a useful tool to model the dispersive surface wave propagation created by small-scale disturbances on the bottom or within the fuid. The technique is designed architecturally in such a way that the passage to fully discretized schemes such as the finite elements would not require complicated integral transformations. A calculation for a dispersive wave field that we did to compare our scheme with a laboratory experiment (Watts 2000) also indicate that the scheme is able to represent very well the first order characteristics of the wave field. References Grilli, S. T., Horillo, J. (1997) Numerical generation and absorption of fully nonlinear periodic waves, J. Eng. Mech. 123, 1060-1069. War, S. N. (2001) Landslide Tsunami, J. Geophys. Res. 106, 201-11, 215. Watts, P. (2000) Tsunami features of solid block underwater landslides, J. Wtrwy, Port Coast and Oc. Engng., ASCE 126, 144-152. 403 403–408. EXPERIMENTAL AND NUMERICAL ASSESSMENT OF VIBRO-ACOUSTIC BEHAVIOR OF RUBBER-DAMPED RAILWAY WHEELS Luděk Pešek and Ladislav Pu ◦ st Institute of Thermomechanics AS CR, Prague, Czech Republic Abstract. The reduction of noise and vibrations is very important task in many industrial and trans- port applications. The sources of intensive noise and vibrations are also tram and railway wheels at high speeds. Therefore the modern types of steel railway wheels contain the visco-elastic paddings. The first problem treated in this contribution is concerned with the theoretical and experimental investigation of the thermo-mechanical properties of rubber-like damping elements loaded with prestress by harmonic force. The dynamic modal and spectral properties of the whole railway wheel with damping elements will be investigated by the 3-D FEM model as the second problem. Key words: vibration, eigenfrequency, damping, thermo-mechanical properties, rubber 1. Introduction The damping elements of elastomers used for reduction of noise in railway wheels, have to work at various operational conditions, i.e. temperatures, forces, defor- mations and frequencies. These factors influence the rheological properties of elastomers and have great effect also on reliability of resilient elements. The study was open by investigation of 2 DOF mathematical model from the point of view of lost energy, heat transfer inside the element, into surrounding air and into the bodies in contact (Pesek et al., 2003a). Mathematical model of resilient element consists of two parts, corresponding to the measuring possibilities; the tempera- ture was measured near the center of resilient element and near to its surface. New methods of measurement of sample temperatures were developed. The numerical 3D model of the rail tram wheel was created by FEM in ANSYS 8.1 (Pesek et al., 2003b). The topology of elements was constructed with accordance of production drawing of the real wheel. Mode shapes of vibrations were computed for more than 10 lowest eigenfrequencies. 2. Investigated model, comparison of measured and computed results The resilient element used for experimental study of rheological properties is a right parallelepiped (50x25x40 mm3). It was loaded by force F(t) generated by Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 404 PEŠEK, PU ◦ ST INSTRON (40 kN) and controlled by system LABTRONIC 8800. Load F(t) con- sists of pre-stress Fst and the time variable component F0 cosωt. Mathematical model with regard to the different temperatures θ1 and θ2 composes of inner and outer parts (m1,m2). S / 2u 2 S /2u 2 S / 2a 2 S12 S12 S 1u S 1b S / 2b 2 E 1u E 12 E 1b E 1a E 1v m 2 2 m 1 S / 2a 2 S 2a E 2v E 2a E 2u E 2b �� �� � E2 m 2 2 S / 2b 2 �� E 1 � Figure 1. Block diagram showing masses, energy sources and flows The total mass m is di- vided into the inner mass m1 and outer mass m2. A block diagram show- ing these masses, energy sources and flows is in Figure 1. The external force F(t) = S 1σ(x, θ1) + S 2σ(x, θ2) = f1(x, θ1) + g1(x, ẋ, θ1) + f2(x, θ2) + g2(x, ẋ, θ2) acts on the up- per surface and is distrib- uted into two parts cor- responding to the masses m1,m2. The only damp- ing components gi(x, ẋ, θi) produce the lost energy during the deformation. Increments of the lost energy in time interval ∆t are ∆Eiv = gi(x, ẋ, θi) ∆x = gi ẋ∆t, i = 1, 2 and equal to the increments of heat. Heats ∆Eiv generated in the resilient element escape partly into surrounding metal bodies or air, and they increase only partly the sample temperatures θ1 and θ2. The heat balances of inner (i = 1) and outer (2) parts are ∆E1ν = ∆E1u + ∆E1b + ∆E1a + ∆E1θ + ∆E12, ∆E2ν = ∆E2u + ∆E2b + ∆E2a + ∆E2θ − ∆E12, (1) where ∆Eiν are increments of heat resulting from the inner damping in ∆t,∆Eiθ are increments of heat staying in the element and increasing its temperature, ∆Eiu,∆Eib are heats drained from the system to the metal parts (“ u” upper,“ b” bottom contact): ∆Eiu = S imαm(θi − θmu), and ∆Eib = S imαm(θi − θmb) , i = 1, 2, S im and αm are contact surfaces with metal and coefficient of heat transfer to metal, θmu and θmb are temperatures of upper and bottom supports. ∆Eia = S iaαa(θi − θa) are heats flowing from the sample into air. ∆E12 = S 12α12(θ1−θ2) is heat draining from the inner into the outer part of element. The heat energies ∆E1θ,∆E2θ staying in the both parts can be calculated from Eq. 1. The increments of the heat causes increments of temperatures ∆θi = ∆Eiθ/cνVi, i = 1, 2, where Vi are the volumes of inner and outer parts of the sample and and cν is the specific heat capacity of the sample. Increments ∆θ1 and ∆θ2 of temperatures in the center of sample and near its air surface during the time interval ∆t are then: EXPER. AND NUMER. ASSESSMENT OF VIBRO-ACOUSTIC 405 ∆θ1 = [∆E1ν − ∆E1u − ∆E1b − ∆E1a − ∆E12]/(cνV1), ∆θ2 = [∆E2ν − ∆E2u − ∆E2b − ∆E2a + ∆E12]/(cνV2). (2) The sample temperatures θ1 and θ2 in the time t = n∆t are ascertained by numerical solution, using values θi(n−1), xn−1,ẋn−1 from the previous step θ1n = θ1(n−1) + [ g1(xn−1, ẋn−1, θ1(n−1))ẋn−1 − S 1mαm(2θ1(n−1) − θmu − θmb)− −S 1aαa(θ1(n−1) − θa) − α12S 12(θ1(n−1) − θ2(n−1)) ] ∆t/(cνV1), θ2n = θ2(n−1) + [ g2(xn−1, ẋn−1, θ2(n−1))ẋn−1 − S 2mαm(2θ2(n−1) − θmu − θmb)− −S 2aαa(θ2(n−1) − θa) + α12S 12(θ1(n−1) − θ2(n−1)) ] ∆t/(cνV2). (3) 0 50 100 150 200 250 300 −10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 F O R C E [k N ] TIME [s] EXCITING FORCE IN TIME frequence 10 Hz 20 Hz 30 Hz Fo [kN] 0.75 0.750.75 1.5 1.5 1.5 2.25 2.25 2.25 Figure 2. Exiting force in time The damping forces gi(x, ẋn−1, θ1)ẋ = S iσg(x, ẋ, θi)ẋ are generally nonlinear and variable with temperature. The lin- ear law σg = (σ0 + σ1(θi − θk))ẋ gives good results at small increase of sam- ple temperature ∆θ ≤ 100C. The bet- ter function for the greater interval of temperature θ ∈ (300, 600) is σg = σg0(1−B1+B1/(1+ (θ−θm)2/γ)). The theoretical model of thermo- mechanical processes was verified by comparison with measured temperature response on dynamic loading. 0 50 100 150 200 250 300 28 29 30 31 32 33 34 35 36 T E M P E R A T U R E [ C ] TIME [s] THETA1cal THETA1exp THETA2cal THETA2exp Figure 3. Comparison of numerical and measured data The sample was loaded by harmonic force with frequency 10; 20; 30 Hz and various series of amplitudes a) 0.25; 0.5; 0.75 kN, b) 0.5; 1; 1.5 kN; c) 0.75; 1.5; 2.25 kN and d) 1; 2; 3 kN. The pre-stress was Fst = 6kN. The load- ing consists of 9 stages characterized by combination of 3 frequencies and 3 amplitudes of force. The whole loading process lasts 300 s. An example of the loading by serie c) is in Figure 2. The temperatures θ1 inside the sam- ple and θ2 near the surface increase ac- cording to the amplitude and frequency of loading. When the sample is loaded by the serie b), these temperatures are 406 PEŠEK, PU ◦ ST shown in Figure 3. Thick lines are values θ1exp, θ2exp gained by experiment. Thin lines, labelled θ1cal, θ2cal , describe the course of temperatures obtained by calculation using the proposed mathematical model. Qualitative agreement is seen. Similar good agreements of measured and calculated values were reached also in other cases. 3. Modal analysis of the tram wheel The modal analysis of conservative numerical wheel model can be described as generalized eigen-value problem as (K + λiM)xi = 0 where stiffness K, mass M are symmetric matrices, λi and xi represent set of eigenvalues and their corresponding eigenvectors, respectively. The ith eigenvalue can be expressed as λi = −ω2i , where ωi is so-called eigenfrequency. The modal analysis of the wheel was numerically solved by Lanczos’s method. The material constants were for rubber layer: Er = 0.522 108Pa, ρr = 1357.4kg/m3, Poisson’s constant µr = 0.45; for steel parts: Es = 2.1 1011Pa, ρs = 7830kg/m3, µs = 0.3. The boundary conditions of the model simulate free support of the wheel. In Table 1 there are results of the selected computed eigenfrequencies and corresponding mode shapes of the whole wheel (W) and its separate steel parts, i.e. rim and disk. The computed shapes of the rim with order numbers (R1..R6), the disk (D1..D5) and the wheel (W5..W9) are put together, if possible, to groups, which are designated by the shape numbers. Each group has a governing mode shape of the whole wheel and the modes of rim and disk that contribute to this governing mode. For clarity, each mode is also described by number of radial (nr) and circumferential (nc) node lines of the disk (nrd, ncd) and the rim (nrr). Since axis symmetry of the problem the double eigenvalues appear for some modes and therefore for these modes there are two forms in the Table 1, e.g. W5, W6 for the number 5. The 5th, 6th and 7th shapes of the wheel and the corresponding group eigen- modes of the separate wheel steel components are drawn in Figure 4. On the basis of our numerical-experimental investigation it was found that the coupling between the steel disk and rim of the wheel is weak due to small acoustic impedance of the rubber layer. It arises different kinds of vibration eigenmodes such as where dominates vibration either of the rim or disk or both these parts vibrate in phase or antiphase and we call this mode “whole” mode. In case when the eigenfrequencies of the disk and rim are close each other and the numbers of node lines of the corresponding eigenmodes are the same there is fulfilled a condition for arising the “ whole” eigenmodes (the 5th and 9th shapes). Sometimes EXPER. AND NUMER. ASSESSMENT OF VIBRO-ACOUSTIC 407 TABLE I. Grouping of mode shapes of the wheel (W), rim (R) and disk (D) no. order no. eigenfreq. [Hz] shape description R1,R2 472 nrr=2, flexural-torsion mode 5 W5,W6 503 nrr=2,nrd=2,“ whole” mode, in-phase motion D1,D2 504 nrd=2 R3,R4 524 nrr=2 6 W9,W10 684 nrr=2,nrd=0 R3,R4 524 nrr=2 7 W11,W12 846 nrr=0,nrd=2 D1,D2 504 nrd=2 8 W13 1055 nrr=0,nrd=0,ncd=1, umbrella mode D3 909 nrd=0,ncd=1, umbrella mode 9 W14,W15 1314 nrr=3,nrd=3 D4,D5 1131 nrd=3 R5,R6 1303 nrr=3, flexural-torsion mode it can be coupled even more modes together. It is for example case of the 5th shape where the rim has close eigenmodes axial and torsion-axial with tilting of cross-section. Figure 4. The 5th, 6th and 7th shapes of the whole wheel (e,d,f) and their corresponding eigenmodes of the disk (c) and rim (a,b) of the wheel In case of the “ whole” mode the eigenfrequency does not change much from the contributed fre- quencies of the disk and rim since they do not influence much each other. On the other hand the changes of eigenfrequencies with dominant vibration of one sep- arate part are substantial due to interaction of both parts. One part vibrates and the other is at still. The still part plays role of rigid body having both additional mass and stiffness and that causes the shift of eigen- frequency (the 7th shape - the disk vibrates, the rim is still). These eigenmodes, however, do not change much, i.e. they keep number of node lines as corre- sponding eigenmodes of separate parts. When there are bigger differences of the eigenfrequencies of the modes with the same number of node lines then due to the 408 PEŠEK, PU ◦ ST interaction the “whole” modes with antiphase vibration of both parts occur (for case of higher shapes). In case of “ whole” resonances when both parts vibrate in phase the rubber layer is very weakly deformed and eigen damping is also small. In case of the “ whole” modes with antiphase vibration the large relative motion of both part arises what causes substantial deformation of rubber layer and the damping increases. 4. Conclusion The experimental-numerical method for ascertaining of rheological model of ther- momechanical properties of rubber-like material is elaborated. The measured de- formation and temperatures are in qualitative agreement with numerical results and enable to ascertain the first approximations of parameters in the general rheo- logical model. The role of the inner rubber isolation layer on the dynamic behavior of com- posed structure of the wheel is described. It concerns the description of the cou- pling of both steel parts by the layer and their mutual interaction. By numerical- experimental modal analysis of wheels was found that, so called, “whole” modes with antiphase vibration of the disk and rim has dominant deformation of rubber layer and by that the highest damping. On the contrary the “whole” modes with phase vibration of both steel parts have smallest relative motions and by that worse damping effect. This analysis can lead to prediction of the wheel dynamics and optimization of the damping effect of this layer. Acknowledgements This research has been solved in a frame of the grant GA CR no.101/05/2669. References Pešek, L., Pust, L., and Vanek, F. (2003)a Contribution to identification of thermo-mechanic inter- action at vibrating rubber-like materials, In Proc. IFAC Symphosium on System Identification, pp. 1–6, Rotterdam. Pešek, L., Vesely, J., and Podruh, J. (2003)b Identification of Tram Wheel with Vibro-dissipation Layer, In Proc. Int. Conf. ICOVP, pp. 1–8, Liberec, Czech Rep. 409 409–414. CHARGE SIMULATION METHOD APPLIED TO VIBRATION PROBLEMS J. Raamachandran Indian Institute of Technology Madras, Chennai, Tamil Nadu 600 036, India Abstract. Charge simulation method is a new boundary type method in which the solution is a linear combination of fundamental solutions with unknown constants. Green’s functions are used as the fundamental solution. In this paper the use of this method is described to solve vibration problem especially plate vibration after explaining the method using a static problem for better understanding. Results are given for various cases and compared with existing solutions. Key words: charge simulation, fundamental solution, pole and field points, distance ratio 1. Introduction There has been several approximate methods available today to solve a structural problem and perhaps the latest one in that series is the charge simulation method. It may be quite appropriate to have an overview of the history of approximate methods so that the relative merits of charge simulation method can be better appreciated. An exact solution which on any day is most preferable is one which satisfies the governing differential (the domain) equation and all the boundary conditions exactly. Such a solution will also make the strain energy a minimum one. However, such a solution becomes very difficult to obtain when the problem becomes very complicated either by way of geometry or loading or boundary conditions. With advancement of engineering technology such problems were becoming more and more common place and a rigorous exact solution were be- coming extremely difficult to obtain. The cost involved was rather prohibitive. Hence, an approximate solution was sought for. The earliest approximate method was the energy method. This method is simply based on the concept stable equilib- rium obtained from physics. An object will be in stable equilibrium if its potential energy is either nil or minimum. This concept is applied to physical problem and an assumed approximate solution is forced to make the energy a minimum one by properly choosing the constants. An extension of this is the Rayleigh-Ritz method the main object of which is to replace the problem of finding the maxima and minima of integrals by that of finding maxima and minima of functions of several variables. The Rayleigh-Ritz method was very popular among engineers since this method relaxes satisfying the boundary condition! Well, in this method one need Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 410 RAAMACHANDRAN to satisfy only the geometric conditions but the natural or force boundary condi- tions need not be satisfied. Such a relaxation was very attractive since it becomes very difficult if not impossible to find an approximate functions satisfying force boundary conditions especially if the geometry is not a regular one. However, if one is able to assume an admissible function satisfying all the boundary conditions then this method gives a solution that is as good as an exact one. After Rayleigh Ritz method several approximate methods were developed and we cannot go through all the methods. The most important methods can all be grouped under one category namely, the weighted residual techniques. The Galerkin, collocation, method of moments, the finite-difference, the Trefftz, the finite element, the boun- dary element methods can all be grouped under this heading. Well, then what is weighted residual technique? We have mentioned earlier that an approximate method can give rise to error either within the domain or on the boundary or in both. All techniques to minimize these errors are called weighted residual methods. But depending on the manner as how the error is minimized we have different brand names! Of all these the Galerkin and the Trefftz methods need some explanation. In the Galerkin method the assumed solution and the weight- ing functions are one and the same. This method demands that all the boundary conditions are satisfied. Though the Rayleigh-Ritz and the Galerkin methods ap- pear to be different because of their starting point, yet fundamentally they are equivalent as evindenced by the fact that both these methods are used in the now well known and widely popular finite element method. Both the methods give the same stiffness matrix is a proof that these methods are philosophically equivalent! The Trefftz method is different from all other weighted residual method in that it is an extension of boundary integral method. Well, this method is the starting point for now known the boundary element method. Here, the domain equation is satisfied fully but the boundary conditions are not. The boundary element method uses the Greens’s function as the fundamental solution which is the solution of the given differential equation of the problem. This actually converts the finite domain problem into an infinite domain problem. Thus, the method is very well suited for off-shore problems, soil-structure interaction, acoustic radiation etc. There are three different types of boundary element method, viz., the direct, the indirect and the semi-direct method. We will not go through them in details here. Any one interested can refer Raamachandran (2000). Since the charge simulation method is an off-shoot of boundary element method it is expedient to see some of the relative merits of this method which will provide a basis for appreciating the charge simulation method. The advantages of boundary element method are (1) the dimensionality of the problem is reduced by one. Hence the data preparation is reduced. (2) the matrix to be solved is much smaller than the finite element method so the solution time is also reduced. (3) the required values at any point can be calculated directly (4) the final matrix to be solved is fully populated so no special technique is needed to invert the same. etc. The primary disadvantage CHARGE SIMULATION METHOD APPLIED TO VIBRATION PROBLEMS 411 of this method is that there is a mathematical difficulty when the source and sink coincide causing thereby singularity. A situation that one has to be very cautious. In the charge simulation method even this hurdle is eliminated and one obtain the solution very methodically. This method is similar to the direct boundary element method in which the physical variables are used directly as the unknowns. 2. Charge simulation method Charge simulation method is a new boundary type method in which the solution is a linear combination of fundamental solutions with unknown constants. Green’s functions are used as the fundamental solution. We will see a little later what are fundamental solutions. Now let us answer the question as to why the method is called the charge simulation method? A method has been devised by Singer et al. in 1974 for solving high voltage problems in electrical engineering. The basic concept used in their work is to replace the distributed charge of conductors and the polarization charges on the dielectric interfaces have been calculated so that their integrated effect satisfies the boundary conditions exactly at a selected num- ber of points on the boundary. Hence the name Charge Simulation Method. The method developed by Singer et al. was first applied to plane elasticity problem by Murashima, Nonaka and Nieda in 1983. The method has been applied to problems in plates and shells mostly by Raamachandran (1986, 1989,1991,1992). The first step in the CSM, as applied to electrical problems is to find out the equivalent simulated charges by the charge simultaneous equations. These charges are always located outside the domain (pole points or source points) where the field distribu- tion is calculated. The point lying on the boundary where the field distribution is prescribed is known as the field points. Once these lumped charges are determined the solution of potential and field strengths anywhere in the domain are computed by the analytical formulation and the superposition principle. 3. BEM vs CSM Before proceeding further we will now compare BEM and CSM to bring out the advantage of CSM. BEM CSM 1.The source is always on the boundary. 1. The source is always outside the So, singularity is a possibility to be checked. boundary. Hence no singularity is expected. 2.The BEM can be either a direct or 2. The CSM is always a direct method and indirect or semi direct method. hence very easy to monitor. 3. The source can be a point or line and 3. The source is always a point and hence numerical integration may be hence no numerical integration is needed a possibility. 412 RAAMACHANDRAN 4. Vibration of plates The governing equation of motion for the transverse displacement of the plate is given by D∇4w = −ρ∂ 2w ∂t2 (1) As we are considering only free vibration, w is taken as w = W(x, y)e−iωt (2) Substituting eqn.(2) into eqn.(1) we get D∇4W = ρω2W (3) The above equation can be written as L{W(x, y)} = (∇2 + λ2)(∇2 − λ2)W(x, y) = 0 (4) where, λ4 = ρω2 D (5) Now, the fundamental solution of eqn.(4) written as L[φ(x, y; xi, yi; λ)] = (∇4 − λ4)φ(x, y; xi, yi; λ) = δ(x, y; xi, yi) (6) is given by φ(x, y; xi, yi; λ) = − i 8λ2 [H10(λr) − H 1 0(iλr)] (7) where, δ= Dirac delta function (xi, yi) = source points r = distance between source and field points The Hankel function is complex function which is a linear combination of Bessel functions of the first and second kinds. Sometimes it is also called as Bessel func- tion of the third kind. To apply the CSM, eqn.(7) is used to assume an expression for W(x, y) as W(x, y) = 2N∑ i AiG(x j, y j; xi, yi; λ) ( j = 1, 2...N) (8) CHARGE SIMULATION METHOD APPLIED TO VIBRATION PROBLEMS 413 That is, W = − 2N∑ i iAi 8λ2 {H10[λ{(x j − xi) 2 + (y j − yi)2} 1 2 ] − H10[iλ{(x j − xi) 2 + (y j − yi)2} 1 2 ]} (9) Here, Ai’s are unknowns to be determined from the boundary conditions. (xi, yi) are poles outside domain and hence the solution eqn.(9) satisfies the partial differential equation inside the plate domain. Both simply supported and clamped boundary conditions are considered. But the more advantage of the method lies in that mixed boundary conditions can be handled as easily as the regular conditions. Now, the boundary is divided into N points, as shown in Fig. 1. As there are two conditions at each point, 2N equations for the 2N unknowns in Ai are obtained. Substituting eqn.(9) into the boundary conditions yields a 2N x 2N algebraic equation of the form [B(φ)]{A} = 0 (10) For nontrivial solution the determinant of eqn.(10) must vanish. That is det[B(φ)] = 0 (11) Figure 1. Scheme of discretisation As it is difficult to obtain true roots in numerical computation, one must evalu- ate the local minima of the determinant. As the eigen-values are non-negative and the determinant is complex the minima must be evaluated in the sense of absolute 414 RAAMACHANDRAN TABLE I. Fundamental frequency parameter for various plates No. Plate λCS M λLeissa,1993 1 Clamped circular plate 9.1847 10.2158 2 Simply supported circular plate 4.468 4.977 3 Clamped elliptical plate 15.291 16.025 4 Simply supported square plate 17.567 19.139 values. The above technique is applied to a few plate problems and the results are shown in Table 1. 5. Discussion We have seen broadly what is the charge simulation method and how it is used in solving vibration problems of plates. As shown here, this method is really a very simple and easy method compared to the FEM and BEM. There are however, few limitations exist that are to be looked into. The distance ratio ( ratio of distance to pole point to distance to field point) is not yet standardized. It is arbitrary and one has to make several attempts to get a good value. From experience a ratio between 2 and 7 gives good results. References Leissa, A. W. (1993) Vibration of Plates, The Acoustical Society of America. Murashima, S., Nonaka, Y., Nieda, H. (1983) The Charge Simulation Method Applied to Two Dimensional Elasticity in Boundary Elements, Published by Computational Mechanics, Southampton. Raamachandran, J. (1986) Large Deflection of Plates Using Charge Simulation Method, Proc. Int. Conf. on Computational Mechanics ICCM 86, Tokyo, Japan 1, 103-108. Raamachandran, J., Bhaskara Reddy, G. C. (1989) Bending of Circular Plates Supported at Number of Points, J. Engg. Mech., ASCE 115, 437-441. Raamachandran, J. (1991) Vibration of Plates Using Charge Simulation Method, Computers and Structures 2, 247-249. Raamachandran, J. (1992) Bending of Plates with Varying Thickness by Charge Simulation Method, Engineering Analysis with Boundary Elements 10, 143-145. Raamachandran, J. (2000) Boundary and Finite Element Method, Narosa Publishing House. Singer, H., Steinbigler, H., Weiss, P. (1974) A Charge Simulation Method for the Calculation of High Voltage Fields, IEEE PAS-93, 1660-1668. Timoshenko, S., Krifger, W. (1959) Theory of plates and shells, 2nd Ed., McGraw Hill, New York. 415 415–427. NONLINEAR STABILITY AND DYNAMICS OF LAMINATED COMPOSITE PLATES AND SHELLS S. Singh,1 B. P. Patel,1 A. Sharma,2 K. K. Shukla3 and Y. Nath1 ∗ 1Applied Mechanics Department, I.I.T. Delhi, New Delhi 110 016, India 2 Thapar Institute of Engineering and Technology, Patiala 147 004, India 3Motilal Nehru National Institute of Technology, Allahabad 211 004, India Abstract. The stability and dynamic characteristics of laminated plates (rectangular, sector) and shells (cylindrical/conical/circular/noncircular) are studied using analytical solutions employing chebyshev polynomials and finite element method. The detailed parametric studies are carried out to highlight the effects of geometrical parameters, lamination schemes, loading and boundary conditions on the stability and dynamic response characteristics of plates and shells undergoing large deformation. Few results are presented. Key words: stability, static, dynamic, plates, shells, thermo-elastic 1. Introduction The advanced laminated composite materials combine a number of unique prop- erties such as high strength-to- and stiffness-to-weight ratios, high fatigue resis- tance, high damping, low coefficient of thermal expansion especially in the fibre direction and tailor-making capability of these characteristics through appropri- ate lamination schemes for achieving minimal weight. The structural elements such as plates/shells of various geometrical shapes made up of such materials are increasingly being used by modern engineering industry such as aerospace, mechanical, petrochemical, marine, nuclear and power, etc. The conical shells are used as transition elements between cylinders of different diameter and/or end closures in various engineering applications such as tanks and pressure vessels, missiles and spacecraft, submarines etc. Either due to the fabrication process or special design requirements like external shape/internal storage compartments in aerospace/spacecraft structural systems, the cylindrical/conical shells may have noncircular cross-section. These structural elements may be subjected to severe static/dynamic thermo- mechanical loading conditions which may cause stability problems and dynamic ∗ Corresponding author: Email:
[email protected]; Fax. +91-11-26581119. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 416 SINGH, PATEL, SHARMA, SHUKLA, NATH Figure 1. Coordinate system and cross-sectional details of the noncircular conical shell response leading to transverse deflections of the order of plate/shell thickness or higher, and having in view the utilization of reserve strength in the design of flight vehicle structures, the study of thermo-mechanical stability/dynamic re- sponse characteristics of plates and shells with different geometrical configura- tions such as square/rectangular/skew/sector/annular plates and circular/noncircular cylindrical/conical shells based on the nonlinear analysis gains importance. Fur- ther the critical loads estimated using the linear theory (eigenvalue approach) may differ significantly and the buckling in the classical sense does not exist due to material coupling in laminated plates and material/curvature coupling in shells. Nonlinear dynamic behaviour of laminated sector plates, critical loads of noncircular cylindrical/conical shells and many other associated problems due to dynamic thermo-mechanical loading environment are not adequately addressed in the open literature (Noor and Burton, 1992; Thornton, 1993; Argyris and Tenek, 1997; Soldatos, 1999; Amabili and Paidoussis, 2003). Here, the stability and dynamic analyses of laminated plates and shells are car- ried out using analytical solutions employing Chebyshev polynomials and finite element method. The studies to highlight the influences of geometrical parame- ters, lamination schemes, loading and boundary conditions on the stability and dynamic response characteristics of plates and shells undergoing moderately large elastic deformation are presented. 2. Formulation and governing equations A laminated composite noncircular conical shell is considered with the co-ordinates x along the meridional direction, y along the circumferential direction and z along the thickness direction having origin at the midsurface of the shell as shown in Figure 1. Using Reissner-Mindlin first order shear deformation theory, the dis- placements u, v, w at a point (x, y, z) are expressed as functions of the mid-plane displacements u0, v0 and w0, and the independent rotations θx and θy of the normal in the xz and yz planes, respectively, as u(x, y, z) = u0(x, y) + zθx(x, y) NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 417 v(x, y, z) = v0(x, y) + zθy(x, y) w(x, y, z) = w0(x, y) (1) Considering moderately large deformations, Green’s strains can be written in terms of mid-plane deformations and rotations as {ε} = { εLp 0 } + { zεb εs } + { εNLp 0 } (2) where, the membrane strains {εLp}, bending strains {εb}, transverse shear strains {εs} and nonlinear in-plane strains {εNLp } in the Eq. (2) are written as (Kraus, 1976) {εLp} = ∂u0 ∂x u0 sin φ r + ∂v0 ∂y + w0 cos φ r ∂u0 ∂y − v0 sin φ r + ∂v0 ∂x ; {εb} = ∂θx ∂x θx sin φ r + ∂θy ∂y ∂θx ∂y + ∂θy ∂x − θy sin φ r ; {εs} = { θx + ∂w0 ∂x θy + ∂w0 ∂y − v0 cos φ r } (3) {εNLp } = 1 2 [( ∂u0 ∂x ) 2 + (∂v0∂x ) 2 + (∂w0∂x ) 2] 1 2 [( ∂u0 ∂y − v0 sin φ r ) 2 + (∂v0∂y + u0 sin φ r + w0 cos φ r ) 2 + (∂w0∂y − v0 cos φ r ) 2] ∂u0 ∂x ( ∂u0 ∂y − v0 sin φ r ) + ∂v0 ∂x ( ∂v0 ∂y + u0 sin φ r + w0 cos φ r ) + ∂w0 ∂x ( ∂w0 ∂y − v0 cos φ r ) (4) where r and φ are the radius of curvature in the circumferential direction and semi-cone angle, respectively. The radius of curvature r is the function of x and y coordinates. If {N} represents the stress resultants (Nxx, Nyy, Nxy) and {M} the moment re- sultants (Mxx, Myy, Mxy), one can relate these to membrane strains {εp} = ({εLp} + {εNLp }) and bending strains {εb} through the constitutive relations as { {N} {M} } = [ [A] [B] [B] [D] ] { {εp} {εb} } − { {N̄} {M̄} } (5) where [A], [D] and [B] are extensional, bending and bending-extensional coupling stiffness coefficients matrices of the composite laminate. {N̄} and {M̄} are the thermal stress and moment resultants, respectively. Similarly, the transverse shear stress resultants {Q} representing the quantities (Qxz, Qyz) are related to the transverse shear strains {εs} through the constitutive relation as {Q} = [E]{εs} (6) where [E] is the transverse shear stiffness coefficients matrix of the laminate. 418 SINGH, PATEL, SHARMA, SHUKLA, NATH For a laminated shell of thickness h, consisting of N layers with stacking angles θi(i = 1, ..., N) and layer thicknesses hi(i = 1, ..., N), the necessary ex- pressions to compute the elements of stiffness coefficient matrices ([A], [D], [B] and [E]) and thermal stress/moment resultants ({N̄} and {M̄}) used here are taken from (Jones, 1975). The potential energy functional U1(δ) (due to strain energy and transverse load) is given by, U1(δ) = 12 ∫ A [ { εp εb }T [ A B B D ] { εp εb } + {εs}T [E]{εs} − { εp εb }T { N̄ M̄ } ]dA − ∫ A qw0dA (7) where δ is the vector of degree of freedom associated to the displacement field in a numerical discretisation and q is the applied external pressure load. The potential energy U2(δ) due to initial state of in-plane stress resultants {N0} = {N0xx N0yy N0xy}T is written as U2(δ) = ∫ A {εNLp }T {N0}dA (8) Following the procedure due to (Rajasekaran and Murray, 1973), the total potential energy functional U(δ)[= U1(δ) + U2(δ)] can be expressed as U(δ) = {δ}T [(1/2)[[K] − [K∆T] + [KG]] + (1/6)[N1(δ)] + (1/12)[N2(δ)]]{δ} −{δ}T {FM} − {δ}T {FT } (9) where [K] is the linear stiffness matrix, [N1] and [N2] are nonlinear stiffness matrices linearly and quadratically dependent on the field variables, respectively. [K∆T] and [KG] are the geometric stiffness matrices due to thermal and initial stress resultants. {FM} and {FT } are mechanical and thermal load vectors. The kinetic energy of the shell is given by T (δ) = 1 2 ∫ A [p(u̇0 2 + v̇0 2 + ẇ0 2) + I(β̇s 2 + β̇θ 2)]dA (10) where p = ∑N i=1 ∫ hi+1 hi ρidz, I = ∑N i=1 ∫ hi+1 hi ρiz2dz and ρi is the mass density of the ith layer. hi and hi+1 are the z coordinate of the inner and outer surfaces of the ith layer. The dot over the variable denotes the derivative with respect to time. Substituting Eqs. (9) and (10) in Langrange’s equation of motion, the govern- ing equations of the shell are obtained as: [M]{δ̈} + [[K] − [K∆T] + [KG] + 1 2 [N1(δ)] + 1 3 [N2(δ)]]{δ} = {F} (11) NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 419 where [M] is the mass matrix. The governing equation (11) can be employed to study the linear/nonlinear static/dynamic response and eigenvalue buckling analyses by neglecting the appropriate terms. The circular shells are analyzed using semi-analytical axisymmetric finite el- ement whereas the noncircular shells are analyzed using two-dimensional (2D) shell finite element developed in house. The presence of asymmetric perturbation in the form of small magnitude lateral pressure spatially proportional to the linear buckling mode shape is assumed to initiate the bifurcation of the shell deformation from axisymmetric to asymmetric one in the nonlinear analysis. The rectangular and sector plate geometries are analyzed using Chebyshev technique (Nath and Kumar, 1998, 2000; Nath and Shukla, 2001a, 2001b). 3. Results and discussion The influences of geometrical parameters, lamination schemes, loading and boun- dary conditions on the stability and dynamic response characteristics of plates and shells undergoing large deformation.The dynamic critical load of laminated plate subjected to biaxial in-plane compression and thermal loading is presented in Table I. It can observed from this table that the critical load decreases due TABLE I. Non-dimensional Dynamic Critical Load for Laminated Plates Non-dimensional Critical Load (λcr) Lamination Boundary λT = 0, λT = 0.5, λT = 0, λT = 0.5, scheme conditions Nx � 0, Nx � 0, Nx � 0, Nx � 0 , Ny = 0 Ny = 0 Ny � 0 Ny � 0 SSSS 13.9 10.04 6.95 4.68 00/900/00/900/00 CCCS 24.45 22.83 13.0 9.37 CSCS 19.96 17.75 10.12 8.82 450/ − 450/450/ − 450/450 CSCS 19.19 17.77 10.59 9.26 00/150/300/450/600 CSCS 9.48 6.57 6.57 5.67 to transient transverse temperature gradient and the decrease in the critical load varies from approximately 6% to 33% depending on the geometric and material configurations. The study on the buckling and free vibration analyses of moderately thick sector plates is presented. The effects of orthotropy, number of layers, annularity (α), thickness ratio (λ), sector angle (Θ), rotary inertia and boundary conditions are studied. The boundary conditions considered here are various combinations of 420 SINGH, PATEL, SHARMA, SHUKLA, NATH simply supported (S ), clamped (C) and free (F) edges. The geometrical details of the sector plate are shown in Figure 2. Figure 2. Geometry of the Sector Plate The results for natural frequency and buckling loads of sector plate are shown in Table II and III, respectively. The effects of thickness ratio λ(= hr0 ) and annular- ity α on the frequencies are shown in Table II. Here, when α is increased from 0.1 to 0.5, the frequencies increase in the absolute sense after removal of the factor (1 − α2) from the non-dimensionalisation factor {(1 − α2)r20 √ ρ Eθh2 }. These results also show that the non-dimensional frequencies decrease with increasing thick- ness parameter λ, but the absolute frequencies actually increase with increasing thickness. It can be observed from Table III that the non-dimensional critical buckling loads decrease with increasing λ and α but the absolute critical buckling loads increase with increasing λ and α. The critical buckling load always decreases with increasing span in terms of Θ. This decrease is largest for the most flexible edge condition SSSS and smallest for the stiffest edge condition CCCC. The nonlinear thermoelstic pre- and post-buckling characteristics of cross- and angle-ply laminated truncated circular cylindrical/conical shells analyzed, reveal that the initially along the prebuckling path the shell deformation grows in the axisymmetric mode with increase in temperature. The increase in the temperature beyond certain value (bifurcation temperature), the shell deformation reveals bi- furcation from axisymmetric to asymmetric pattern. The bifurcation temperature decreases significantly with the increase in the semi-cone angle of the conical NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 421 TABLE II. Effect of λ and α on non-dimensional natural frequencies {ω(1 − α2)r20 √ ρ/Eθh2} of sector plates with combinations of clamped, simply supported and free edges, (Θ = π/3, n = 10, (0/900)5, Er/Eθ = 40) Mode Sequence number Boundary Condition λ α 1 2 3 4 5 6 0.2 0.1 16.912 25.827 26.103 34.996 35.719 36.125 0.5 6.169 8.335 10.952 11.103 12.469 13.916 CCCC 0.1 0.1 31.063 48.085 48.540 66.492 67.688 67.778 0.5 11.611 15.621 20.843 21.098 23.686 26.774 0.01 0.1 70.061 131.034 132.817 209.955 213.194 213.588 0.5 30.413 42.818 66.041 75.149 82.599 95.771 0.2 0.1 12.363 15.256 17.874 22.997 25.194 25.666 0.5 4.526 5.690 6.190 7.847 8.122 10.841 SSSS 0.1 0.1 23.542 24.726 35.753 43.973 44.807 45.994 0.5 9.048 9.240 12.380 14.084 15.702 20.102 0.01 0.1 32.950 80.714 81.747 146.511 149.461 152.020 0.5 14.103 24.953 45.986 50.504 56.406 72.422 0.2 0.1 12.363 15.321 22.997 23.409 25.260 25.672 0.5 5.733 6.190 8.156 10.851 10.974 12.224 CSSS 0.1 0.1 23.713 24.726 44.256 44.819 45.994 46.880 0.5 9.734 12.380 14.413 20.275 20.332 23.322 0.01 0.1 32.957 81.140 81.747 149.005 149.461 152.020 0.5 19.447 27.324 46.263 61.078 65.585 72.422 0.2 0.1 12.363 15.828 22.997 25.280 25.971 33.719 0.5 5.958 6.190 8.297 10.852 11.076 12.224 CSCS 0.1 0.1 24.726 26.539 45.052 45.994 46.795 64.262 0.5 10.825 12.380 15.126 20.474 20.804 23.396 0.01 0.1 44.831 96.755 99.805 169.006 171.967 173.459 0.5 28.395 35.281 54.054 74.190 78.093 81.545 0.2 0.1 9.814 17.604 19.022 26.967 29.328 29.493 0.5 2.874 4.675 5.209 7.103 7.515 8.178 FCFC 0.1 0.1 16.125 31.069 32.864 48.827 52.400 53.757 0.5 4.705 8.237 9.361 14.204 14.913 15.066 0.01 0.1 26.729 65.853 67.176 126.189 127.602 130.000 0.5 7.917 18.436 20.707 35.562 38.880 38.945 0.2 0.1 6.537 7.475 8.273 9.086 17.392 18.741 0.5 2.681 4.177 4.784 4.877 6.003 7.566 SFSF 0.1 0.1 9.784 11.644 14.964 18.173 29.713 31.911 0.5 5.367 8.037 8.198 8.362 12.007 12.229 0.01 0.1 13.016 14.982 50.913 54.671 57.529 112.487 0.5 12.818 13.076 18.495 35.188 49.728 50.084 422 SINGH, PATEL, SHARMA, SHUKLA, NATH TABLE III. Effect of λ , α and Θ on non-dimensional isotropic critical buckling load {Ncr(1 − α2)r20/Eθh3} of sector plates with combinations of clamped and simply supported edges, (n=10, (0/900)5, Er/Eθ = 40) Boundary Condition λ α Θ CCCC CSCS SSSS CSSS 0.2 0.1 π/6 8.905 8.623 8.275 8.348 0.1 π/3 8.494 7.817 7.118 7.362 0.1 π/2 8.145 7.163 6.298 6.648 0.1 2π/3 7.861 6.655 5.785 6.155 0.5 π/6 2.768 2.661 2.552 2.577 0.5 π/3 2.681 2.481 2.345 2.397 0.5 π/2 2.631 2.404 2.312 2.347 0.5 2π/3 2.606 2.371 2.321 2.331 0.1 0.1 π/6 32.209 29.719 25.258 26.974 0.1 π/3 27.481 23.027 16.975 19.716 0.1 π/2 24.312 18.708 12.915 15.542 0.1 2π/3 22.309 15.894 10.984 13.132 0.5 π/6 10.103 9.105 7.721 8.245 0.5 π/3 9.185 7.283 6.198 6.648 0.5 π/2 8.816 6.664 6.179 6.316 0.5 2π/3 8.610 6.481 6.198 6.261 0.01 0.1 π/6 274.543 202.934 82.371 132.804 0.1 π/3 114.687 73.694 32.817 50.205 0.1 π/2 75.696 44.035 20.281 30.982 0.1 2π/3 60.185 32.014 16.052 22.623 0.5 π/6 85.671 63.041 24.010 40.371 0.5 π/3 51.997 21.095 14.305 16.398 0.5 π/2 47.234 16.275 14.930 15.064 0.5 2π/3 45.521 15.800 14.305 14.825 shells while smaller diameter is kept fixed as shown in Figure 3. It is also ob- served that for certain moderately thick and short conical shells the bifurcation of the equilibrium path may not occur. The critical bifurcation temperature pre- dicted from nonlinear analysis are higher than those evaluated using eigenvalue approach for moderately thick/long and thin shells whereas for two-layered thick and short shells the eigenvalue may overestimate the critical temperature. The per- centage difference in the critical temperature predicted using the two approaches, NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 423 in general, increases with semi-cone angle and reduction in the bending-stretching coupling due to lamination scheme. The analysis of the angle-ply shells reveals that the combined influence of meridional and hoop stress resultants leads to the onset of asymmetric thermoelastic buckling at lowest critical temperature for shells with 150 and 750 ply-angles and axisymmetric buckling/deformation for shells with 450 ply-angle. Figure 3. Nonlinear pre- and post-buckling characteristics of conical shells: (a) (00/900), (b)(150/ − 150) The nonlinear thermoelastic stability characteristics of laminated oval cylin- drical/conical shells subjected to uniform temperature rise reveal that circular shells exhibit distinct bifurcation point whereas the fairly noncircular shells reveal a smooth transition form prebuckling to postbuckling equilibrium path as depicted in Figure 4. The influence of thermal degradation of material properties on the buckling and post-buckling characteristics is brought out for the conical and cylindrical shells. The comparison of thermoelastic pre- and post-buckling characteristics of shells with temperature dependent (TD) material properties is made with those considering constant material properties (TID), and the behavior is found to be significantly different at higher temperature range depending upon the material properties and shell parameters. The typical response behavior is shown in Fig- ure 5. For the shell parameters yielding lower bifurcation/buckling temperature, the influence of thermal degradation of material properties is found to be less and hence its incorporation for the response prediction in such situations may not be essential Figure 5(b). The modal participation of axisymmetric and asymmetric displacement com- ponents in the dynamic response of conical shell subjected to harmonic modal 424 SINGH, PATEL, SHARMA, SHUKLA, NATH excitation is highlighted in Figure 6. It can be observed from this figure that the energy exchange among different modes results in beating and circumferentially travelling wave non-stationary response. The first four free flexural vibration frequencies of thermally stresses cylindri- cal shells in prebuckling and post-buckling equilibrium configuration are studied. The typical results showing the prebuckling and post-buckling equilibrium path along with the frequency parameter (Ω2) variation is highlighted in Figure 7. It can be seen from this figure that the Ω2 values decrease with increase in tem- perature up to the critical bifurcation point. After the bifurcation point Ω2 values pertaining to some modes may be negative indicating that the corresponding static equilibrium configuration is unstable shown as dotted lines in Figure 7(a). 4. Conclusions The stability and dynamic analyses of laminated plates and shells are carried out employing analytical and finite element methods. The postbuckling characteristics of conical/cylindrical circular/noncirculer shells reveal interesting behaviour. The nonlinear dynamic response of the shells shows circumferentially travelling wave non-stationary response with the participation of the companion mode. The sta- bility study of the postbuckling curves reveals that the thermal postbuckling path with symmetric variation of displacement along meridional direction is unstable while the one with anti-symmetric variation is of stable nature. Figure 4. Nonlinear thermoelastic response of cross-ply oval cylindrical shells with different ovality parameter (ξ) NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 425 Figure 5. Nonlinear response curves for cross-ply laminated simply-supported conical shells: (a) Material 1, L/r1 = 1, r1/h = 500, (b) Material 2, L/r1 = 1, r1/h = 500 Figure 6. Steady state response of conical shell with semi-cone angle=600, ωF/ωL = 0.81 426 SINGH, PATEL, SHARMA, SHUKLA, NATH Figure 7. Equilibrium configuration and variation of free flexural vibration frequencies about First (I) and second (II) equilibrium configurations for a cross-ply laminated heated simply-supported cylindrical shells (R/h = 200, L/R = 1, (00/900)4) of Composite Plates and Shells with strong Nonlinearities, Applied Mechanics Reviews, 50(5), 285-305. Jones R. M., 1975, Mechanics of Composite Materials, Mc-Graw Hill, New York. Kraus H., 1976, Thin Elastic Shells, John Wiley and Sons, New York. Nath Y. and Kumar S., 1998, Effect of Transverse Shear on Static and Dynamic Buckling of Anti- symmetric Laminated Polar Orthitropic Shallow Spherical Shells, Composite Structures, 40(1), 67-72. References Amabili M. and Paidoussis M. P., 2003, Review of Studies on Geometrically Nonlinear Vibra- tions and Dynamics of Circular Cylindrical Shells and Panels, with and without Fluid-Structure Interaction, Applied Mechanics Reviews, 56, 349-381. Argyris J. H. and Tenek L., 1997, Recent Advances in Computational Thermostructural Analysis NONLINEAR STABILITY AND DYNAMICS OF PLATES AND SHELLS 427 Nath Y. and Shukla K. K., 2001b, Nonlinear Transient Analysis of Moderately Thick Laminated Composite Plates, Journal of sound and vibration, 247(3), 509-526. Noor A. K. and Burton W. S., 1992, Computational Models for High-Temperature Multilayered Composite Plates and Shells, Applied Mechanics Reviews, 45(10), 419-446. Rajasekaran S. and Murray D. W., 1973, Incremental Finite Element Matrices, ASCE Journal of the Structural Division, 99, 2423-2438. Soldatos K. P., 1999, Mechanics of Cylindrical Shells with Noncircular Cross-Section: A survey, Applied Mechanics Reviews, 52, 237-274. Thornton E. A., 1993, Thermal Buckling of Plates and shells, Applied Mechanics Reviews, 46(10), 485-506. Nath Y. and Kumar S., 2000, Nonlinear Dynamic Response of Axisymmetric Thick Laminated Shallow Spherical Shells, International Journal of Non-linear Mechanics and Numerical Simulation, 1(3), 225-233. Nath Y. and Shukla K. K., 2001a, Analytical Solution for Buckling and Post-Buckling of Angle- Ply Laminated Plates Under Thermomechanical Loading, International Journal of Non-linear Mechanics, 36, 1097-1108. 429 429–436. ON THE STABILITY OF A VIBROISOLATION SYSTEM WITH MORE DEGREES OF FREEDOM Jan Michal and Antonin Skarolek Fakulty of machinery construction, Technical University in Liberec, Halkova 6,461 17, Liberec, Czech Republic Abstract. The spring suspension of an ambulance couch has three degrees of freedom. Its guide spacious mechanisms is parallelogram for vertical suspension; on the upper base of which is a double Cardan suspension with horizontal axes. The vibroisolation is realized by pneumatic springs with a position control. So that all proper frequencies might be lower than 1 Hz., it is necessary to equip the springs with additional volumes. There have been determined the conditions of the stability and the ill-effect of the position control has been proven. Key words: stability, suspension, vibroisolation, pneumatic springs 1. Introduction The vibroisolation system of a ambulance couch is realized by pneumatic springs and by hydraulic dampers. The locating mechanism is a parallelogram, on its upper base there is a double Cardan suspension (see Fig. 1). Figure 1. It is supposed, to elimi- nate a vertical translation and rotation around both horizon- tal axes of the carriage. The detailed deduction of the mo- tion equations of this system is in (Skliba et al., 2005) and the analysis of the possibility of the decreasing of the sus- pension stiffness is in (Skliba et al., 2005). This requirement leads mostly to the application of the additional volume at the pneumatic springs. The analy- sis made in (Skliba et al., 2005) and (Skliba et al., 2005) concern surely the uncontrolled vibroisolation system. Vibration Problems ICOVP 2005, c© 2007 Springer. ˘ Sivc̆a ′ k (eds.), . şE. Inan and A. Kırı Skliba, 430 The vibroisolation system is expected to be equipped with the control system. The requirements on the optimal transfer will be met. On the other hand side, it is well-known from both empiric experience and many theses that it is necessary to meet the stability condition: The steepness of the position control has an upper limit (see (Skliba, 1996)). Therefore it will be suitable to analyze the stability condition by the application of the PI control. 2. Formulation of the dynamic system We suppose, that the kinematic excitation has only one degree: vertical translation ζ(t). In the coordinate system connected with the carriage (θ + θ0) is the angle of the parallelogram levers with the loading surface and ϕ resp. ψ the angle deflec- tions of the first resp. second Cardane frame (see Fig. 1). When we assume, that the deflections ζ(t), θ, ϕ, ψ are small, we can make the general system, deduced in (Skliba et al., 2005), linear: A�̈q + B0�̇q + (C0 + C1(t)) �q = �E0 + �E1(t) (1) This system is deduced in (Skliba et al., 2005); we denote �q (ϑ, ϕ, ψ) vector of generally coordinates. The mass matrix is symmetrical and for its coefficients is valid: A11 = (mR + m4 + m5 + m6)R 2 + 4JRy A22 = J5y + J6y + m6(x 2 56 + z 2 56 + 2x56xT6 + 2z56zT6) (2) A12 = −R[m5(xT5 cosϑ0 + zT5 sinϑ0)+m6((xT6+ x56) cosϑ0+(zT6 + z56) sinϑ0)] A13 = m6R cosϑ0yT6, A23 = −D6xy − m6x56yT6, A33 = J6x At the same time it is obvious, that Ai j are the functions of masses and inertia moments of the individual members of the located mechanism, of its geometrical configuration, of parallelogram lever R, and of the coordinates of the mass cen- ters. Regarding to the general position of the human body on the coach, we also respekt the resultant centrifugal moment of the second frame. The stiffness matrix is diagonal and its members C011 = −(2mR + m4 + m5 + m6)gR sinϑ0 − 4r2pϑ n(pa + p4)S 204 V4 + p4S 14 + + 4p4S 04rpϑ sinϑ0 C022 = −(m5zT5 + m6z56 + m6zT6)g − r2pϕ1 n(pa + p51)S 205 V5 + p51S 15 − − r2pϕ2 n(pa + p52)S 205 V5 + p52S 15 (3) ŠKLlBA, SIV ÀK, SKAROLEK ′ Č ON THE STABILITY OF A VIBROISOLATION SYSTEM 431 C033 =−m6gzT6 − r2pψ1 n(pa + p61)S 206 V6 +p61S 16 −r2pψ2 n(pa + p62)S 206 V6 +p62S 16 are the functions of pressures, effective areas and volumes of pneumatic springs, of the coordinates of its suspension and coordinates of mass centers. The damping matrix is also symmetrical and its members B011 = 4b1ir 2 Tϑi cos 2 ϑ0, B022 = 2∑ j=1 b1 jr 2 Tϕ j, B033 = 2∑ j=1 b1 jr 2 Tψ j (4) are the functions of the gradients of the static velocity characteristic of dampers and of the arms of its suspensions. According to the presumption of the vertical kinematic excitation, the members of the matrix of the parametric excitation are defined by relation: C111 = −(m4 + m5 + m6 + 2mR)Rζ̈(t) sinϑ0 (5) the all another members of matrix C1(t) are zero. The vector of the external excitation is E11 = (m4 + m5 + m6 + 2mR)R cosϑ0ζ̈(t), E12 = 0, E13 = 0 (6) Vector �E0 is realized by the moments of the gravity forces and by moments of the pneumatic springs, which must be balanced in the equilibrium state. The condition: �E0 = �0 expresses this state: (m4 + m5 + m6 + 2mr) gR cosϑ0 = 4p4S 04(rpϑ cosϑ0) m5gxT5 + m6g(x56 + xT6) = 2∑ j=1 p5 jS 05rpϕ j (7) m6gyT6 = 2∑ j=1 p6 jS 06rpψ and along with the conditions of the forces balance − (m5 + m6)g + p51S 05 + p52S 05 = 0, −m6g + p61S 06 + p62S 06 = 0 (8) define the preliminary adjustment of the pressures in the springs. The position of the human body can be general. The applied valve springs DUNLOP have the 432 natural frequency 4Hz. The analysis of its application has demonstrated that the stiffness of the complete vibroisolation system must be reduced (see (Skliba et al., 2005)). On the Figures 2a . . . 2d there are the results of the numerical simulation of the analysed system for the aplicated springs DUNLOP (“real spring”) and for fictive springs with a reduced stiffness “imaginary springs”). The frequency of the kinematic excitation is 2 Hz - near the first natural frequency of a chasis. On the Figures 2a, 2b, 2c, there are the time responses of the relative displacements of the springs (The immediate lengths of the suspension spring of the palelogram (Fig. 2a) first Cardan frame (Fig. 2b) and second Cardan frame (Fig. 2c)). On the Fig. 2d there are the time responses of kinematic excitation and of the absolute displacement of the mass center of the second Cardan frame. The vibroisolation effect is obvious by a “imaginary spring”. Figure 2. ŠKLlBA, SIV ÀK, SKAROLEK ′ Č ON THE STABILITY OF A VIBROISOLATION SYSTEM 433 The sleeve springs FIRESTONE or CONTINENTAL with the forming piston will be applied. 3. Stability of controlled system In case that it is necessary to lover the stiffness of the suspensions by the ap- plication of additional volumes, we have to expand the dynamic system (1) by other equations for the equilibrium of the mass flows between the spring and the volume. The equations quoted in (Skliba et al., 2005), describe the case, when the vibroisolation system is not controlled in the dynamic state. In this case, the stability is reached by the adjusting of the hydraulic dampers. In the opposite case, we will assume that the control is realized by PI controllers. Then it is necessary to expand the above mentioned equations by corrective flows Gcor and join the equations for PI controllers: V4 ṗ4 + p4r4pS 4 (ϑ) ϑ̇ = −D41(p4 − p7) + Gcor7, V7 ṗ7 = D41(p4 − p7), V51 ṗ51 + p51r51pS 51 (ϕ) ϕ̇ = −D51(p51 − p8) + Gcor8, V8 ṗ8 = D51(p51 − p8), V52 ṗ52 + p52r52pS 52 (ϕ) ϕ̇ = −D52(p52 − p9) + Gcor9, V9 ṗ9 = D52(p52 − p9), V61 ṗ61 + p61r61pS 61 (ψ) ψ̇ = −D61(p61 − p10) + Gcor10, V10 ṗ10 = D61(p61 − p10), V62 ṗ62 + p62r62pS 62 (ψ) ψ̇ = −D62(p62 − p11) + Gcor11, V11 ṗ11 = D62(p62 − p11). (9) Figure 3. where we denote p7 . . . p11 the pressure in additional volume V7 . . .V11, D7 . . . D11 the resultant coefficients of the constriction between the spring and the volume and S (q) the effective areas of the springs. The control mass flows Gcori are the functions of in- let and outlet pressure and the constriction area S c which is proportional of the increment of the voltage U. For of the every control circuit (on the Fig. 3) is valid: KPq̇i + KIqi = U̇i, i = 7 . . . 11 (10) Gcori = Kout (Ui − U0i) √ pi (pi − pa), Ui > U0i 434 Gcori = Kin (Ui − U0i) √ pz (pz − pa), Ui < U0i (11) The resultant system is formed by 3 equations of the second order (1), 10 equations of the first order (2) and 5 equations of the first order (10). The solutions of the stability of the system of the 21 order by the application of Hurwitz conditions, proves to be complicated. We will try to determine the starting conditions for this task on the base of the following considerations: The control of the individual pneumatic springs is in fact autonomic. In the case, when the both cardan frame are lock, we go out from the first equations of the system (1), (2) and (10), but we must consider, that is now the function of variables ϑ, p4. Then we have: A11(ϑ)ϑ̈ + B11(ϑ)ϑ = Mpneu + Mgrav V4(ϑ) ṗ4 + p4S 4 (ϑ) r4p (ϑ) ϑ̇ = −D41 (p4 − p7) + Gcor7(U7, p4) (12) V7 ṗ7 = D41 (p4 − p7) KPϑ̇ + KIϑ = U̇7 where is: Mpneu = r4p (ϑ) S (ϑ) p4, Mgrav = (m4 + m5 + m6 + 2mR) gR cosϑ = MgR cosϑ in the equilibrium state belong ϑ = ϑo, p4 = p7 = p0, U = U0 we introduce to system (12) the linearised disturbance system. A11(ϑ0)ϑ̈ + B11(ϑ0)ϑ̇ + Cϑ − Ep4 = 0 Hϑ̇ + V4(ϑ0) ṗ4 + D41 p4 − D41 p7 + AcorU7 = 0 −D41 p7 + (V7 ṗ7 + D41 p7) = 0 (13) KPϑ̇ + KIϑ − U̇7 = 0 where for the shortening of the record we have introduced C = −∂ ( Mpneu + Mgrav ) ∂ϑ =−r2p (ϑ0) p dS dl ∣∣∣∣∣ l=l0 = drp (ϑ) dϑ ∣∣∣∣∣∣ ϑ=ϑ0 pS (ϑ0) + MgR sinϑ0 E = −∂Mpneu ∂p4 = −rp (ϑ0) S (ϑ0) (14) H = (p0 + pa) S (ϑ0) rp (ϑ0) Acor = 1 2 [ Kin √ pz(pz − p0) + Kout √ p0(p0 − pa) ] where its characteristic equation is of fifth degree 5∑ j=0 a5− jλ j = 0 and its coeffi- cients are given by the relations: a0 = A11V4V7 ŠKLlBA, SIV ÀK, SKAROLEK ′ Č ON THE STABILITY OF A VIBROISOLATION SYSTEM 435 a1 = B1V4V7 + A11D4 (V4 + V7) (15) a2 = CV4V7 + B1D41 (V4 + V7) + EHV7 a3 = CD4(V4 + V7) + E(AcorKpV7 + HD4) = a3(Kp) a4 = EAcorKPD4 + EAcorKIV7 = a4 (KP, KI) a5 = KI EAcorD4 = a5 (KI) All the coefficients are positive and the last Hurwitz conditions are: ∆3 = a1a2 − a0a3 = Φ1 (KP) > 0 (16) ∆4 = a3∆3 − a4a21 + a5a0a1 = Φ2 (KP, KI) > 0 Figure 4. It is obviously, that Φ1 is a linear form and Φ2 a quadratic form of its variables KP, KI . Both the forms limit the areas of parameters KP, KI , where the equilibrium state is stabile. It is obvious, the areas depend on additional volumes V , and on constriction coefficient D. For a choice alternative of parameters KP, KI , there are demonstrated the response of functions Φ1 (see Fig. 4a) and Φ2 (see Fig. 4b) 4. Conclusion It will be necessary to make an analogic analysis for the first and second Cardan frame and this to gain a preliminary orientation on the limits of stability before we start the general stability solution for complex system of the order twenty one. 436 Acknowledgment This work was supported by Czech Ministry of Education under project MSM 4674788501. References Skliba J. (1996) Problem jedne aplikace tlumene pneumaticke pruziny s polohovou regulaci, VII International congress on the theory of mechanisms, Liberec, Czech Republic. Skliba J., Prokop J., Sivcak M. (2005) Vibroisolation system of ambulance couch with three degrees of freedom, Engineering Mechanics 12. Skliba J., Prokop J., Sivcak M. (2005) About the possibility of the stiffness reducing of vibroiso- lation system, Proceedings of National Conference Engineering Mechanics, Svratka, Czech Republic. ŠKLlBA, SIV ÀK, SKAROLEK ′ Č 437 437–442. THE STABILITY AND VIBRATION OF CONICAL SHELLS COMPOSED OF SI3 N4 AND SUS304 UNDER AXIAL COMPRESSIVE LOAD 1 and Ali Deniz2 1 Department of Civil Engineering, SDU, Isparta, Turkey 2 Department of Mathematics, AKU, Uşak, Turkey Abstract. In this study, an analytic solution is provided for the stability and vibration behavior of truncated conical shells composed of Si3N4and SUS304 under the axial compressive load. The material properties of the FG conical shells are assumed to vary continuously through the thickness of the shell according to quadratic and inverse quadratic distribution of the volume fraction of the constituents. The fundamental relations and governing equations for thin conical shells composed of Si3N4and SUS304 is obtained by using Love’s shell theory and a solution is obtained by applying Galerkin’s method. The effects of the constituent volume fractions on the critical axial load and frequency parameter of the truncated conical shell are also elucidated. Key words: stability, vibration, conicall shells 1. Introduction Functionally graded materials (FGMs) are composite materials with gradient com- positional variation of the constituents (e.g. metallic and ceramic) from one sur- face of the material to the other, which results in continuously varying material properties. The concept of FGM, initially developed for super heat resistant ma- terials to be used in space planes or nuclear fusion reactors, is now of interest to designers of functional materials for energy conversion (Koizumi, 1993). In recent years, the stability and vibration behaviors of FGM shell structures have attracted increasing research effort, most of which have been limited to the vibration analysis (Pitakthapanaphong and Busso, 2002, Loy et al., 1999, and aeronautical structures, mainly as transition elements between cylinders of different diameters. For stability analysis of isotropic conical shells under axially compressive loads, a simple formula for critical axial loads of the moderately long shells was given in (Seide, 1956). In addition, there is much literature on stability and vibration analysis of the axially compressed isotropic and orthotropic conical shells (Tani and Yamaki, 1970, Irie et al., 1982 and Agamirov, 1990). Vibration Problems ICOVP 2005, c© 2007 Springer. Abdullah H. Sofiyev ¨ ¨ Patel et al., 2005 and Sofiyev, 2005). Conical shells are often used in marine (eds.), . şE. Inan and A. Kırı 438 SOFIYEV, DENİZ It is well known, due to the difficulties emerging during the solution and the- oretical analysis of the stability and vibration problems of conical shells, such compressive loads have not been studied sufficiently. From the literature survey, it is seen that few studies have been made for stability and vibration analyses of truncated conical shells under axial compressive load made of functionally graded materials. In this work, the stability and vibration of functionally graded truncated conical shells subjected to axial compressive load is studied. 2. Theoretical formulations Consider a conical shell made of FGM as shown in Fig. 1, R1 and R2 indicate the radii of the cone at its small and large ends, respectively γ denotes the semi-vertex angle of the cone, H is height of truncated conical shell, S axis lies along the Figure 1. The geometry and coordinate system of a truncated conical shell generator on the curvilin- ear reference surface of the cone, the θ axis lies in the circumferential direction on the reference surface of the cone and the z axis, being perpendicular to the plane of the first two axes, lies in the inwards normal direction of the cone, and the distances from the vertex to the small and large bases are S 1 and S 2 respectively. Here we consider an FGM shell, which is made from a mixture of Si3 N4 and SUS304. We assume that the composition is varied from the top to bottom surface, i.e. the top surface (z = h/2) of the shell is Si3 N4-rich whereas the bottom surface (z = −h/2) is SUS304-rich. The volume fractions Si3 N4 and SUS304 are Vc and Vm related by Vc + Vm = 1 (1) The compositional gradation of the Si3 N4 /SUS304 shell is defined by the volume fraction of the ceramic phase. Here, the following functions of Vc will be considered [2]: 1. Quadratic Vc = (z̄ + 0.5)2 2. Inverse Quadratic Vc = 1 − (0.5 − z̄)2 where z̄ = z/h is dimensionless thickness coordinate. The effective Young’s modulus E, Poisson ratio ν and the mass density ρ of an FGM shell can be written as E = (Et − Eb)Vc + Eb, ν = (νt − νb)Vc + νb, ρ = (ρt − ρb)Vc + ρb (2) VIBRATION OF CONICAL SHELLS 439 It is assumed that Et, Eb, νt, νb, ρt and ρb are functions of temperature and position. ν and ρ functions of temperature and position. As the properties and the independent variables s = s1eξ and F = F1e2ξ are considered, the dynamic stability and strain compatibility equations of FG conical shell will be as follows: L(ξ, ϕ, t) = A2e2ξ ∂4F1 ∂ξ4 − 4A2e2ξ ∂ 3F1 ∂ξ3 + ( 4A2 + s1eξ cot γ ) e2ξ ∂ 2F1 ∂ξ2 −s1 cot γe2ξ ∂F1∂ξ + A2e 2ξ ∂4F1 ∂ϕ4 + 2 (A1 − A5) e2ξ ∂ 4F1 ∂ξ2∂ϕ2 +4 (A5 − A1) e2ξ ∂ 3F1 ∂ξ∂ϕ2 + 2 (A1 − A5 + A2) e2ξ ∂ 2F1 ∂ϕ2 −A3 ∂ 4w ∂ϕ4 − 2 (A4 + A6) ∂ 4w ∂ξ2∂ϕ2 + 4 (A4 + A6) ∂ 3w ∂ξ∂ϕ2 −2 (A4 + A6 + A3) ∂ 2w ∂ϕ2 − A3 ∂ 4w ∂ξ4 + 4A3 ∂ 3w ∂ξ3 − 4A3 ∂ 2w ∂ξ2 −s21e 2ξN ( −∂w∂ξ + ∂2w ∂ξ2 ) − ρ1h∂ 2w ∂t2 = 0 (3) B1 ( ∂4F1 ∂ξ4 − 4∂ 3F1 ∂ξ3 + 4∂ 2F1 ∂ξ2 + ∂ 4F1 ∂ϕ4 ) + 2(B5 + B2) ( ∂4F1 ∂ξ2∂ϕ2 − 2 ∂ 3F1 ∂ξ∂ϕ2 ) + +2 (B5 + B2 + B1) ∂2F1 ∂ϕ2 = = e−2ξ B4 ∂ 4w ∂ϕ4 + 2 (B3 − B6) ∂ 4w ∂ξ2∂ϕ2 + 4 (B6 − B3) ∂ 3w ∂ξ∂ϕ2 + +2 (B3 + B4 − B6) ∂ 2w ∂ϕ2 + B4 ∂ 4w ∂ξ4 − 4B4 ∂ 3w ∂ξ3 + + ( 4B4 − s1eξ cot γ ) ∂2w ∂ξ2 + s1eξ cot γ∂w∂ξ (4) where ρ1 = 0.5∫ - 0.5 ρdz̄, Ai, Bi (i = 1 ÷ 6) are the parameters depending on the material properties and shell characteristics (see (Sofiyev, 2005). 3. The solution of equations Consider a truncated conical shell with simply supported edge conditions. The solutions for Eq.(4) take the following form (Agamirov, 1990): w = f1(t)e ξ sin β1ξ sin β2ϕ (5) where f1(t) is time dependent amplitude and the following definitions apply: β1 = mπ ξ0 , β2 = n sin γ , ξ0 = ln s2 s1 , ξ = ln s s1 (6) Function (5) satisfies the following geometrical boundary conditions (Agamirov, 1990): 440 SOFIYEV, DENİZ w = 0 at ξ = 0 and ξ = ξ0 (7) ∂2w ∂ξ2 − ∂w ∂ξ = 0 at ξ = 0 and ξ = ξ0 (8) Substituting Eq. (5) into Eq. (4) and can be obtained: F1 = ( K1 sin β1ξ + K2 cos β1ξ+ +K3e−ξ sin β1ξ + K4e−ξ cos β1ξ ) f1(t) sin β2ϕ (9) where the following definition apply: K1 = β1(β1u0+u2) u20+u 2 2 S 1 cot γ, K2 = β1(β1u2−u0) u20+u 2 2 S 1 cot γ, K3 = u4u1 u21+u 2 3 , K4 = − u4u3u21+u23 (10) where ui(i = 1 ... 4) are the parameters depending on the material properties and shell characteristics. Substituting Eqs.(5) and (9) into Eq.(3) and by applying Galerkin’s method, results in the following equation is obtained: ξ0∫ 0 2π sin γ∫ 0 L(ξ, ϕ, t)eξ sin(β1ξ) sin(β2ϕ)dξdϕ = 0 (11) After integrating Eq. (11), the following differential equation is expressed: ∂2 f1 ∂t2 + a10 ρ1ha9 [ J1a1 + J2a2 + J3a3 + J4a4 + a8 a10 − N ] f1(t) = 0 (12) where ai(i = 1 ... 10)are the parameters depending on the material properties and shell characteristics. At statically case, for the dimensionless critical axial load, the following ex- pression is obtained: N̄cr = J1a1 + J2a2 + J3a3 + J4a4 + a8 a10E1h (13) When N = 0, from Eq. (12) for the frequency of free vibration the following expression is obtained: ω = √ J1a1 + J2a2 + J3a3 + J4a4 + a8 ρ1ha9 , (14) VIBRATION OF CONICAL SHELLS 441 4. Numerical results and discussion To validate the analysis, results for simply supported homogeneous elastic trun- cated conical shells are compared with ref. (Agamirov, 1990) (see Table 1). Com- putations have been carried out for the following data: material properties are E = 2.1 × 105 MPa, ν = 0.3, ρ = 8000kg/m3 and shell properties are h = 5 × 10−4 m, H/R2 = 2. The comparisons show that the present results agreed well with those in the Ref. (Agamirov, 1990). TABLE I. Comparison of values of critical axial loads with those in (Agamirov, 1990) (Agamirov, 1990) Present study R2(m) γ 0.25 10o 0.25 20o N̄cr x103 ncr 1.42 19 1.56 18 N̄cr x103 (m,n) 1.415 (18, 18) 1.553 (20, 17) TABLE II. Variations of critical axial load and dimensionless frequency parameter ω̄ = ωR22ρ 1/2 1 /(hE 1/2 2 ) for FG conical shell composed of Si3 N4 /SUS304 with quadratic and inverse quadratic compositional profile and γ (H/R2 = 2 and R 2=0.2m) Si3N4 Si3N4/SUS304 Quadratic γ 100 150 200γ 100 150 200 N̄cr x103 (m,n) 1.769 (19,15) 1.875 (20,15) 1.935 (1,5) SUS304 1.140 (19,15) 1.209 (20,15) 1.248 1,5 N̄cr x103 (m,n) 1.359 (20,14) 1.440 (20,15) 1.487 (1,5) Si3N4/SUS304 Inv. Quadratic 1.528 (17,16) 1.619 (20,15) 1.667 (1,5) Si3N4 Quadratic γ 100 150 200γ 100 150 200 ω̄ (m,n) 36.99 (1,6) 41.92 (1,6) 49.72 (1,5) SUS304 29.70 (1,6) 33.66 (1,6) 39.09 (1,5) ω̄ (m,n) 32.34 (1,6) 36.77 (1,6) 43.59 (1,5) Inv. Quadratic 34.43 (1,6) 38.86 (1,6) 46.14 (1,5) In Table 2, variations of values of critical axial loads and dimensionless fre- quency parameters for FG conical shell composed of Si3 N4/SUS304 with quadratic and inverse quadratic compositional profile and semi vertex angle are presented. For all compositional profiles, when the semi-vertex angle γ increased, values of critical axial load and dimensionless frequency parameters also increased. Com- paring the values of critical axial loads for FG conical shell made by Si3N4with the values for FG conical shell composed of Si3N4/SUS304, the effect of compo- sitional profiles to the critical axial load for quadratic case is bigger as 23%. Comparing the values of dimensionless frequency parameters for FG conical shellmade by Si3N4with thevalues forFGconicalshellcomposedof Si3N4/SUS304, the effect of compositional profiles to the critical axial load for quadratic case is bigger as 12,57%. 442 SOFIYEV, DENİZ Comparing the values of critical axial loads for FG conical shell made by SUS304 with the values for FG conical shell composed of Si3N4/SUS304, the effect of compositional profiles to the critical axial load for inverse quadratic case is bigger as 25.4%. Comparing the values of dimensionless frequency parameters for FG con- ical shell made by SUS304with the values for FG conical shell composed of Si3N4/SUS304, the effect of compositional profiles to the critical axial load for inverse quadratic case is bigger as 13.74%. It is seen that the minimum values of critical axial loads and dimensionless frequency parameter are obtained for different wave numbers, but when γ = 20o, they are obtained for the same wave numbers. 5. Conclusions The stability and free vibration of functionally graded truncated conical shells subjected to axial compressive load is studied. The material properties of FG shell are varied in a state of quadratic and inverse quadratic law distribution along the thickness. Galerkin method is applied to determine the critical axial load and fre- quency parameter. The results show that the critical axial load and dimensionless frequency parameter are affected by the configurations of the constituent materials and the variation of the semi-vertex angle. Comparing results with those in the literature validates the present analysis. References Agamirov, V. L. (1990) Dynamic Problems of Nonlinear Shells Theory, Nauka, Moscow. Irie, T., Yamada, G., Kaneko, Y. (1982) Naturel frequences of truncated conical shell, J. Sound Vibr 92 447-453. Koizumi, M. (1993) The concept of FGM. Ceram Trans, Funct Gradient Mater 34 3-10. Loy, C. T., Lam, J. N., Reddy, J. N. (1999) Vibration of functionally graded cylindrical shells, Int J Mech Sci 41 309-324. Patel, B. P., Gupta, M. S., Loknath, M. S., Kadu, C. P. (2005) Free vibration analysis of a func- tionally graded elliptical cylindrical shells using higher-order theory, J Compos. Struct in press. Pitakthapanaphong, S., Busso, E. P. (2002) Self-consistent elasto-plastic stress solutions for func- tionally graded material systems subjected to thermal transients, J. Mech. and Phys. Solids 50 695-716. Seide, P. (1956) Axisymmetrical buckling of circular cones under axial compression, J. Appl. Mech. 23 626-628. Sofiyev, A. H. (2005) The stability of compositionally graded ceramic-metal cylindrical shells under a-periodic axial impulsive loading, J Compos Struct 69 257-67. Tani, J., Yamaki, Y. (1970) Buckling of truncated conical shells under axial compression, AIAA Journal 8 568-570. 443 443–448. DYNAMIC BUCKLING OF ELASTO-PLASTIC CYLINDRICAL SHELLS UNDER AXIAL LOAD 1 and E. Schnack2 1 2Institute of Solid Mechanics, Karlsruhe University, Karlsruhe, Germany Abstract. The dynamic buckling of a cylindrical shell subject to a uniform axial compression, which is a linear function of time, is examined within the framework of small strain elasto-plasticity. The material of the shell is incompressible and the effect of the elastic unloading is not taken into consideration. Initially, employing the deformation theory, the fundamental relations and Donnell type stability equations have been obtained. Finally, by using the Galerkin’s methods the closed form solutions are presented. Comparing the results with those in the literature validates the present analysis. Key words: dynamic buckling, elastoplasticity, shell 1. Introduction Plastic buckling of circular cylindrical shell under different loading has been sys- tematically studied since the late 1940s due to its importance in aerospace engi- neering. The basic information, and mathematical and technical formulation of elasto-plastic stability theory of the structures was described and summarized in (Ilyushin, 1947 and Hill, 1983). Numerous studies of the elasto-plastic stability of cylindrical shells under axial load have been conducted with various theoretical models and solution approaches (Gerard, 1957, Lee, 1962, Mao and Lu, 2001, and Ore and Durban, 1992). Some research works have been published concern- ing the dynamic buckling of elasto-plastic cylindrical shells under axial impact (Karagizova and Jones, 2002 and Wand and Tian, 2003). The dynamic buckling problem of the shells under compressive axial loading has been much less studied in contrast to the dynamic buckling under axial impact. In the buckling of the shells under axial compressive loads, the effect of stress wave propagation in middle surface is usually not taken into consideration. One of the most important studies about this subject is Wolmir (1975) that studied elasto-plastic behaviour of cylindrical panel, which has initial imperfection under dynamic axial compressive load by using the Runge-Kutta method. Vibration Problems ICOVP 2005, c© 2007 Springer. A. Sofiyev ¨ (eds.), . şE. Inan and A. Kırı Department of Civil Engineering, SDU, Isparta, Turkey 444 SOFIYEV, SCHNACK From the literature survey, it is seen that few studies have been made for elasto- plastic buckling analysis of circular cylindrical shells under linear depending of time compressive axial loads. In this study, the elasto-plastic behaviour of circular cylindrical shells subjected to time dependent axial compressive load varying as a linear function of time is studied, by using the small elasto-plastic deformation theory and the Ritz type variational method. 2. Fundamental relations and basic equations The cylindrical shell as shown in Fig. 1, is assumed to be thin and of length L, radius R and thickness h. The x and y axes are in the axial and tangential directions, respectively and the z-axis is directed radially positive inwards. The cylindrical shell subjected to axial compression, σx = σi, σxy = 0, σy = 0 (1) and axial load varying as a linear function of time in the form: T 0x = σxh = −T0t (2) where T 0x is pre-buckling membrane force, T0 is loading speed and t is time. The Figure 1. Geometry and the coordinate system of a cylindrical shell material of the shell is homogeneous isotropic and incompressible and the effect of the elas- tic unloading is not con- sidered. The intensity of the stress σi and the in- tensity of the strain εi of the material are given by σ2i = σ 2 x − σxσy + σ2y + 3σ2xy, ε2i = 4 9 [ ε2x − εxεy + ε2y + 2 ( εx + εy )2 + 3ε2xy ] (3) where σx, σy,σxy and εx, εy, εxy are stress and strain components, respectively. In agreement with the laws for the elasticity and plasticity of materials the stresses and strains are connected by the relations (Ilyushin, 1947): sx = σx − 0.5σy = Esεx, sy = σy − 0.5σx = Esεy, sxy = σxy = 2 3 Esεxy (4) where Es = σi/εi is the secant modulus. The variations of formulae (4) have the form δsx = Esδεx − (Es − Et) sx σi δεi, δsy = Esδεy − (Es − Et) sy σi δεi ELASTO-PLASTIC CYLINDRICAL SHELLS 445 δsxy = δσxy = 2 3 Esδεxy − (Es − Et) sxy σi δεi (5) where Et = dσi/dεi is the tangent modulus, δ is the symbol of variation and the following substitutions are introduced: δsx = δσx − 0.5δσy, δsy = δσy − 0.5δσx (6) Let the surface z = z0 represent the boundary between the regions, one of which is elastic after instability, the other plastic. For its determination we shall suppose that in the elastio-plastic zone, the plastic zone adjoins the shell surface z = +h/2 and the elastic zone which originates as a result of unloading adjons the surface z = −h/2. In the region of plastic strain and in the zone of loading (z > z0), the following relations can be used instead of (4): δsx = Esδεx − Es−EtEt sx ∏ (σ,δs) σ2i z0−z z0 δsy = Esδεy − Es−EtEt sy ∏ (σ,δs) σ2i z0−z z0 δsxy = 23 Esδεxy − Es−Et Et sxy ∏ (σ,δs) σ2i z0−z z0 (7) where σ and ε are stress and strain functions and the following definitions apply: Π (σ, δs) = Π (s, δσ) = σxδsx + σyδsy + 3σxyδsxy = Etσiδεi (8) In the region of elastic strain and in the zone of unloading (z < z0), the variation of the stress and strain relations are written in the form: δsx = Eδεx, δsy = Eδεy , δsxy = 2 3 Eδεxy (9) where E is the Young’s modulus. In the region of elasto-plastic strains, the stresses σx, σy, σxy have different expressions for z > z0 and for z < z0. Hence the integrals in the expressions for the forces and moments must be split into two parts. For the region z0 ≥ z ≥ −h/2 we take δsx, δsy, δsxy according to (7) and for the region h/2 ≥ z ≥ z0, according to (9). Finally, should no effect of elastic unloading take place ( z̄0 = 1, the following equations are obtained: Esh3 9 [( 1 4 + 3ϕts 4 ) ∂4w ∂x4 + 2 ∂4w ∂x2∂y2 + ∂4w ∂y4 ] +T0t ∂2w ∂x2 − h R ∂2Φ ∂x2 +ρh ∂2w ∂t2 = 0 (10) where the following definitions apply: ϕts = Et/Es 446 SOFIYEV, SCHNACK 3. Solution of the basic equations Assuming the cylindrical shell to have hinged supports at the ends, the solution of the equation set (10, 11) is sought in the following form: w = ξ (t) sin m1x R cos ny R , Φ = ζ (t) sin m1x R cos ny R (12) where m1 = mπR/L, m is the half wave length in the direction of the x-axis, n is the wave number in the direction of the y-axis, ξ (t) and ζ (t) are the time dependent amplitudes. Substituting expressions (12) in the equation set (10, 11) and applying Galerkin’s method and eliminating ζ (t) , the following differential equation is obtained: d2ξ(τ) dτ2 + {Λ1 − Λ2τ} ξ(τ) = 0 (13) where the dimensionless time parameter τ satisfies 0 ≤ τ ≤ 1,Λ1 and Λ2 are the parameters depending on the material properties, loading speed and the characteristics of the shell. The function satisfying the initial condition ξ (0) = ξ,τ (0) = 0 is in the following: ξ (τ) = Aξ1 (τ) = Ae pττ [ (p + 2) / (p + 1) − τ] (14) where A is found from the condition of transition to the static condition. The values of p will be determined numerically (Sofiyev and Schnak, 2004). For a cylindrical shell, the plastic buckling under axial compression is mainly axisymmetric, by taking this factor into consideration and substituting Eq. (14) in (13) then applying the Ritz type variational method to equation (13), for the static critical load and dynamic critical load the following expression is obtained: Tcrs = 2 3 h R (EsEt) 1/2 (15) Tcrd = T0tcr h = Ω0 3 Esh R 2 + √ 4 + 36T 20Ω1ρ (Ω0Es)3 R4 h4 1/2 (16) where Ω0, Ω1 are the double integral expressions and depending on τ and p. The expression (15) is firstly obtained in (Gerard, 1957). Es R ∂2w ∂x2 + ( 3 4 + 1 4ϕts ) ∂4Φ ∂x4 + ( 3 − 1 ϕts ) ∂4Φ ∂x2∂y2 + 1 ϕts ∂4Φ ∂y4 = 0 (11) ELASTO-PLASTIC CYLINDRICAL SHELLS 447 4. Numerical results and discussion In Table 1 the experimental results in (Lee, 1962, Mao and Lu, 2001, and Ore and Durban, 1992) and the calculated results with the present study are compared. The calculationss were performed with simple support cylindrical shells, made of Al 3003-0 with the material parameters E = 7.0 × 1010(Pa), N = 4.1, σy = 2.362 × 107(Pa), ρ = 2.77x103kg/m3. Ramberg-Osgood Equation, characterizes Al 3003-0 shells. The comparison shows that the deformation theory gives good results but the flow theory predicts much too high critical static loads. Their results obtained by using the deformation theory correspond well with those from the present study. TABLE I. Comparison of the critical static stress (MPa) with experimental and analytical results Geometry of Lee Ore and Durhan Mao and Lu Present the shell study R/h L/R Experimental Deform. Flow Deform. Flow Deform theory theory theory theory theory 9.36 4.21 96.87 88.49 162.33 89.71 165.46 88.21 19.38 4.10 78.60 74.09 122.03 74.87 124.25 72.68 29.16 4.06 64.74 67.06 103.98 67.70 106.0 64.91 TABLE II. Variations of the critical parameters for elastic and elasto-plastic buckling of cylindrical shells with different R/h and T0(R/L = 0.25) Elas. buck. Elas.-plas. buck. Elas. buck. Elas.-plas. buck. Al 3003-0 (T0 = 5×108 Pa×m/sec) R/h Tcrd (MPa) Tcrd /Tcrs 50 75 100 883.02(p=66) 89.50(p=6) 1.042 1.612 605.68(p=39) 87.48(p=5) 1.072 1.785 468.48(p=27) 86.65(p=4) 1.106 1.941 R/h Al 3003-0 (T0 = 10×108 Pa×m/sec) 50 75 100 903(p=42) 108.36(p=4.2) 1.067 1.953 630(p=25) 108.21(p=3.6) 1.114 2.216 495(p=18) 108.18(p=3.3) 1.167 2.451 In Table 2 are given the variations of the critical parameters for the elastic and elasto-plastic buckling of cylindrical shells with different R/h and T0. In the elasto-plastic buckling, the values of the dynamic critical load are con- siderably lower than the corresponding dynamic critical load in the elastic buck- ling. With an increase of the ratio R/h, the values of the dynamic critical load in the elastic and elasto-plastic buckling of the shell decrease, however, the values of the ratio Tcrd/Tcrs increase. When the ratio of R/h is increased, the values of the dynamic critical load are decreased in a more acute way in the elastic buckling, but this decrease is less in the elasto-plastic buckling. Furthermore, when the ratio of R/h is increased, the values of the ratio dynamic critical load to static critical are 448 SOFIYEV, SCHNACK increased more slowly in the elastic buckling, but this increase is more important in the elasto-plastic buckling. The increase of the ratio R/h affects more the values of the ratio dynamic critical load to static critical in the elasto-plastic buckling and this effect are even increased in aluminum shells. With an increase of the loading speed the values of the dynamic critical load and the values of the ratio Tcrd/Tcrs increase, whereas, that on the p decreases. 5. Conclusions The dynamic buckling of a cylindrical shell subject to a uniform axial compres- sion, which is a linear function of time, is examined. The material of the shell is incompressible and the effect of the elastic unloading is not taken into considera- tion. Initially, employing small elasto-plastic deformation theory, the fundamental relations and Donnell type dynamic stability equation of a cylindrical shell have been obtained. Then, applying the Galerkin method and Ritz type variational method the critical parameters have been found. Finally, carrying out some com- putations, the effects of the varying loading speed and of the varying radius to the thickness ratios on the critical parameters of the shells has been studied. References Gerard, G. (1957) Plastic stability theory of thin shells, Journal of Aeronautic Science 24 264-279. Hill, R. (1983) The Mathematical Theory of Plasticity, (Chapter 2), Oxford University Press, London. Ilyushin, A. A. (1947) The elasto-plastic stability of plates. National Advisory Commitee for Aeronautics, Technical Memorandum, No 1188, Washington. Karagizova, D., Jones, N. (2002) On dynamic buckling phenomena in axially loaded elasto-plastic cylindrical shells, Int. J. Non-linear Mech. 37 1223-1238. Lee, L. N. H. (1962) Inelastic buckling of initially imperfect cylindrical shells subjected to axial compresion, Journal of Aerospace Science 29 87-95. Mao, R., Lu, G. (2001) Plastic buckling of cicular cylindrical shells under combined in-plane loads, Int. J. of Solids and Structures 38 741-757. Ore, E., Durban, D. (1992) Elasto-plastic buckling of axially compressed circular cylindrical shells, Int. J.Mech. Sci. 34 727-742. Sofiyev, A. H., Schnack, E. (2004) The stability of functionally graded cylindrical shells under linearly increasing dynamic torsional loading, J. of Eng. Struct. 26 1321-1331. Wang, A. W., Tian, W. Y. (2003) Twin-characteristic-parameter solution of axisymmetric dynamic plastic buckling for cylindrical shells under axial compression waves, Int. J. of Solids and Structures 40 3157-3175. Wolmir, A. S. (1967) Stability of Deformable Systems, Moscow: Nauka. Wolmir, E. A. (1975) The Behavior of cylindrical panels under dynamic axial compression in elasto- plastic zone (in Russian), X International Conference on Theory of Plates and Shells 352-355, Kutaisi-Moscow. 449 449–454. DYNAMIC BUCKLING ANALYSIS OF IMPERFECT ELASTICA Ümit Sönmez Mechanical Engineering, İstanbul Technical University, Gümüşsuyu, 80191, İstanbul, Turkey Abstract. Dynamic buckling analysis of a pinned-pinned flexible beam attached to a sliding mass is investigated using nonlinear Elastica Theory. Initially straight flexible buckling beam having pinned end boundary conditions loaded at one of its end with curved imperfection is considered. Large de- flection analysis of flexible beam is studied using nonlinear elastica theory. Imperfection analysis of the flexible beam is investigated considering imperfection as an initial curvature. Dynamic response of the flexible beam is studied using numerical simulation procedures under step loading conditions; assuming this member buckles in its first mode. Key words: buckling dynamics, large deflection theory, imperfection analysis, elastica 1. Introduction Hoff and Brooklyn (1951) investigated the motion of a perfectly elastic initially slightly curved column using small deflection theory. Prathap and Varadan (1978) studied the large amplitude vibration of a pinned beam taking into account dis- placement terms up to cubic order. Dynamic response of a tip-loaded elastica with an end-mass is investigated (Snyder and Wilson, 1990). Cameron Boyle and al. (2003) designed and synthesized a compliant constant force robot arm using polynomial fits to elastica. 2. Large deflection analysis of flexible beams The geometry of the initially circular shaped beam is shown in Fig. 1. One end of the beam is fixed and the other end is subjected horizontal applied load P. The initial shape of the curved beam is a circular arc with radius ρ, and the curved beam angle α. Under the influence of horizontal P load, the beam end undergoes horizontal and vertical deflections (h0-h) and (b0-b) respectively. Nonlinear de- flection theory of curved beams originally derived by (Nordgren, 1956) and later investigated by (Sathyamoorty, 1998). These equations have not been used before in literature to investigate the imperfection buckling analysis. Large deflection analysis of curve beams may be summarized as follows. Assuming a slender beam Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 450 SÖNMEZ with an inextensional neutral axis, using the Bernoulli-Euler relation: Change in the beam’s curvature is proportional to beam’s bending moment. The following relation might be obtained Figure 1. Curved beam subjected to a horizontal load EI( dθ ds − 1 ρ ) = P(b − y) (1) Differentiating above equation with respect to length coordinate s and assigning a2 = P/EI, considering differential arc length ds2 = dx2 + dy2 Eq.1 reduces to d2θ ds2 = −a2 sin θ (2) Integrating Eq. (2) once and applying boundary conditions at the free end (θ = θ0), the following equation might be obtained 1 2 ( dθ ds )2 = a2(cos θ − cos θ0) + 1 2 ρ2 (3) At the beam fixed end (θ = 0) and dθds = ( 1 ρ ) + a2b. Using above Eq. (3) and further simplifying, the normalized vertical deflection of curved beam might be obtained as a quadratic polynomial with an unknown end angle θ0, ( b ρ )2 + 2 a2ρ2 ( b ρ ) − 2 a2ρ2 (1 − cos θ0) = 0 (4) Introducing a new variable β such that; cos β = cos θ0 − 0.5a2ρ2, curvature as dθds = dθ dx dx ds = dθ dx cos θ = √ 2a(cos θ − cos β)0.5 and integrating the following relation may be obtained h = 1 √ 2a ∫ cos θ (cos θ − cos β)0.5 dθ (5) Above integral relation can be converted to an integral equation containing elliptic integrals of first and second kind with amplitude γ0 and modulus k if variables transform as (1 − cos θ) = 2k2 sin2 γ = (1 − cos β) sin2 γ ( h ρ ) = 1 aρ [ 2E(γ0, k) − F(γ0, k) ] = 0 (6) BUCKLING DYNAMICS OF ELASTICA 451 Where F(γ0, k) and E(γ0, k) are elliptic integrals of first and second kind; and γ0 is given by sin2 γ0 = (1 − cos θ0)/(1 − cos θ0 + (1/2a2ρ2)) (7) Using inextensibility criteria L = ρα the following relationship could be found, aρα = ∫ 1 (1 − k2 sin2 γ)0.5 dγ = F(γ0, k) (8) The above equations may be solved numerically to obtain nonlinear load de- flection characteristics of a curved beam subjected to a horizontal load. The load deflection characteristic of a fixed-free beam with length 0.5L subjected to a horizontal load is same as the pin-pin beam with length L due to the symme- try. The forces acting on a simple pinned-pinned segment are collinear along the line between the two pin joints. Moreover deflections of the flexible pinned- pinned segment are along this line. Pin ends do not carry moments therefore pinned-pinned flexible beam is a two-force member. 3. Load deflection plots of curved beams Load deflection plots of curved beams with three different subtending angles are shown in Fig. 2. Three different subtending angles α = 0.1rad, α = 0.01rad and α = 0.001rad are considered in this investigation. 0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25 Normalized Deflection N or m al iz ed L oa d Load Deflection Characteristics of Imperfect Beam 0.001 0.01 0.1 Figure 2. Normalized load-deflection plots of the imperfect beams While the angle decreases also the imperfection decreases and the beams be- come more straight. Load deflection plots are obtained from solution of highly nonlinear Eq. (4) and Eq. (6) that takes into account large deflections and rotations. Deflection solutions are obtained until the pin ends touch each other. Imperfection of α = 0.001 beam is not noticeable to a naked eye when the curve beam is plotted. Normalized load (p = PL2/EI) and normalized deflection 452 SÖNMEZ (u = y/L) results are plotted in Fig. 2. While 0.001 and 0.01 beams exhibit obvious critical buckling loads; 0.1 beam does not have one, but the load-deflection-slope change rate switches from a decreasing one to an increasing one at a certain load, this load could be considered as the critical load for comparison purposes. All three beams show softening spring characteristics with very little change in de- flections until the critical buckling load, and then hardening spring characteristics with large change in deflections. 4. Dynamic response of imperfect curved beams Dynamic response of imperfect beams under a step loading condition; attached to a sliding mass at one end, pinned at both ends (as shown in Fig. 4) are investigated solving corresponding nonlinear equation of motion numerically. Fourth order Runge-Kutta routine in MATLAB (ODE45) is used to simulate the equation of motion. Figure 3. Free body diagram of a slider attached to a curved beam The slider’s equation of motion may be written as, mÿ + cẏ + FNL = P (9) Where m is the slider’s mass, c is the viscous damping coefficient and FNL nonlin- ear load deflection characteristics of corresponding flexible curve beam segment and y is the axial displacement of the curve beam. Step loads are applied to a steel flexible beam with elasticity modulus E = 2.07 ∗ 1011N/m2, cross sectional area A = 16mm ∗ 0.1mm and length L = 10cm. Flexible beam cross section dimension is chosen from the commonly used standard steel spring strips described in the Handbook of Spring Design (SMI, 1991). This beam deflects elastically until the maximum deflection occurs; when the pinned ends touch each other. The stress at the middle of the beam (maximum stress location) should be kept below the yield stress of steel spring σmax = 800MPa and σyield = 1000MPa). Slider BUCKLING DYNAMICS OF ELASTICA 453 mass m = PEuler/g is used in all examples (where g is the gravitational constant g = 9.81N/m2). 5. Curved beam under the influence of a step input Dynamic response of the curved beams is obtained under the influence of a step load with magnitudes 0.5PE , and 1.0PE , with no damping, where PE is the cor- responding Euler buckling load of initially straight beam (perfect beam with the same cross sectional area and length). When 0.5PE is applied to the imperfect beams, the deflections are quite small, because 0.5PE is half of the buckling crit- ical buckling load of the perfect beam and therefore beams does not buckle. The deflection responses oscillate around the static equilibrium points with the follow- ing magnitudes and their ratios: y001max = 1.387E−7m, y01max = 1.312E−5m and y1max = 1.293E − 3m. Curved beam length L = 100mm, y1max/y01max = 99 and y1max/y001max = 9300. The imperfect beams vibrate along the pinned-pinned axis with the following frequencies: f001 = 976.6Hz > f01 = 97.6Hz > f1 = 9.76Hz. These frequencies are exact multiples of each other f001/ f1 = 100 and f01/ f1 = 10. These results are confirmed with the linear theory results, where there is an inverse proportional relationship between frequencies and corresponding beam angles for small deflections. Figure 4. Dynamic response for step input 0.5PE Step input PE causes curved beams to oscillate with amplitudes:0.6cm,1cm, 4cm, these displacements are large deflections and curved beam vibrations have the following frequencies: f001 = 1.26Hz < f01 = 1.71Hz < f1 = 1.76Hz. The order of the vibration frequencies of large deflections are reversed that of the small ones. 6. Conclusions The dynamic buckling response of imperfect straight beams with pinned ends is investigated using elastica theory for curved beams. The results show significant 454 SÖNMEZ Figure 5. Dynamic response for step input PE differences in vibration frequencies and deflections under the influence of same applied load as the beam imperfection parameter, very small subtending curved beam angle, changes. If a step load smaller than the critical Euler buckling load is applied, the vibration frequencies of the curved beam increase as the beam imperfection decreases. However the same is not true for the loads greater than the Euler buckling load, the vibration frequency order reverses. The vibration frequencies increases as the imperfection increases and the frequencies are of the same order (close to each other). Similar observations are made regarding to deflections: Step input loads less than Euler bucking load cause small deflections with distinct magnitudes, greater than buckling load causes large deflections of the same order. References C. Boyle C., L. L. Howell, S. P. M. and Evans, M. S. (2003) Dynamic Modeling of Compliant Constant-Force Compression Mechanisms, Mechanisms and Machine Theory 38, 1469–1487. Conway, H. D. (1956) The Nonlinear Bending of Thin Circular Rods, Journal of Applied Mechanics 23, 7–10. Hoff, N. J. and Brooklyn, N. Y. (1951) The Dynamics of the Buckling of Elastic Columns, Journal of Applied Mechanics pp. 68–75. Prathap, G. and Varadan, T. K. (1978) The Large Amplitude Vibration of a Hinged Beam, Journal of Applied Mechanics 45, 959–961. Sathyamoorty, M. (1998) Nonlinear Analysis of Structures, CRC Press, New York. SMI (1991) Handbook of Spring Design, Spring Manufacturers Institute. Snyder, J. M. and Wilson, J. F. (1990) Dynamic of the Elastica with End and Follower Loading, Journal of Applied Mechanics 57, 203–208. 455 455–461. SPECIFICATION OF FLOW CONDITIONS IN THE MATHEMATICAL MODEL OF HYDRAULIC DAMPER Rudolf Svoboda1, Jan S̆kliba2 and Radek Matĕjec2 1Techlab Ltd, Praha, Czech Republic 2Technical University of Liberec, Czech Republic Abstract. Hydraulic dampers represent one of the basic instruments for absorption of vibration in dynamic systems. The damper is substituted either by its hydraulic velocity characteristic, or directly by a mathematical model as a dynamic subsystem. The standard damper model does not provide satisfactory results especially those concerning the strokes of all four damper valves. To improve these results it is necessary to simulate the flows through valves more precisely and, last but not least, to set adequately correct values to all essential parameters of both mechanical and hydraulic parts of the damper. In the paper is presented a new, corrected formula for discharge flow coefficients based on the measurements of flow characteristics of throttle elements of the damper with constant as well as variable slot width. The experimental equipment used for identification process is described as well. Key words: micromechanical scanners, torsional frequency, torsional rigidity, orthotropic material 1. Introduction Hydraulic damper modelling forms a part of theoretical and applied mechanics which is not getting publicity that would deserve. Leaving aside the works of corporate research and development which are usually not made public, during the last thirty years we can name just a few results published in prestigious journals or international conferences (Weigel et al., 2002). Nevertheless, thanks to the intense progress in computer techniques, computer times needed for damper model simu- lations on digital computers are presently comparable with former simulations on hybrid computers (Skliba, 1983) and the time is coming when the mathematical model of hydraulic damper will work in real-time dynamic systems. Therewith relate possibilities of damper model improvements consisting in application of powerful modern computing methods on the one hand and in exploitation of ex- perimental results and their implementation into the model on the other hand. It will be possible to use the improved mathematical model of hydraulic damper either separately for damper designing and development purposes or as a com- ponent of the precise model of a dynamic system (for instance a vibroisolating Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 456 SVOBODA, S̆KLIBA, MATĔJEC system) in which the damper is applied and where it is not sufficient to substitute the damper by its static velocity characteristics. 2. Problem formulation Owing to the intended utilizations of dynamic model of hydraulic damper in calculations of real-time dynamic systems it is unacceptable to solve damper dynamics in full complexity using finite element methods for calculations of me- chanical damper parts movements and finite volumes method for description of flows inside the damper. To ensure reasonable computer time all mechanical parts of the damper are modelled using concentric parameters and the flows through the valves are modelled on the bases of laws of hydraulics. Therefore, the present dy- namic models of hydraulic damper deal with pressures concentrated into separate regions (e.g. regions above the piston, under the piston and in the accumulator, supply channels of valves etc.); the flows among regions of the damper are sup- posed to be one-dimensional, the liquid is compressible, in the state of unsaturated solution (it does not contain free air), viscose. Figure 1. A damper cross-section view and corresponding scheme of damper model MATHEMATICAL MODEL OF HYDRAULIC DAMPER 457 The standard mathematical model of the damper consists of: a) systems of differential equations describing equations of motion of valves, b) equations of dynamic balance of flows among the damper regions, c) equations describing evolution and destruction of columns of saturated vapors, d) algebraic equations describing dependence of discharge flow coefficients on corresponding Reynolds numbers. In case of enhanced model of hydraulic damper where the flows through valves are modeled more precisely and pressure gradients in supply channels are respected, equations for unknown flows Q j and pressure PC, j in supply channel of j-th valve ( j = 1, ..., 4) must be added to the model. But in many cases even this enhanced models do not provide fully acceptable results. The computed valve strokes do not fully correspond to experimentally obtained strokes and the courses of transient valve vibrations can differ from the reality too. Moreover, on certain combinations of definition parameters some zones with strong vibrations of self-excited type can occur in the solution. These vibrations arise in steady parts of both damper expansion and rebound phases where all transient phenomena have gone off long ago and working valves are fully open. It yields from analysis of linearized hydraulic damper model that the system is not stable for a wide range of damper definition parameters and that self-excited valve vibrations are generated by the oil flowing through the opened valve slot. Because the instability is strongly influenced by this flow the correct determination of valve definition parameters is crucial for description of vibration phenomena. 3. Specification of flow conditions From the point of view of mathematical damper model fidelity the replacement of empirical values used till this time by experimentally verified input data should be as broad as possible. Especially exact determination of flow definition parameters of both pressure and discharge valves is decisive for true description of flow phe- nomena in the damper. On the bases of precise experiments it was found that the standard formula used for flow description through the valve slot Q = α. √ ν ρ S O √ ∆P α = α(Re), Re = Q.dH S 0ν (S 0 slot area, ∆P pressure gradient, ρ oil density, α dimensionless discharge coefficient dependent on the Reynolds number, that represents a correction of 458 SVOBODA, S̆KLIBA, MATĔJEC the flow regarding to the actual valve design, dH hydraulic diameter, ν kinematic viscosity) does not hold exactly. It results from evaluated experiments that both asymptotic value of the discharge coefficient and the steepness of its course in the start-up zone depend not only on Reynolds number but also on the valve stroke. This dependency on actual working conditions is especially strong on transient conditions of valve closing and opening. The results of regress analysis show that the identified dependency can be treated in the following approaches: a) To replace the area S 0(x) of the valve slot by a reduced slot area S red(x) b) In the formula for discharge flow coefficient α = α0 exp(α1.Re) to respect the dependency on the valve stroke x, so α0 = α0(x), α1 = α1(x). c) To replace the exponential formula for discharge flow coefficient by the rela- tion α(Re, x) = F(Re, x) = α00(x) + α0(x) α1(x)Re 1+α1(x)Re , which better complies with measured data. These results implemented in the damper model provide to improve further findings concerning the problems of flow-induced self-excited vibrations of damper valves. 4. Experimental equipment A special test stand (Figure 2) intended for determination of flow characteristics of throttle valve elements was constructed in laboratory of Technical university of Liberec. The main parts of the test arrangement form: a)source of pressurized liquid, b) working cylinder, c) generator of alternate liquid flow and d) the test chamber. For the purpose of flow characteristics identification the following values must be measured during the measurements: − pressure gradients caused by the flow through the throttle elements, − temperatures of flowing oil, − flow area of throttle elements with variable slot (typically valves). These quantities are necessary for determination of dependence of the pres- sure gradient on the volume flow through this element, from which the discharge coefficients are further identified. The assembly of the piston with the discharge and back valves used for exper- imental purposes is shown on Figure 4; the points for pressure and temperature measurement are presented on the configuration scheme on Figure 3. With respect MATHEMATICAL MODEL OF HYDRAULIC DAMPER 459 Figure 2. Test chamber and alternate flow generator to the piston and valves dimensions an application of expensive miniature pressure and stroke sensors was inevitable. Figure 3. Measuring points, (pressure and temperature) Figure 4. Components of the piston, dis- charge valve and back valve The main difficulty, which is not satisfactorily solved till this time, consists in measurement of strokes of valve plates. Valve strokes attain tens of microns 460 SVOBODA, S̆KLIBA, MATĔJEC only and must be therefore measured with accuracy to microns, which is a hardly viable problem in itself. Moreover, it deals with installation of stroke sensors into the areas with severe local conditions (high pressure up to 3Mpa, temperature up to 250◦C). For the present an alternative method of stroke measurement on the bases of plate stops is used; however, the method is too work and time consuming and not very precise. Figure 5. Measured characteristic of the discharge valve In Figure 5 are presented typical courses of pressure gradients and static char- acteristics of the discharge valve which were obtained in the process of measure- ments. On the bases of these (and another) measured static damper characteristics and on the bases of other supplementary flow characteristics (flow area, oil tem- perature, viscosity and modulus, etc.) the required flow discharge coefficients of the tested throttle element in dependence on Reynolds number were determined. 5. Conclusions It is evident that the first stage of the damper improvements will be crowned by experimentally verified model of hydraulic damper, which is oil-filled in the state with pulled out piston rod, where the pressure does not fall under the limit of atmospheric pressure and with presumption that working liquid has unchanging concentration of free and dissolved air. From the point of view of mathematical damper model fidelity the replacement of empirical values used till this time by experimentally verified input data should be as broad as possible. Especially exact determination of flow definition parameters of both pressure and discharge valves is decisive for true description of flow phenomena in the damper. The next damper improvements complying with the variable air content in liquid and with the possibility of formation of cavitation will demand heavy effort in future yet. MATHEMATICAL MODEL OF HYDRAULIC DAMPER 461 Acknowledgement This work has been supported by Ministry of Education of Czech Republic in the frame of the project No. 4674788501. References Derbaremdiker A. D. (1985) Hydraulic dampers of transport machines (in Russian), Mashinostroie- nie, Moscow, Russia. Lang H. L. (1978) A study of the characteristics of automotive hydraulic dampers at high stroking frequencies, Doctor dissertation, University of Michigan. Škliba J. (1983) Zur Problematik der Modellierung eines hydraulishen Dampfer, Proceedings of International Conference Dynamics of Machines, Praha-Liblice, Czech Republic. Škliba J., Svoboda R. (2000) Pressure distribution in a hydraulic damper, 8th International Conference on the theory of Machines and Mechanisms, 623-628 Liberec, Czech Republic. Škliba J., Svoboda R. (2004) On the necessity of hydraulic damper model specification, Proceedings of International Conference Flow Induced Vibration, Paris, France. Svoboda R., Škliba J. (2003) Identification of stiffness of hydraulic damper valves, Proceedings of 6th International Conference on Vibration Problems, 78-80 Liberec, Czech Republic. Weigel M., Mach W., Riepl A. (2002) Nonparametric shock absorber modelling based on standard test data, Vehicle Syst. Dynamics 38 415-432. 463 463–468. NONLINEAR TRANSIENT ANALYSIS OF RECTANGULAR COMPOSITE PLATES Hakan Tanrıöver and Erol Şenocak Faculty of Mechanical Engineering, İstanbul Technical University, Mechanical Engineering Department, Gümüşsuyu, TR-34437, İstanbul, Turkey Abstract. The geometrically nonlinear analysis of laminated composite plates under dynamic loading is considered. Galerkin method with the use of Newmark’s scheme in association with Newton-Raphson method is applied to obtain the dynamic nonlinear response of the plates. First order shear deformation theory based on Mindlin’s hypothesis and von Kármán type geometric nonlinearity are utilized. The governing differential equations are solved by choosing suitable poly- nomials as trial functions to approximate the plate displacements. The solutions are compared to that of Chebyshev series, finite strips and finite elements. A very close agreement has been observed with these approximating methods. In the solution process, analytical computation has been done wherever it is possible, and analytical-numerical type approach has been made for all problems. Key words: Galerkin method, laminates, shear deformation, large deflection, transient analysis 1. Introduction Various numerical techniques can be utilized to investigate the geometrically non- linear dynamic response of laminated composite plates. Reddy (1983) applied the finite element method (FEM), Chen et al. (2000) developed a semi-analytical finite strip method (FSM) and Nath and Shukla (2001) used Chebyshev series (CS) technique for the nonlinear transient analysis of laminates. Note that the first order shear deformation theory (FSDT) based on Mindlin’s hypothesis and von Kármán type geometric nonlinearity were employed in these works. In the present paper geometrically nonlinear transient analysis of laminated composite plates is performed using the Galerkin method (GM). FSDT based on Mindlin’s hypothesis with von Kármán nonlinearity is utilized and the dynamic nonlinear analysis is performed through using the Newmark method in association with the Newton-Raphson method. In the solution process, computations have been carried out analytically wherever it is possible and analytical-numerical type approach has been made for all cases. Suitable polynomials are chosen as trial functions to approximate the plate displacements. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 464 TANRIÖVER, ŞENOCAK 2. Governing equations Consider a rectangular laminated plate with dimensions a, b and uniform thick- ness h. The origin of the coordinate system is chosen to coincide with the center of the midplane of the undeformed plate. The plate is assumed to be subjected uni- form transverse pressure qo, and it is constructed of finite homogenous orthotropic layers perfectly bonded together. Under the assumptions of first order shear deformation theory based on Mindlin’s hypothesis; let u, v, w denote the displacements at an arbitrary point of the plate in the x, y, z directions and u0(x, y), v0(x, y), w0(x, y) are the dis- placements at a corresponding point of the midplane of the plate in the x, y and z directions respectively. Then the displacement field of the first order theory is of the form (Mindlin, 1951): u = u0 + φxz, v = v 0 + φyz, w = w 0, (1) where φx and φy are the rotations of a transverse normal about the y and x axes respectively. The corresponding total strains could be expressed as follows: εx = u0,x + 1 2 w 2 ,x + κxz, εy = v 0 ,y + 1 2 w 2 ,y + κyz, γxy = u0,y + v 0 ,x + w,xw,y + κxyz, γxz = w,x + φx, γyz = w,y + φy. (2) where differentiations are denoted by comma. Midplane curvatures and twist of the plate are the following: κx = φx,x, κy = φy,y, κxy = φx,y + φy,x. (3) For a plate with an arbitrary number of layers, the constitutive relations are { N M } = [ A B B D ] { ε0 κ } , { Qy Qx } = K [ A44 A45 A45 A55 ] { γ0yz γ0xz } , (4) where N and M are the resultant forces and moments conjugate to ε0 and κ respec- tively. Qx and Qy are transverse forces and the parameter K is shear correction factor taken as 5/6 in the analyses. Five governing equations of motion for the plate can be written as follows in the general form (Reddy, 1997): R1 = Nx,x + Nxy,y − I0u0,tt − I1φx,tt = 0, R2 = Nxy,x + Ny,y − I0v0,tt − I1φy,tt = 0, R3 = Qx,x + Qy,y + (w,xNx + w,yNxy),x +(w,xNxy + w,yNy),y + qo − I0w,tt = 0, R4 = Mx,x + Mxy,y − Qx − I2φx,tt − I1u0,tt = 0, R5 = Mxy,x + My,y − Qy − I2φy,tt − I1v0,tt = 0. (5) NONLINEAR TRANSIENT ANALYSIS 465 For the Galerkin approach, the displacements of the plate are approximated in the form shown below: u0 = M∑ m=0 N∑ n=0 amn( t )Umn(x, y), v 0 = M∑ m=0 N∑ n=0 bmn( t )Vmn(x, y), w = M∑ m=0 N∑ n=0 cmn( t )Wmn(x, y), φx = M∑ m=0 N∑ n=0 dmn( t )S mn(x, y), φy = M∑ m=0 N∑ n=0 emn( t )Tmn(x, y), (6) where amn, bmn, cmn, dmn and emn are unknown functions of time, Umn, Vmn, Wmn, S mn and Tmn are the trial functions, and M and N are the number of terms in x and y directions respectively. Herein, polynomials are used as trial functions, which are chosen to satisfy the geometric boundary conditions, where as natural boundary conditions are not satisfied. In this case, simultaneous approximation is made to the solutions of differential equations and to the boundary conditions. ∫ +b/2 −b/2 ∫ +a/2 −a/2 UmnR1dxdy − ∫ +b/2 −b/2 UmnNx|x=±a/2dy − ∫ +a/2 −a/2 UmnNxy|y=±b/2dx = 0,∫ +b/2 −b/2 ∫ +a/2 −a/2 VmnR2dxdy − ∫ +a/2 −a/2 VmnNy|y=±b/2dx − ∫ +b/2 −b/2 VmnNxy|x=±a/2dy = 0,∫ +b/2 −b/2 ∫ +a/2 −a/2 WmnR3dxdy = 0,∫ +b/2 −b/2 ∫ +a/2 −a/2 S mnR4dxdy − ∫ +1 −1 S mnMx|x=±a/2dy = 0,∫ +b/2 −b/2 ∫ +a/2 −a/2 TmnR5dxdy − ∫ +1 −1 TmnMy|y=±b/2dx = 0. (7) 3. Solution procedure In the application of the Galerkin method the geometrical boundary conditions are satisfied by choosing appropriate trial functions. Trial functions are weighted polynomials given as follows: Umn = Φ1xmyn, Vmn = Φ2xmyn, Wmn = Φ3xmyn, S mn = Φ4xmyn,Tmn = Φ5xmyn, (8) where Φi (i = 1, . . . 5) denote the weight functions. Substituting Eq. (6) into Eqs. (7), nonlinear equations in terms of unknown coefficients amn, bmn, cmn, dmn and emn are obtained. These equations are solved by employing the Newton-Raphson methodology. 466 TANRIÖVER, ŞENOCAK TABLE I. Boundary conditions and corresponding weight func- tions u0 = v0 = w = Mx = φy = 0 at x = ±a/2, u0 = v0 = w = My = φx = 0 at y = ±b/2. SS–1 Φi = (x2 − a2/4)(y2 − b2/4) (i = 1, . . . 3), Φ4 = (y2 − b2/4), Φ5 = (x2 − a2/4). Nx = v0 = w = Mx = φy = 0 at x = ±a/2, Ny = u0 = w = My = φx = 0 at y = ±b/2. SS–2 Φ3 = (x2 − a2/4)(y2 − b2/4), Φi = (y2 − b2/4) (i = 1, 4), Φi = (x2 − a2/4) (i = 2, 5). Two different boundary conditions are considered and shown in Table 1. Note that whole plate models are analyzed in all cases presented here. To integrate Eqs. (7), Newmark’s direct integration scheme (Newmark, 1959) is employed. In the Newmark scheme the first time derivative of the displacement field U̇ and the solution U are approximated at (n + 1) time step (i.e., at time t = tn+1 ≡ (n + 1)∆t) by the following expressions: U̇n+1 = U̇n + ∆t[(1 − γ)Ün + γÜn+1], Un+1 = Un + U̇n∆t + ∆t 2 2 [(1 − 2β)Ün + 2βÜn+1], (9) where γ and β parameters are chosen as 1/2 and 1/4 respectively in all of the present analyses. The method gives nonlinear equations in terms of unknown co- efficients amn, bmn, cmn, dmn and emn in each time step. These equations are solved by employing the Newton-Raphson methodology. Note that the evaluations of integrals are symbolically computed by using a commercial computer math code MathematicaT M (Wolfram, 1988). 4. Numerical examples Geometrically nonlinear transient analyses of an orthotropic and a cross-ply lam- inate are accomplished as a verification of the present technique. The orthotropic laminate is under SS–2 type boundary condition. The ma- terial, geometric and loading data are taken from (Chen et al., 2000) and given below. E1 = 525000 N/mm2, E2 = 21000 N/mm2,G12 = G13 = G23 = E2/2, ν12 = 0.25, a = b = 250 mm, h = 5 mm, ρ = 800 kg/m3, qo = 1 N/mm2. Here, the time step ∆t is taken as 10 µsec. Results of GM (M = N = 5), FSM (Chen et al., 2000) and commercial FEM program ABAQUS are given in Fig. 1. NONLINEAR TRANSIENT ANALYSIS 467 In the ABAQUS analyses presented here 10 × 10 mesh with S4 type elements is used. 0.25 0.5 0.75 1 1.25 1.5 1.75 2 time 0 1 2 3 4 c e n t e r d e f l e c t i o n ABAQUS FSM GM (a) 0.25 0.5 0.75 1 1.25 1.5 1.75 2 time 0 1 2 3 4 c e n t e r s t r e s s ABAQUS FSM GM (b) Figure 1. Response of an orthotropic laminate in time (millisec): (a) Central deflection (w/h). (b) Central stress (525 σx/E1) Nonlinear transient analysis of an unsymmetric cross-ply [0◦/90◦/0◦/90◦] laminate is considered. The cross-ply laminate is under SS–1 boundary condition. Material properties, geometry, loading data and related normalized variables of the plate are taken from (Nath and Shukla, 2001) and given below. E1/E2 = 25,G12 = G13 = E2/2,G23 = E2/5, ν12 = 0.25, a = b, ξ = a/h = 10, q′ = qoa 4 E2h4 = 125,w′ = wh , τ = t √ 4A22 I0h2ξ2 Here, the time step ∆τ is taken as 0.1. Results of GM, CS techniques (?) and ABAQUS are given in Fig. 1. 5. Conclusions Geometrically nonlinear transient analysis of composite plates based on FSDT is performed by using Galerkin approach with the use of Newmark’s scheme in association with Newton-Raphson method. The choice of trial functions is cru- cial to approximate the two dimensional displacement field. The present solution methodology may be used to solve dynamic large deflection analysis of the lam- inates in an easy and effective way with the help of a symbolic math package. The method is found to determine closely the displacements with a few number of terms. The results are compared to that of known other approximating methods (Chebyshev series and finite strips), and commercial FEM code ABAQUS. A very good agreement is observed. 468 TANRIÖVER, ŞENOCAK 5 10 15 20 25 30 35 40 time �0.25 0 0.25 0.5 0.75 1 1.25 c e n t e r d e f l e c t i o n A CS GM 5 GM 3 Figure 2. Center deflection (w′) of the cross-ply laminate in time (τ). GM 3: M = N = 3 in GM, GM 5: M = N = 5 in GM, CS: Chebyshev series results. A: ABAQUS results References Chen, J., Dawe, D. J., and Wang, S. (2000) Nonlinear transient analysis of rectangular composite laminated plates, Composite Structures 49, 129-139. Mindlin, R. (1951) Influence of rotatory inertia and shear on flexural motions of isotropic, elastic plates, Journal of Applied Mechanics 18, 31-38. Nath, Y. and Shukla, K. (2001) Non-linear transient analysis of moderately thick laminated composite plates, Journal of Sound and Vibration 247, 509-526. Newmark, N. M. (1959) A Method of Computation for Structural Dynamics, Journal of the Engineering Mechanics Division 8, 67-94. Reddy, J. N. (1983) Geometrically Nonlinear Transient Analysis of Laminated Composite Plates, AIAA 21, 621-629. Reddy, J. N. (1997) Mechanics of Laminated Composite Plates, New York, CRC Press. Wolfram, S. (1988) Mathematica: A System for Doing Mathematics by Computer, Redwood City, CA. 469 469–474. ON THE “RESONANCE” VALUES OF THE DYNAMICAL STRESS IN THE SYSTEM COMPRISES TWO-AXIALLY PRE-STRETCHED LAYER AND HALF-SPACE Fatih Taşçı,1 İbrahim Emiroğlu1 and Surkay D. Akbarov2 1Faculty of Chemistry and Metallurgy, Department of Mathematical Engineering, YTÜ, Davutpasa Campus, No:127, 34210 Esenler, İstanbul, Turkey 2Inst. of Math. and Mech. of National Academy of Sciences of Azerbaijan, Baku, Azerbaijan Abstract. This paper is devoted to the study of the influence of the pre-stretching of a covering layer on the “resonance” values of a stress. The investigations are performed within the framework of the piecewise homogeneous body model with the use of the Three Dimensional Linearized Theory of Elastic Waves in Initially Stressed Bodies (TLTEWISB). Numerical results are presented for some selected materials. Key words: elastic waves, bi-axially pre-stretch layer, resonance 1. Introduction In the present paper the investigations started in the papers by (Akbarov and Ozaydın, 2001a), (Akbarov and Ozaydın, 2001b), (Akbarov and G‘̀ uler, 2005), (Güler and Akbarov, 2004), (Emiroglu et al., 2004) are developed for the forced vibration of the half-space covered with the bi-axially pre-stretched layer. In this case the attention is adressed to the determination of the “resonance” values of the frequency of the external point-located force. The investigations are carried out within the framework of the piecewise homogeneous body model with the use of the Three Dimensional Linearized Theory of Elastic Waves in Initially Stressed Bodies (TLTEWISB). Numerical results are presented for concrete se- lected materials. The influence of the pre-stretching of the covering layer on the absolute maximum values of the interface stresses under “ resonance” frequency is analysed. Throughout the study, repeated indices indicates summation over their ranges. However underlined repeated indices are not summed. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 470 TAŞÇI, EMİROĞLU, AKBAROV 2. Formulation of the problem Consider the half-space covered by a bi-axially per-stretched layer. For generality, we assume that the half-space is also bi-axially pre-stressed. With the covering layer we associate a lagrangian coordinate system, Ox1x2x3, which is in the un- deformed state, would coincide with a cartesian coordinate system. Note that the covering layer and the half-space occupy the regions {−∞ < x1 < +∞,−h1 ≤ x2 ≤ 0,−∞ < x3 < +∞} and {−∞ < x1 < +∞,−∞ < x2 ≤ −h,−∞ < x3 < +∞}, respectively. We assume that, before contact, the layer and the half-space are pre- stressed separately in the directions of Ox1 and Ox3 axes, and homogenous initial stress states appear in both materials. The values related to the layer and half-space are denoted by upper indices (1) and (2), respectively. The values related to the initial stresses are denoted by upper indices (m), 0 where m = 1, 2. The linearly elastic material of the layer and the half-space are to be taken homogenous and isotropic. The initial stresses in the layer and the half-space are determined within the framework of the classical linear of elasticity as follows σ(m),011 = const1m, σ (m),0 33 = const3m, σ (m),0 i j = 0 f or i j � 11; 33. (1) According to [Guz 2004], the equations for the small initial deformation state version of the TLTEWISB are ∂σ(m)i j ∂x j + σ (m),0 11 ∂2u (m) i ∂x21 + σ (m),0 33 ∂2u (m) i ∂x23 = ρ (m) 0 ∂2u (m) i ∂t2 , i; j = 1, 2, 3; m = 1, 2. (2) In Eq. (2), ρ(m)0 denotes the density of the m-th material in the natural state. For an isotropic compressible material one can write the following mechanical relations. σ (m) i j = λ (m)θ(m)δi j + 2µ (m)ε (m) i j , θ (m) = ε(m)11 + ε (m) 22 + ε (m) 33 (3) where λ(m), µ(m) are Lame’s constants and ε(m)i j = 1 2 ( ∂u(m)i ∂x j + ∂u(m)j ∂xi ). (4) We assume that the following complete contact conditions exist between the layer and the half-space. u(1)i |x2=−h = u (2) i |x2=−h, σ (1) i2 |x2=−h = σ (2) i2 |x2=−h, i = 1, 2, 3. (5) In the free face plane of the covering layer, the following conditions are satis- fied σ(1)32 |x2=0 = σ (1) 12 |x2=0 = 0, σ (1) 22 |x2=0 − P0δ(x1)δ(x3)e iwt (6) TWO-AXIALLY PRE-STRETCHED LAYER AND HALF-SPACE 471 In addition to these, we also assume that as x2 → −∞ there is no reflection, which means all waves travel in the negative x2 direction. In other words, we assume that |u(2)i |, |σ (2) i j | < M = constant, as x2 → −∞. (7) This completes the formulation of the problem. Thus, within the framework of the equations (1)-(4), the contact conditions (5), the boundary conditions (6), and the assumption (7), we investigate the forced vibration caused by time-harmonic point-located force in the bi-axially pre-stressed half-space covered by the bi- axially pre-stretched layer. It should be noted that in the case where σ(m),011 = 0, σ (m),0 33 = 0(m = 1, 2), the problem transforms to the corresponding formulation of the classical linear theory. 3. Method of solution We prefer to apply the double integral (Fourier transformation) method for the solution of the considered problem. We attempt to use Lame’s representations for displacements (Eringen and Suhubi; 1975) u1 = ∂φ ∂x1 + ∂ψ3 ∂x2 − ∂ψ2 ∂x3 , u2 = ∂φ ∂x2 + ∂ψ ∂x3 − ∂ψ3 ∂x1 , u3 = ∂φ ∂x3 + ∂ψ2 ∂x1 − ∂ψ1 ∂x2 (8) In the hypothesis of time-harmonic motion with circular frequency ω, every field g(x1, x2, x3, t) can be expressed as g(x1, x2, x3)eiωt and the following equations are obtained for φ, ψ1, ψ2 and ψ3 from (2)-(4): ∇2φ(m) + σ (m),0 11 λ(m) + 2µ(m) ∂2φ(m) ∂x21 + σ (m),0 33 λ(m) + 2µ(m) ∂2φ(m) ∂x23 + (C(1)2 ) 2 (C(m)1 ) 2 Ω2φ(m) = 0, ∇2ψ(m)n + σ (m),0 11 µ(m) ∂2ψ (m) n ∂x21 + σ (m),0 33 µ(m) ∂2ψ (m) n ∂x23 + (C(1)2 ) 2 (C(m)2 ) 2 Ω2ψ (m) n = 0, ∂ψ(m)1 ∂x1 + ∂ψ(m)2 ∂x2 + ∂ψ(m)3 ∂x3 = 0, (9) where Ω = ωh c(1)2 , c(m)2 = √ (µ(m)/ρ(m)0 , c (m) 1 = √ (λ(m) + 2µ(m))/ρ (m) 0 . From Eqs. (9), we employ the double Fourier transformation with respect to the coordinates x1 and x3: f13F(s1, x2, s3) = ∫ +∞ −∞ ∫ +∞ −∞ f (x1, x2, x3)e −i(s1 x1+s3 x3)dx1dx3. (10) 472 TAŞÇI, EMİROĞLU, AKBAROV After applying the transformation (10) to Eqs. (9), the functions φ(m)13F and ψ (m) n13F are determined as follows: φ(1)13F = A (m) 1 (s1, s3)e γ(1)1 (s1,s3)x2 + A(1)2 (s1, s3)e −γ(1)1 (s1,s3)x2 , φ(2)13F = A (2) 1 (s1, s3)e γ(2)1 (s1,s3)x2 , ψ(1)n13F = B (1) 1n (s1, s3)e γ(1)2 (s1,s3)x2 + B(1)2n (s1, s3)e −γ(1)2 (s1,s3)x2 , ψ(2)n13F = B (2) 1n (s1, s3)e γ(2)2 (s1,s3)x2 (11) where (γ(m)1 ) 2 = s21(1 + σ (m),0 11 λ(m) + 2µ(m) ) + s23(1 + σ (m),0 33 λ(m) + 2µ(m) ) − (c2) 2 (c (m) 1 ) 2 Ω2, (γ(m)2 ) 2 = s21(1 + σ (m),0 11 µ(m) ) + s23(1 + σ (m),0 33 µ(m) ) − (c(1)2 ) 2 (c(m)2 ) Ω2, (12) From last equation in (9) the unknowns B(m)12 and B (m) 22 are expressed throught B(m)11 , B (m) 13 , B (m) 21 and B (m) 23 . Thus, we obtain from eqs. (5) and (6) the algebraic sys- tem of equations for unknowns A(m)1 , ..., B (m) 23 . After determining these unknowns we get the expressions for u(m)n13F , σn j13F and ε (m) n j13F from relations (8),(3) and (4). The original unknowns that were sought can now be represented as {u(m)n , σ (m) n j , ε (m) n j } = 1 4π2 ∫ +∞ −∞ ∫ +∞ −∞ {u(m)n13F , σ (m) n j13F , ε (m) n j13F}e i(s1 x1+s3 x3)ds1ds3 (13) The algorithm for calculation of the integral (13) are presented in the papers by (Emiroglu et al., 2004). Therefore we do not consider this algorithm here. 4. Numerical results and discussions The following materials are selected for numerical consideration: Steel (shortly St) with properties ρ0 = 7.86x10−3kg/m3, v = 0.29, c1 = 5890m/s, c2 = 3210m/s; Aluminium (shortly Al) with properties ρ0 = 2.7x10−3kg/m3, v = 0.35, c1 = 6420m/s, c2 = 3110m/s, where ρ0, v, c1 and c2 denote the density, Poisson’s ratio, the speed of dilatation and distortion waves, respectively. The numerical investi- gation is carried out for following two cases: Case I: (Al+St), Layer= Aluminium, Halfspace= Steel, Case II: (St+Al), Layer=Steel, Halfspace=Aluminium. We in- troduce the notation σ22 = σ (1) 22 (0,−h, 0) = σ (2) 22 (0,−h, 0). We shall evaluate the numerical results obtained only for the stress σ22. The validity of the algorithm TWO-AXIALLY PRE-STRETCHED LAYER AND HALF-SPACE 473 and programmes are established as in (Emiroglu et al., 2004). We introduce the parameter η = σ(1),011 /µ (1) for the estimation of the initial stretching of the covering layer, and assume that σ(1),011 > 0, σ (1),0 33 = σ (2),0 11 = σ (2),0 33 = 0 and 0 < Ω ≤ 3. Figures 1 and 2 show the graphs of the dependencies between σ22h2/P0 and Ω for the cases I and II respectively for various values of η. According to these graphs the behavior of the considered system comprises a covering layer and half-space under forced vibration is similar to the forced vibration of the system comprises a mass, a spring and a dashpot. 474 TAŞÇI, EMİROĞLU, AKBAROV The values of Ω for which the absolute values of σ22 have its maximum we name as the “resonance” value of Ω. It follows from the numerical results that the absolute values of σ22 for “resonance” frequency decrease with increasing of initial stretching of the covering layer. 5. Conclusion In the present paper within the framework of the piecewise homogeneous body model with the use of the TLTEWISB the forced vibration of the half-space covered by the bi-axially pre-stretched layer is studied. It is established that the behavior of considered system is similar to the behavior of the system comprises a mass, spring and dashpot. The “resonance” values of the frequency of the point- located external force is determined for the concrete selected materials. It is shown that the absolute values of the interface normal stress arising under “resonance” frequency of the external force decrease with the pre-stretching of the covering layer. References Akbarov S. D., Ozaydin O. (2001a) The effect of initial stresses on harmonic stress field within a stratified half-plane, Europ. J. Mechnanics A/Solids 20, 385-396. Akbarov S. D., Ozaydin O. (2001b) On the Lamb’s problem for a prestressed stratified half-plane, Int. Appl. Mech. 37(10), 138-142. Akbarov S. D., Guler C. (2004) Dynamical (harmonic) interface stress field in the half-plane covered by the pre-stretched layer under a strip load, J.Strain Analysis 40(3), 225-235. Akbarov S. D., Guler C. (2005) Dynamical (harmonic) interface stress field in the half-plane covered by the prestretched layer under a strip load, J. Strain Analysis 40(3), 225-235. Emiroglu I., Tasci T., Akbarov S. D. (2004) Lamb’s problem for a half-space covered with a two- axially pre-stretched layer, Mechanics of Composite Materials 40(3), 227-236. Gladwell G. M. L. (1968) The calculation of mechanical impedances related with the surface of a semi-infinite elastic body, Jour. Sound Vibr. 8, 215-219. Guler C., Akbarov S. D. (2004) Dynamic (harmonic) interfacial stress field in a half-plane covered with a prestretched soft layer, Mechanics of Composite Materials 40(5), 379-388. Guz, A. N. (2004) Elastic waves in bodies with initial (residual) stresses, Kiev: “A. C. K.” -672p. Lamb H. (1904) On the propagation of tremors over the surface of an elastic solid, Philosophical Transactions of the Royal Socety A203, 1-42. 475 475–480. OPTIMIZATION OF VIBRATING ARCHES BASED ON GENETIC ALGORITHM Nildem Tayşi, M. Tolga Göğüş, Mustafa Özakça Department of Civil Engineering, University of Gaziantep, 27310, Gaziantep, Turkey Abstract. The present work is concerned with the structural shape optimization of vibrating arches. Natural frequencies and mode shapes are determined using curved, variable thickness, C(0) conti- nuity Mindlin-Reissner finite elements. The whole shape optimization process is carried out by integrating finite element analysis, cubic spline shape and thickness definition, automatic mesh generation and Genetic Algorithm (GA). An example is included to highlight various features of the optimization procedure. Key words: arches, free vibration, optimization, genetic algorithm 1. Introduction The predominant emphasis and the most significant progress in the field of the optimal design of structures have been concerned with shape optimization in static situations (Tadjbakhsh, 1981; Uzman et al., 1999; Hinton et al., 1992). However, relatively few authors have concentrated on the shape optimization of vibrat- ing structures using GA; see for example resent surveys by (Schittkowski et al., 1994; Arora et al., 1999). One possible explanation is that the dynamic response of structures is often considered of less significance than the static response. Nevertheless, the optimal design of structures subjected to dynamic loading can be extremely useful leading to improved dynamic characteristics. The underlying objective of the present study is to develop a robust and reliable shape optimiza- tion tool for vibrating arches based on Mindlin-Reissner theory and integrating ‘locking free’ finite element analysis, cubic splines to define the arch geometry, automatic mesh generation and GA. The type of optimization considered is maxi- mization of the natural frequency by constraining the volume of the arch material (weight). Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 476 TAYŞİ, GÖĞÜS, ÖZAKÇA Figure 1. Structure chart of genetic algorithm 2. The structure of optimization using genetic algorithm Figure 1 illustrates the structure chart for the GA (Tayşi, 2005). All the necessary data will be read in and the process of the GA will start for the first genera- tion. The initial population will be generated randomly. The constraint violation may be computed so that the objective functions can be modified. By applying some statistics procedures, the average, the maximum and the fittest design will be found, and the convergence criteria will be checked. If the convergence is achieved the GA process will be terminated otherwise the GA process will re- sume by storing the best individual which will be used in the next generation. By producing the mating pool, the creation of the next population is started by applying the crossover operator, and the GA process will proceed continuously until convergence, or the maximum generation is achieved. OPTIMIZATION OF VIBRATING ARCHES BASED ON GA 477 2.1. DEFINITION OF FITNESS FUNCTION AND SELECTION OF OBJECTIVE AND CONSTRAINT FUNCTIONS The objective function is a mathematical function expressed in terms of the design vector s, which quantifies (in a mathematical sense) the worth of any design. It is a criteria which has to be chosen for comparing the different alternative ac- ceptable designs and for selecting the best one. The choice of the objective and constraint functions is governed by the nature of the problem. Typical objective functions f (s) are weight minimization or fundamental frequency maximization under frequency or weight constraint. Weight minimization: Summing for the number of elements ne, the individual masses (the unmodified objective functions) for individual i and population nz are calculated from: f (s)i = ne∑ j=1 ρAl f or : j = 1, . . . , ne and i = 1, . . . , nz (1) Fundamental frequency maximization: In the case of the frequency maximization, objective function f (s) for the structure is computed as the f (s)i = ωp (2) ω is the associated frequency. The details of the volume and fundamental fre- quency computation is presented for arch structures in (Tayşi, 2005). In order to complete the formulation of the problem, some restrictions must be imposed on the values of the design variables for the mathematical model to be meaningful. The constraints in an optimization problem can be geometric con- straints setting a fixed volume for the structure throughout the entire optimization process. Alternatively, there are behavior constraints imposing limiting values on the frequency. Fundamental frequency constraint: the constraint from the fundamental fre- quency for each individual are calculated by: ci = ωi ωinitial (3) where, ωi is the individual fundamental frequency to be compared against the initial fundamental frequency ωinitial. The constraint violation violi is obtained from the fundamental frequency constraints ci: violi = ci − 1.0 (4) where, ωi and individual fundamental frequency must be larger than the initial fundamental frequency values of ωinitial , without incurring constraint violation violi. 478 TAYŞİ, GÖĞÜS, ÖZAKÇA Weight constraint: Structural weight is kept constant by using the target weight WT , which is initial weight of structure. ci = Wi WT f or : Wi > WT (5) ci = WT Wi f or : Wi ≤ WT (6) The constraint violation violi is obtained from the weight constraints ci: violi = ci − 1.0 (7) where, and individual weight Wi must lie within the initial weight values of WT , without incurring constraint violation violi. This constraint is used together with fundamental frequency maximization. Fitness function: After computing objective function f (s) from Eq. 1 or 2 and constraint violations from Eq 3–7, the individual modified objective function f̄ (s)i can be derived: f̄ (s)i = f (s)i + pc ∑ violi 2 (8) Note that the influence of viol on the above modified objective function is con- trolled by an input penalty multiple pc . The maximum and minimum modified objective function ( f̄ (s)i)max and ( f̄ (s)i)min in the population of individuals can then be used to calculate the fitness value of each individuals designs: f iti = f̂i/ fav fav = ∑ndv dv=1 f (s)i nz (9) where: f̂i = f̄ (s)max + f̄ (s)min + f̄ (s)i (10) 3. Geometry definition and mesh generation Among the various types of the curves used to representing shape, cubic splines are the most popular. The cubic spline is the spline of lowest degree with C(2) continuity which meets the needs of most problems arising in practical appli- cations. The detail treatment of cubic spline curves and varieties of cubic spline functions can be found in (Tayşi, 2005). The automatic mesh generator is of prime importance in the automated structural optimization. In the present work, a mesh generator for arches has been developed based on approach of (Sienz, 1994). The mesh generator can generate meshed of two- three- and four noded elements on the natural line of the arch which is defined using parametric cubic splines. Moreover the thickness within the arch are also interpolated from the key points to the nodal points using cubic splines. OPTIMIZATION OF VIBRATING ARCHES BASED ON GA 479 Figure 2. Pinned-pinned arches of continuously varying cross-section 4. Example Problem definition: The following example deals with a continuously varying cross-section arch supported by pin joints at both ends as shown in Figure 2. The arch was originally analyzed (but not optimized) by (Gutierrez and Laura, 1989). The arches are optimized for following cases (a) maximization of the fundamental frequency with a constraint that the total material volume of the structure should remain constant, and (b) volume (or weight) minimization subject to the constraint that the fundamen- tal frequency should remain constant. The location of the design variables and boundary condition for the arch is shown in Figure 2 where ratio of thickness of beginning section to end section tb/te = 0.43 and β = 60◦. The geometry of the arch is modeled using five key points and one segment. Six design variables are considered. These are thick- nesses of five key points and width of arch. The width of the arch at the key points is same. Discussion of results: Table 1 shows initial and optimum values of design variables and objective functions together with bounds on the design variables for cubic thickness variation. For case (a), the problem of fundamental frequency maximization — a 77% increase in the fundamental frequency from 5.58 to 9.69 is obtained. In case (b) involving volume minimization, a reduction of 53% from 104.61 to 49.09 is obtained. Large numbers of thickness design variables, apart from leading to impractical geometries, can sometimes lead to negative thick- nesses between key points. Constraints on the bounds of the design variables are used to guard against negative or zero element thickness. As seen in Table 1 the optimum thicknesses of arch shows that, the symmetry varying of thickness gives better results and upper part of arch thickness is smaller than lower parts. 480 TAYŞİ, GÖĞÜS, ÖZAKÇA TABLE I. Initial and optimum design values of continuously varying arch DV Minumum Initial Maximum Optimum design variables Frequency maximizaiton Weight Minimization t1 0.1 0.6 2.0 1.2302 1.6118 t2 0.1 0.6 2.0 1.9472 0.9711 t3 0.1 1.0 2.0 1.4721 0.8596 t4 0.1 1.2 2.0 1.9853 0.9209 t5 0.1 1.4 2.0 1.5835 1.6508 w1 1.0 1.0 2.0 0.5586 0.4436 Optimum frequency 9.6905 — Optimum weight — 49.0884 5. Conclusions In the present work, computational tools have been developed for the geomet- ric modeling, automatic mesh generation, analysis and GA optimization of arch structures. It is observed that the number of the design variables and the type of thickness variation used for thickness definition greatly affects the amount of improvement and smoothness of the final optimum shape. The optimum thickness variation may not be practical, however, we believe that the tools developed in the present work could be assisted engineering their quest for novel structural forms. References Arora, J.S., Huang, M.W. and Hsieh, C.C. (1999) Methods for Optimization of Nonlinear Problems with Discrete Variables: A Review, Structural Engineering and Mechanics 8, 465–476. Farin, G. (1996) Curves and Surfaces for CAGD, Academic Press, 5 edition. Gutierrez, R. and Laura, P., In-Plane Vibration of Non-Circular Arcs of Non-Uniform Cross Section, Journal of Sound and vibration 129. Hinton, E., Özakça, M. and Jantan, M.H. (1992) A Computational Tool For Determining Optimum Shapes Of Vibrating Arches, Struct. Eng. Review. 4, 162–174. Schittkowski, K., Zillober, C. and Zotemantel, R. (1994) Numerical Comparison of Nonlinear Programming Algorithms for Structural Optimization, Structural Optimization 7, 1–19. Sienz, J. (1994) Integrated Structural Modeling, Adaptive Analysis and Shape Optimization, Ph.D. thesis, University of Wales, Swansea. Tadjbakhsh, I. (1981) Stability and Optimum Design of Arch-type Structures, Int. J. of Solids and Structures 17, 565–574. Tayşi, N. (2005) Analysis and Optimum Design of Structures Under Static and Dynamic Loads, Ph.D. thesis, University of Gaziantep, Gaziantep. Uzman, Ü., Daloğlu, A. and Saka, M.P. (1999) Optimum Design Of Parabolic And Circular Arches With Varying Cross-Section, Structural Engineering and Mechanics 8, 465–476. 481 481–486. PROPAGATION OF LONG EXTENSIONAL NONLINEAR WAVES IN A HYPER-ELASTIC LAYER Mevlüt Teymür Faculty of Science and Letters, İstanbul Technical University, 34390, İstanbul, Turkey Abstract. Propagation of small but finite amplitude waves in a nonlinear hyper-elastic plate of uniform thickness is considered. By employing a perturbation expansion, extensional waves are examined under the long wave limit. It is shown that the asymptotic wave field is governed by a Korteweg-DeVries (K-dV) equation. Then the propagation characteristics of the asymptotic waves are discussed via the well known solutions of the K-dV equation. Key words: nonlinear waves, hyper-elastic, KDV, extensional waves 1. Introduction The propagation of linear elastic waves in wave guides such as rods, plates, lay- ered half spaces, etc. have been studied extensively because of their important applications in certain areas such as geophysics, non-destructive testing of mate- rials, electronic signal processing devices, etc. (Ewing et al., 1957, Farnell, 1978, Morgan, 1998). In these wave propagation problems because of repeated reflec- tion processes which take place at the boundaries of wave guides, waves become dispersive, i.e. the phase velocities of waves depend on the wave number. In recent years the effect of the constitutional nonlinearity on the propagation characteristics of dispersive elastic waves has been the subject of many investiga- tions for somewhat similar reasons mentioned above. By employing the asymp- totic perturbation methods previously used in the fields of fluid mechanics, plasma physics, etc., to investigate the propagation of weakly nonlinear waves (Jeffrey and Kawahara, 1981, Johnson, 1997), several problems related with the propagation of nonlinear dispersive elastic waves are examined (Ahmetolan and Teymur, 2003, Fu, 1994, Mayer, 1995, Teymur, 1988, 1996). In these works as a balance between nonlinearity and dispersion, various scalar nonlinear evolution equations, which were obtained previously in the above mentioned fields, have been derived to describe the wave motions asymptotically. Then several aspects of problems under consideration, such as nonlinear instability of modulated waves, the existence of solitary waves, etc., were discussed on the basis of these equations. Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 482 TEYMÜR In the present work the propagation of small, but finite amplitude extensional waves in a nonlinear elastic plate of uniform thickness is considered. The pertur- bation analysis developed in (Teymur, 1996) is applied to examine the problem. The balance between the nonlinearity and dispersion in the analysis yields a K- dV equation for the asymptotic wave field. Then the propagation characteristics of nonlinear extensional waves are discussed via the well known solutions of the K-dV equation. 2. Formulation of the problem Let (x1, x2, x3) and (X1, X2, X3) be the spatial and material coordinates of a point in space referred to the same rectangular Cartesian system of axes respectively. Consider a plate of uniform thickness 2h occupying the region; D = {(X1, X2, X3)| − h ≤ X2 ≤ h, −∞ < X1 < ∞, −∞ < X3 < ∞} (1) in the reference frame XK and assume that the boundaries X2 = h and X2 = −h are free of traction. Then a wave motion described by the equations; x1 = X1 + u1(X1, X2, t), x2 = X2 + u2(X1, X2, t), x3 = X3 (2) is supposed to propagate along the X1-axis in the plate, where u1 and u2 are displacements in the (X1, X2)−plane in X1 and X2 directions respectively, and t is the time. In the absence of body forces, the equations of motion in the reference state take the following forms; Tβ1,β = ρ0ü1 , Tβ2,β = ρ0ü2 (3) where TKk is the first Piola-Kirchoff stress tensor. The subscripts preceded by a comma indicate partial differentiation with respect to coordinates X1 or X2 and an over-dot represents the partial differentiation with respect to t. The assumption of vanishing of traction on the free surfaces of the plate imposes the boundary conditions; T21 = T22 = 0 on X2 = ±h. (4) The constituent material is assumed to be homogenous, isotropic and compress- ible elastic. Since the propagation of small but finite amplitude waves are consid- ered the stress constitutive relation is taken in the following quadratic nonlinear form; TKk = [ λ(trẼ) + λ2 up,Mup,M + 1 2 (6l + 3m + n)(trẼ) 2 − 12 (m + n)(trẼ 2) ] δKk + [ 2µ − (m + n)(trẼ)]ẼKLδLk + [ λ(trẼ) ] uk,K +2µẼKLuk,L + µup,Kup,LδLk + nẼKN ẼNLδLk (5) PROPAGATION OF LONG EXTENSIONAL NONLINEAR WAVES 483 where Ẽ is the linear Lagrangian deformation tensor with the components EKL = (1/2)(uk,KδLk + uk,LδKk) and λ, µ are the Làme (linear elastic) constants, l,m,n are Murnaghan (nonlinear elastic) constants. 3. Extensional waves We now investigate the propagation of small, but finite amplitude extensional (symmetric) waves by a perturbation method. In the analysis, the following non- dimensional variables are defined x = X1/L, y = X2/h, τ = (c/L)t, u = u1/L, v = u2/h, δ = h/L (6) where c and L denote a characteristic velocity and a characteristic wavelength respectively. Here δ is the ratio of vertical to horizontal length scales. Note that, for extensional waves we have the following symmetry conditions to be satisfied by the displacement functions; u(x, y, τ) = u(x,−y, τ), v(x, y, τ) = −v(x,−y, τ). (7) We consider the unidirectional waves propagating along the positive x−axis gener- ated by a time dependent boundary loading impinged on x = 0. For this boundary value problem to develop a uniformly valid asymptotic solution, we use the as- ymptotic perturbation method given in (Teymur, 1996). As in (Teymur, 1996) we introduce new independent variables x0 = x, x1 = εx, τ = τ, y = y where x1 = εx characterizes the slow spatial variation along the x−axis, and assume that u and v are functions of x0, x1, τ, and y. We now expand u and v in the following asymptotic power series in a small parameter ε > 0 which measures the degree of nonlinearity; u = ∞∑ n=1 εnun(x0, x1, , y, τ) , v = ∞∑ n=1 εnvn(x0, x1, y, τ). (8) Here, we also make an assumption on δ, the ratio of vertical to horizontal length scales, that is δ2 = O(ε) as ε → 0, i.e.ε = δ2 = h2/L2. Hence employing the expansions (8) in the equations of motion (3) and the boundary conditions (4), after using the constitutive relation (5) in them, and then collecting the terms of like powers in ε we obtain a hierarchy of problems from which it is possible to determine un and vn, successively. The solutions of the first and second order problems are found to be; u1 = A(x0, x1, τ) , v1 = ( 1 − 2c2T − c 2 L )∂A ∂ξ y , ξ = τ − x0 (9) u2 = ( 1 − 2c2T − c 2 L )∂2A ∂ξ2 y2 2 + J1(x0, x1, τ) (10) 484 TEYMÜR where J1 is an arbitrary function to be determined in higher order perturbation problems and A satisfies the wave equation; c2 c2T ∂2A ∂τ2 − 4(1 − c2T − c2L )∂2A ∂x20 = 0. (11) If we now choose c2 as c2/c2T = 4 ( 1 − c2T − c 2 L ) (12) the equation (11) reduces to ∂2A ∂τ2 − ∂ 2A ∂x20 = 0. (13) A close examination of (12) reveals that c = √ E/(1 − ν2)ρ0. Hence, c is nothing but the longitudinal thin plate velocity. The solution of this equation for the waves propagating along the positive x−axis is of the form A = A(τ − x0, x1). Note that, at this order the function A could not be determined completely. The result implies that A remains in a frame of reference moving with the unit velocity. The dependence of A on τ−x0 and x1 is determined proceeding to the next order. From the third order problem it is found that J1 satisfies the following inhomogeneous wave equation; ∂2J1 ∂x20 − ∂ 2J1 ∂τ2 = 2 ∂2A ∂ξ∂x1 − Λ0 ∂ ∂ξ (∂A ∂ξ )2 − Λ1 ∂4A ∂ξ4 . (14) Here, while the constant Λ1 depends only on linear elastic constants, Λ0 also depends on nonlinear ones and it becomes zero when the nonlinear effects are neglected. Note that, the solution of this equation will exhibit a secular behavior. But this can be eliminated by equating the terms on the right-hand side of (14) to zero; ∂2A ∂ξ∂x1 − Λ0 2 ∂ ∂ξ (∂A ∂ξ )2 − Λ1 2 ∂4A ∂ξ4 = 0. (15) By doing this not only the uniformity of the asymptotic expansions (8) are main- tained, but also a differential equation for the function A, which was left unde- termined in the previous steps, is deduced. Under the condition (15) the solution of (14) for the waves propagating along the positive x−axis is of the form J1 = J1(τ− x0, x1). Note that, at this order it is only revealed that J1 remains constants in the frame of reference moving with the unit velocity. The function could not have been determined completely. The dependence of on τ − x0 and x1 can be determined in higher order problems. But, since this work is centered around the small but finite amplitude waves it is aimed to obtain just the uniformly valid first PROPAGATION OF LONG EXTENSIONAL NONLINEAR WAVES 485 order solutions u1 and v1. Therefore, to obtain this solution the determination of A as a solution of (15) will be sufficient. Now let us define a function w(ξ, x1) by ∂A/∂ξ = (1/|Λ0|) w, (16) we then write the equation (15) as wx1 − αwwξ − βwξξξ = 0, (17) where α = sgnΛ0 = ±1, β = Λ1/2. This new equation is the K-dV equation and once a solution for w is derived from (17) for a given initial value of the form w(0, ξ) = w0(ξ), then the first order solutions u1 and v1 can be constructed via (16). The initial condition w0(ξ) is related with an applied boundary traction creating a symmetric motion in the plate. Note that, if α = −1, then the substitution w → −w, ξ → −ξ, x1 → x1 in (17) reduces it to the following standard form; wx1 + wwξ + βwξξξ = 0. (18) If α = +1, the substitution w → w, ξ → −ξ, x1 → x1 yields this form. It is known that this equation has the solitary wave solution (Jeffrey and Kawahara, 1981); w = w∞ + a sech 2( √ a/12β ζ), ζ = ξ − (w∞ + a/3)x1, (19) if w → w∞ as ζ → −∞ or +∞. Here w∞ represents the uniform state at infinity and a is the amplitude of the solitary wave. For w∞ = 0, we obtain the displacements as; u1= ( √ 12aβ/|Λ0|) tanhΘ , v1= (1 − 2c2T − c 2 L)(a/|Λ0|) sech 2Θ y, Θ= √ a/12β ζ. They respectively represent a shock wave and a solitary wave propagating along the positive x−axis with the same speed. References Ahmetolan S., Teymur M. (2003) it Nonlinear modulation of SH waves in a two layered plate and formation of surface SH waves, Int. J. Non-Linear Mechs. 38 1237-1250. Ewing W. M., Jardetsky W. S., Press F. (1957) Elastic Waves in Layered Media, McGraw Hill, New York. Farnell G. W. (1978) it Types and properties of surface waves, Acoustic Surface Waves, (Ed: Oliner A. A), 24 13, Springer, Berlin. Fu Y. B. (1994) On The propagation of nonlinear travelling waves in an incompressible elastic plate, Wave Motion 19 271-292. Jeffrey A., Kawahara T. (1981) Asymptotic Methods in Nonlinear Wave Theory, Pitman, Bostan. Johnson R. S. (1997) A Modern Introduction to the Mathematical Theory of Water Waves, Cambridge Univ. Press. Mayer A. P. (1995) Surface acoustic waves in nonlinear elastic media, Physics Reports 256 (4&5). 486 TEYMÜR Morgan D. P. (1998) History of SAW devices, IEEE Intl. Frequency Control Symposium, 3450-3458. Teymur M. (1988) it Non-linear modulation of Love waves in a compressible hyperelastic layered half space, Int. J. Eng. Sci. 26 907-927. Teymur M. (1996) Small but finite amplitude waves in a two-layered incompressible elastic medium, Int. J. Eng. Sci. 34 227-241. 487 487–492. QUENCHING OF SELF-EXCITED VIBRATIONS IN A SYSTEM WITH TWO UNSTABLE VIBRATION MODES ∗ Aleš Tondl,1 Radoslav Nabergoj2 and Horst Ecker3 1Zborovská 41, Prague 5, CZ-l5000 Czech Republic 2University of Trieste, Department of Naval Architecture, Ocean and Environmen- tal Engineering, I-34127 Trieste, Italy 3Vienna University of Technology, Institute of Mechanics and Mechatronics, A-1040 Vienna, Austria Abstract. A self-excited chain system consisting of three-mass is investigated. The equilibrium position of the system is unstable in two vibration modes. Two of the masses are self-excited by a Van der Pol type self-excitation, the third mass is positively damped and parametrically excited due to a stiffness variation of its elastic mounting. The results from numerical simulations show that single mode vibrations prevail, either with the first or the third mode. These vibrations are locally stable and there exist two domains of attraction. Only one mode vibration can be partly or fully suppressed with single-harmonic variation of spring stiffness at parametric anti-resonance frequency. Key words: self-excitation, Van der Pol type, parametric excitation, stiffness variation 1. Introduction This study investigates a three-mass system with nonlinear self-excitation forces acting on two masses. The equilibrium position is unstable for two vibration modes, the first and the third mode, while the second mode is stable. From this system the following vibrations can be expected in principal: single-frequency vi- brations with the first or the third mode, double-frequency vibrations or eventually chaotic vibrations. It is known from (Tondl, 1998) and other references that parametric excita- tion due to a periodic variation of a spring stiffness can be used for vibration quenching. The suppression effect to a certain vibration mode can be achieved at a Parametric Excitation (PE-)frequency given by the difference of the natural frequencies corresponding to the vibrating mode due to self-excitation and to the natural frequency of the stable vibration mode. Single-mode vibration can be even fully cancelled and therefore this phenomenon was called parametric ∗ This work was partly funded by the Austrian Science Fund under Grant P-16248 E. Inan and A. (eds.), Vibration Problems ICOVP 2005, c . © 2007 Springer. K r şıı 488 TONDL, NABERGOJ, ECKER anti-resonance. Several questions arise on what may happen when parametric excitation is applied to the considered system, where a multi-mode instability of the equilibrium position occurs. Special attention will be given to PE-frequencies close to Ω2 − Ω1 (for suppressing the first mode vibration) and to Ω3 − Ω2 (for suppressing the third mode vibration). 2. Equations of motion and basic considerations Figure 1. Schematic representation of the sys- tem with parametric excitation at foundation Let us consider a three-mass system as shown in Figure l. The masses are con- nected by springs of stiffness k j ( j = 1, 2, 3). The upper masses m1, m2 are self-excited, e.g. by a constant flow, which can be described by a Van der Pol model. Mass m3 represents a foun- dation mass being elastically mounted and damped by linear viscous damping. The stiffness of the elastic mounting has constant and periodic components, e.g. k3 = k0 (1 + ε cosωt), which represents the source of parametric excitation. The deflections from equilibrium positions are y j ( j = 1, 2, 3). Thus, the considered system is governed by the following equations m1ÿ1 − ( b1 − d1y21 ) ẏ1 + k1 (y1 − y2) = 0 , m2ÿ2 − ( b2 − d2y22 ) ẏ2 − k1 (y1 − y2) + k2 (y2 − y3) = 0 , (1) m3ÿ3 + b3ẏ3 − k2 (y2 − y3) + k3y3 = 0 . For the sake of simplicity let us assume that k1 = k2 = k0 = k, m1 = m2 = m, m3 = m/2. Using the time transformation ω0t = τ, where ω0 = √ k/m, the following equations are obtained: y′′1 − ( β1 − δ1y21 ) y′1 + y1 − y2 = 0 , y′′2 − ( β2 − δ2y22 ) y′2 + y2 − 1 2 (y1 + y3) = 0 , (2) y′′3 + κy ′ 3 − 1 2 y2 + (1 + ε cos ντ) y3 = 0 , where βk = bk/mkω0, δk = dk/mkω0, for k = 1, 2, and κ = b3/m3ω0, ν = ω/ω0. QUENCHING OF SELF-EXCITED VIBRATIONS 489 It is useful to transform the sets of differential equations (1) and (2) into the quasi-normal form of x′′s + Ω 2 s xs + 3∑ k=1 ( Θsk x ′ k + Qsk xk cos ντ ) = 0 , (s = 1, 2, 3) . (3) For both alternatives this transformation reads: y1 = x1 + x2 + x3 , y2 = √ 3 2 x1 − √ 3 2 x3 , (4) y3 = 1 2 x1 − x2 + 1 2 x3 . In the case when Θ j j is negative and Θss positive, then the vibration mode corresponding to the natural frequency Ω j can be stabilized in the interval, see (Tondl, 1998): ∣∣∣Ω j −Ωs ∣∣∣ − ∆ js < ν < ∣∣∣Ω j −Ωs ∣∣∣ + ∆ js , ( j, s = 1, 2, 3; s � j) (5) where ∆ js = Θ j j + Θss 2 √ |Θ j jΘss| ( Q jsQs j 4Ω jΩs + Θ j jΘss )1/2 . (6) The following conditions must be met to achieve full vibration suppression: Θ j j + Θss > 0 , (7) Q jsQs j 4Ω jΩs + Θ j jΘss > 0 . (8) For the considered system (2) the following relations hold: Θ11 = Θ33 = 1 6 [− (2β1 + 3β2) + κ ] , Θ22 = 1 3 (−β1 + 2κ) , (9) Q12 = Q32 = ε 12 , Q21 = Q23 = ε 6 . (10) From these conditions it is evident, that Θ22 would be the first to become posi- tive when starting from zero and increasing the dimensionless damping parameter κ. Coefficients β1, β2 and κ were chosen such that Θ11 + Θ22 and Θ33 + Θ22 will be of positive value. 490 TONDL, NABERGOJ, ECKER Figure 2. Extreme values [x1], [x2] and [x3] of normal mode vibration amplitudes x1, x2 and x3 versus applied parametric excitation frequency ν for κ = 0.15, β1 = 0.05, β2 = 0.05, δ1 = 0.50, δ2 = 0.50, and ε = 0.25. (Ω1 = 0.37, Ω2 = 1.0, Ω3 = 1.37) 3. Numerical results Numerical simulation results are presented in diagrams showing the extreme val- ues of xk (k = 1, 2, 3) (denoted as [xk]) in dependence of the parametric exci- tation frequency ν. For this ν was increased by small steps, starting from zero up to a maximum value. Also a second run was calculated by decreasing the PE-frequency to check the nonlinear behavior and to look for hysteresis effects. Figure 2 shows a result for a system with parameters as defined in the figure annotation. Note that normal mode deflections are plotted in this and the following diagram. Within a broad interval around ν = Ω3 − Ω2 the single-frequency vibra- tion with the third mode is suppressed, but the first mode vibration is initiated. For QUENCHING OF SELF-EXCITED VIBRATIONS 491 ν � Ω2 − Ω1 the first mode vibration is suppressed, but the third mode vibration is initiated. Strong parametric resonances occur, especially for the first mode x1: the para- metric resonance of the first kind at ν ≈ 2Ω1 and the combination resonances at ν = Ω1 + Ω2 and ν = Ω1 + Ω3 can easily be seen. As mentioned earlier, a system with double component parametric excitation has been investigated also. This system is governed by equations: y′′1 − ( β1 − δ1y21 ) y′1 + y1 − y2 = 0 , y′′2 − ( β2 − δ2y22 ) y′2 + y2 − 1 2 (y1 + y3) = 0 , (11) y′′3 + κy ′ 3 − 1 2 y2 + {1 + ε [cos (Ω2 −Ω1) τ + cos (Ω3 −Ω2) τ]} y3 = 0 . Normal mode deflections [x1], [x2], [x3] have been calculated for ε ranging from zero to (a rather high value of) 0.5. From Figure 3 one can see that with increasing ε also [x1] increases from zero value, but [x3] starts from a non-zero value and decreases. Close to ε = 0.2 [x3] reaches its minimum close to 0.1. Further increase of ε results in vibrations having higher extreme values than for ε = 0. Single-frequency vibrations with the third vibration mode turn into vibrations resulting from more than one vibrational mode. This means that full suppression cannot be achieved. From this it follows that the double-frequency parametric excitation using stiffness variation of one spring cannot fully suppress both modes of vibration. For a large amplitude value parametric excitation can even results in more pronounced vibrations than it is the case without parametric excitation. One possibility of quenching both modes of vibration with a single-frequency parametric excitation would be to tune the system so that the relation Ω2 − Ω1 = Ω3 − Ω2 is met, which may not be an easy matter. Another approach could be a combination of passive means together with parametric excitation such that only one mode would be necessary to quench and therefore the single-frequency parametric excitation would be sufficient. 4. Conclusions In the case when the equilibrium position is unstable in two vibration modes then using a single-frequency parametric excitation can fully suppress only one mode of self-excited vibrations, even when conditions (7) and (8) are met at certain frequency of parametric excitation. When just condition (7) is met, then only partial suppression can be expected. The application of a two-frequency component parametric excitation corre- sponding to different parametric anti-resonances was not successful with respect to full vibration cancelling and needs further investigations. 492 TONDL, NABERGOJ, ECKER 0.0 0.1 0.2 0.3 0.4 0.5 -2.0 -1.0 0.0 1.0 2.0 V ib ra tio n A m pl itu de , [ x 1 ] 0.0 0.1 0.2 0.3 0.4 0.5 -2.0 -1.0 0.0 1.0 2.0 V ib ra tio n A m pl itu de , [ x 2 ] 0.0 0.1 0.2 0.3 0.4 0.5 Parametric Excitation Amplitude , ε -2.0 -1.0 0.0 1.0 2.0 V ib ra tio n A m pl itu de , [ x 3 ] Excitation Down Excitation Down Excitation Up Excitation Down Excitation Up Excitation Up Figure 3. Extreme values [x1], [x2] and [x3] in dependence of the PE-amplitude ε for κ = 0.15, β1 = 0.05, β2 = 0.05, δ1 = 0.50, δ2 = 0.50, and ν1 = Ω2 −Ω1, ν2 = Ω3 −Ω2 References Ecker H., Tondl A. (2000) Suppression of flow-induced vibrations by a dynamic absorber with parametric excitation, Proc. 7th Int. Conference on Flow-Induced Vibrations - FIV. Nabergoj R., Tondl A. (2001) Self-excited vibration quenching by means of parametric excitation, Acta Technica ČSAV 46 107-211. Tondl A. (1998) To the problem of quenching self-excited vibrations, Acta Technica ČSAV 43 109-116. Tondl A. (2002a) Two parametrically excited chain systems, Acta Technica ČSAV 47 67-74. Tondl A. (2002b) Three-mass self-excited systems with parametric excitation, Acta Technica ČSAV 47 165-176. Tondl A., Ecker H. (2003) On the problem of self-excited vibration quenching by means of parametric excitation, Archive of Applied Mechanics 72 923-932. 493 493–498. FREE VIBRATION ANALYSIS OF LAMINATED PLATES USING FIRST-ORDER SHEAR DEFORMATION THEORY Umut Topal and Ümit Uzman Karadeniz Technical University, Department of Civil Engineering, 61080, Trabzon, Turkey Abstract. This paper deals with free vibration analysis of simply supported laminated composite plates using first-order shear deformation theory (FSDT). The displacement field of a laminated composite plate is given for FSDT. The numerical studies are conducted to determine the effect of width-to-thickness ratio, degree of orthotropy, fiber orientation, aspect ratio on the nondimension- alized fundamental frequency for laminated composite plates. Also, the effect of shear deformation, rotatory inertia and shear correction coefficient on the nondimensionalized fundamental frequency is examined. A MATLAB code is written for free vibration of laminated plates. However, the problem is modeled using finite element package program ANSYS for different meshes. Finally, the results are given in graphical and tabular form and compared. Key words: laminated composite plate, first-order shear deformation theory, free vibration analysis, non-dimensional fundamental frequency 1. Introduction The application of laminated fiber reinforced composites as structural plate ele- ments or members has been steadily increasing in civil, aerospace, mechanical and marine structures, etc. due to their advantages of high stiffness and strength- to-weight ratios (Guo et al., 2002). An understanding of the composite plates’ behaviour under dynamic loading is essential because the loading can cause severe damage in composite plates, such as, matrix cracking, fibre cracking and delami- nating (Latheswary et al., 2004). Determining the free vibration characteristics of a structural system often appears to be the fundamental task in dynamic analysis. Many researchers have investigated the vibration of laminated plates in the past (Dai et al., 2004), (Leissa and Narita, 1989), (Afshari and Widera, 2000), (Ray, 2003),(Bert and Chen, 1978), (Dawe and Wang, 1993). Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 494 TOPAL, UZMAN 2. First order displacement fields Based on the first-order shear deformation plate theory, the displacement field of a laminated plate is expressed as u(x, y, z) = uo(x, y) + zφx(x, y) (1) v(x, y, z) = vo(x, y) + zφy(x, y) (2) w(x, y, z) = wo(x, y) (3) The strain-displacement and stress-strain relations for any ply are given by ε = LU, σ = Qε (4) where UT = [u v w φx φy], ε T = [�xx �yy γyz γxz γxy] (5) L = ∂x 0 0 z∂x 0 0 ∂y 0 0 z∂y 0 0 ∂y 0 1 0 0 ∂x 1 0 ∂y ∂x 0 z∂y z∂x ,Q = Q11 Q12 0 0 Q16 Q12 Q22 0 0 Q26 0 0 Q44 Q45 0 0 0 Q45 Q55 0 Q16 Q26 0 0 Q66 (6) in which σi j, εi j, Qi j are the stress, strain components and plane-stress reduced constitutive matrix, respectively. 3. Numerical results and discussions The numerical results are given for laminates having the following properties. E1 = 25x10 6N/m2, E2 = 1x10 6N/m2, ν12 = 0, 25 G12 = G13 = 0, 5x10 6N/m2,G23 = 0, 2x10 6N/m2, ρ = 1kg/m3 3.1. EFFECT OF WIDTH-TO-THICKNESS RATIO Simply supported four-layer cross-ply and angle-ply laminates with symmetric and anti-symmetric arrangement and having different width-to-thickness ratios are analysed. The variation of non-dimensional fundamental frequency is shown in Fig. 1. As seen, the frequency increases abruptly as b/h increases in the thick plate, but the increase is small beyond b/h=20. The value of non-dimensional fundamental frequency becomes practically constant for very thin plates. Also, FREE VIBRATION ANALYSIS OF LAMINATED PLATES 495 symmetric lay-up has higher non-dimensional frequency as compared to anti- symmetric lay-up in the case of cross-ply laminates whereas the reverse is the case with angle-ply laminates. Also in Fig. 1, the non-dimensional fundamental frequency obtained from ANSYS for (10x10 and 15x15 meshes) is shown for (0/90/90/0) laminate. 9 11 13 15 17 19 21 23 0 20 40 60 80 100 120 0/90/90/0 0/90/0/90 45/-45/-45/45 45/-45/45/-45 8 10 12 14 16 0 20 40 60 80 100 120 0/90/90/0,FSDT,RI=0 0/90/90/0,ANSYS,10X10 0/90/90/0,ANSYS,15x15 b/h Figure 1. Variation of non-dimensional fundamental frequency with b/h ratio 3.2. EFFECT OF MATERIAL ANISOTROPY Three-layer cross-ply (0/90/0) and angle-ply (45/-45/45) laminates with b/h=10 and 100 are analysed for this case. An increase of E1/E2 ratio, keeping same E2, leads to an increase in the fundamental frequency in the case of both cross-ply and angle-ply laminates in thin and thick plates (Fig. 2). This is due to the increase in stiffness with increase in E1. The rate of increase is more in the case of thin plates. Also in Fig. 2, the non-dimensional fundamental frequency obtained from ANSYS is shown for (0/90/0)laminate. 7 9 11 13 15 0 10 20 30 40 50 0/90/0,FSDT,RI=0 0/90/0,ANSYS,10x10 0/90/0,ANSYS,15x15 21 E/E 7 10 13 16 19 22 25 28 0 10 20 30 40 50 0/90/0(b/h=10) 0/90/0(b/h=100) 45/-45/45(b/h=10) 45/-45/45(b/h=100) Figure 2. Variation of fundamental frequency with material anisotropy 496 TOPAL, UZMAN 3.3. EFFECT OF FIBRE ORIENTATION Symmetric (α/-α/-α/α) and antisymmetric (α/-α/α/-α) laminates with the fibre orientation varying from 0o to 45o with b/h=10 and 100 are analysed. A change in fibre orientation angle leads to an increase in the fundamental frequency in the (b/h=10) and (b/h=100) plates as shown in Fig. 3. The difference being more for higher values of α . The non-dimensional fundamental frequency obtained from ANSYS is shown for (α/-α/-α/α) laminate (b/h=10) in Fig. 3. 12 14 16 18 20 22 0 15 30 45 /- /- / (b/h=10) /- /- / (b/h=100) /- / /- (b/h=10) /- / /- (b/h=100) 12 13 14 15 16 17 0 15 30 45 /- /- / ,FSDT,RI=0 /- /- / ,ANSYS,10x10 /- /- / ,ANSYS,15x15 Figure 3. Variation of fundamental frequency with fibre orientation angle 3.4. EFFECT OF ASPECT RATIO Cross-ply (0/90/0) laminates with b/h=10 and 100 by varying a/b ratio, keeping the value of ‘b’ constant are analyzed. As seen, the fundamental frequency is found to decrease gradually with increase in aspect ratio (Fig. 4). This is due to the decrease in stiffness of the plate with increasing aspect ratio. In Fig. 4, the non- dimensional fundamental frequency obtained from ANSYS is shown for (0/90/0) laminate (b/h=10). 5 7 9 11 13 15 17 0,8 1 1,2 1,4 1,6 1,8 2 2,2 b/h=10 b/h=100 2 4 6 8 10 12 14 0,8 1 1,2 1,4 1,6 1,8 2 2,2 0/90/0,FSDT,RI=0 0/90/0,ANSYS,10x10 0/90/0,ANSYS,15x15a/b Figure 4. Variation of fundamental frequency with aspect ratio FREE VIBRATION ANALYSIS OF LAMINATED PLATES 497 3.5. EFFECT OF SHEAR DEFORMATION, ROTATION INERTIA AND SHEAR CORRECTION COEFFICIENT In Table 1, the effect of shear deformation, rotatory inertia (RI) and shear co- efficient (K) on nondimensionalized frequencies of simply supported cross-ply (0/90/0) square plates for a/h=10 is given for different m and n modes. (*) corre- sponds to K=1 and the second line corresponds to K=5/6. As seen, nondimension- alized frequencies are higher for classical plate theory (CLPT) than those for first order shear deformation (FSDT) theory for both with RI and without RI cases. The effect of the K is to decrease the frequencies, i.e. the smaller K, the smaller is the frequency. The RI also has the effect of decreasing frequencies. TABLE I. Effect of shear deformation, rotatory inertia and shear coefficient on nondimen- sionalized natural frequencies of simply supported symmetric cross-ply (0/90/0) square plates for a/h=10 m n CLPT CLPT FSDT FSDT w/o RI with RI w/o RI with RI 1 1 15,2278 15,1041 12,5931 12,5269* 12,2235 12,163 1 2 22,8772 22,4209 19,4402 19,2032 18,9422 18,7294 1 3 40,2992 38,7377 32,4963 31,9219 31,4213 30,9322 2 1 56,8855 55,7508 33,0978 32,9316 31.1316 30.9916 8 10 12 14 16 0 20 40 60 80 100 0/90/0,FSDT,RI=0 0/90/0,CLPT,RI=0 0/90/0,FSDT,RI 0 0/90/0,CLPT,RI 0 a/h Figure 5. Effect of transverse shear deformation and rotary inertia on fundamental frequency for a/h 498 TOPAL, UZMAN In Fig. 5, the nondimensionalized fundamental frequency versus side-to thick- ness ratio (a/h) for simply supported symmetric cross-ply (0/90/0) laminates for K=5/6 is given. As seen, the effect of shear deformation decreases with increasing values of a/h. The effect of rotatory inertia is negligible in FSDT, whereas it is significant in CLPT only for very thick plates. 4. Conclusions In this paper, free vibration analysis of laminated plates based on first-order shear deformation theory is presented. The non-dimensional fundamental frequency of vibration is found to increase with increase in width-to-thickness ratio, material anisotropy and angle of fibre orientation. The fundamental frequency of vibration decreases gradually with increase in aspect ratio. The nondimensionalized natural frequencies are higher for classical plate theory (CLPT) than those first order shear deformation (FSDT) theory for both with RI and without RI cases. The effect of the K is to decrease the frequencies, i.e. the smaller K, the smaller is the frequency. The RI also has the effect of decreasing frequencies. The effect of rotatory inertia is negligible in FSDT, whereas it is significant in CLPT only for very thick plates. References Afshari P., Widera G. E. O. (2000) Free vibration analysis of composite plates Transactions of the ASME, 122 390-398. Bert C. W., Chen T. L. C. (1978) Effect of shear deformation on vibration of antisymmetric angle-ply laminated rectangular plates, Int. J. Solids Struct. 14 465-473. Dai K. Y., Liu G. R., Lim K. M., Chen X. L. (2004) A mesh-free method for static and free vibration analysis of shear deformable laminated composite plates, Journal of Sound and Vibration, 269 633-652. Dawe D. J., Wang S. (1993) Free vibration of generally-laminated, shear-deformable, composite rectangular plates using a spline Rayleigh-Ritz method, Structural Mechanics 25 77-87. Guo M., Harik I. E., Ren W. X. (2002) Free vibration analysis of stiffened laminated plates using layered finite element method. Structural Engineering and Mechanics, 14 245-262. Latheswary S., Valsarajan K. V., Rao Y. V. K. S. (2004) Free vibration analysis of laminated plates using higher-order shear deformation theory, IE(I) Journal-AS, 85 18-24. Leissa A. W., Narita Y. (1989) Vibration studies for simply supported symmetrically laminated rectangular plates, Composite Structures, 12 113-132. Ray M. C. (2003) Zeroth-order shear deformation theory for laminated composite plates, Transac- tions of the ASME, 70 374-380. 499 499–504. THE GENERALIZED BOLOTIN METHOD AS AN ALTERNATIVE TOOL FOR COMPLETE DYNAMIC STABILITY ANALYSIS OF PARAMETRICALLY EXCITED SYSTEMS: APPLICATION EXAMPLES Özgür Turhan Faculty of Mechanical Engineering, İstanbul Technical University, Gümüşuyu 34437, İstanbul, Turkey Abstract. Stability analysis of the solutions of linear differential equations with periodic coeffi- cients or Mathieu-Hill equations constitutes a topic of constant research interest. Many methods have been devised for this purpose, among which is the generalized Bolotin method developed by the present author. This study presents some application examples of that method, by comparing them with the results obtained via monodromy matrix method. Key words: Mathieu-Hill equations, stability, infinite determinant method 1. Introduction Many problems of mathematical physics lead to linear ordinary differential equa- tions with periodic coefficients (or Mathieu-Hill equations), which can be written ẍ + 1 ω C(τ)ẋ + 1 ω2 K(τ)x = 0; C(τ) = C(τ + 2π), K(τ) = K(τ + 2π) (1) where τ = ωt, ω = 2π/T , T=Period in t domain. Systems whose behaviour is governed by an equation of form (1) are referred to as parametrically excited systems and the theory of that class of equation is called Floquet theory (Ce- sari, 1959). Parametrically excited systems have special resonance conditions and the important problem of determining these conditions is called dynamic sta- bility analysis. It is customary to differentiate between two kinds of resonance: Parametric resonances encountered in both single and multi-d.o.f systems and combination resonances peculiar to multi-d.o.f ones. The problem of devising comprehensive methods for dynamic stability analysis of Mathieu-Hill equations remained a challenge since more than one century. References to the existing methods may be found in (Nayfeh and Mook, 1979) and (Seyranian and Maily- baev, 2003). We also cite (Guttalu and Flashner, 1996) and (Pernot and Lamarque, 2001). Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 500 TURHAN The generalized Bolotin method is a complete dynamic stability analysis met- hod for non-canonical systems, proposed by the present author (Turhan, 1998). It consists in a generalization of the well-known Bolotin method (Bolotin, 1964), which is an infinite determinant boundary trace method valid for single-d.o.f sys- tems (parametric resonances) only, so as to cover multi-d.o.f systems (combina- tion resonances) too. Here, we present some application results of that method and compare them to the results of another, essentially numerical scatter plot method known as monodromy matrix method. 2. The method According to the Floquet theory, a fundamental solution of Eq. (1) can be written x(τ) = q(τ) · eρτ; q(τ) = q(τ + 2π) (2) where the constant ρ is called a Floquet exponent. Inserting Eq. (2) into Eq. (1), introducing the complex Fourier series expansions x(τ) = eρτ ∞∑ k=−∞ Dkeikτ, C(τ) = m∑ p=−m Cpeipτ, K(τ) = m∑ p=−m Kpeipτ (3) and requiring harmonic balance, one is led to an infinite set of linear homoge- neous algebraic equations for the unknown Fourier coefficients’ vectors Dk, the solvability condition of which yields det [ ρ2I + ρ(E0 + 1 ω E1) + (F0 + 1 ω F1) + 1 ω2 F2) ] = 0 (4) where Ei, Fi’s are infinite dimensional hyper-matrices made up of nxn sub-matrices Ek,q0 = 2ikIδkq, E k,q 1 = Cp F k,q 0 = −k 2Iδkq, F k,q 1 = ikCp, F k,q 2 = Kp (5) where p=k-q, δkq is the Kronecker delta and the superscripts k and q refer to the hyper-row and column indices. Given the value of the Floquet exponent ρ on the stability boundaries, Eq. (5) can be used to calculate the corresponding ω values. It is known that on parametric resonance boundaries a certain s-th exponent takes the value ρs = 0 (a harmonic resonance boundary) or ρs = i2 (a sub-harmonic res- onance boundary), while in a non-canonical system, on a combination resonance boundary a certain pair ρs, ρt of Floquet exponents take values so that ρs + ρt = 0 (Turhan, 1998). Whence, substituting ρ = 0 into Eq. (5), one has for harmonic resonance boundaries det [ F0 + 1 ω F1 + 1 ω2 F2 ] = 0 (6) and substituting ρ = i2 one has for sub-harmonic parametric resonance ones, det [[ F0 + i 2 E0 − 1 4 I ] + 1 ω [ F1 + i 2 E1 ] + 1 ω2 F2 ] = 0. (7) GENERALIZED BOLOTIN METHOD AND APPLICATION EXAMPLES 501 As for the combination resonance boundaries, first linearize the matrix polynomial of Eq. (4) in ρ, to obtain the infinite Hill’s determinant of the problem det [[ U0 + 1 ω U1 + 1 ω2 U2 ] − ρI ] = 0; U0 = [ −E0 −F0 I 0 ] ,U1 = [ −E1 −F1 0 0 ] ,U2 = [ 0 −F2 0 0 ] , (8) then introduce the bialternate sum matrices B(Ui) of the matrices Ui, whose eigen- values are the sums of the eigenvalues of the argument matrix taken in pairs (Fuller, 1968), and write det [ B(U0) + 1 ω B(U1) + 1 ω2 B(U2) ] = 0 (9) as a condition for ρs + ρt = 0. ω values on the stability boundaries can be ap- proximately calculated from finite dimensional portions of the determinants of Eqs. (6), (7) and (9) (obtained by truncating the first Fourier series of Eq. (3) at k=K), by solving an eigenvalue problem. Details of the method and some notes on its implementation can be found in (Turhan, 1998) and (Turhan and Bulut, 2005). Let one be contented here with noting that some of the ω values calculated through Eq. (9) must be eliminated as they correspond to unconverged ρs,t values. 3. Application examples Here we present some application examples of the generalized Bolotin method and compare its results to those of the monodromy matrix method. The former method is implemented by means of a special FORTRAN code and the latter is implemented by using the MATLAB package. As a first example, we consider a system of two coupled damped Mathieu equations depending on a parameter λ, ẍ + [ 0.1 0 0 0.1 ] ẋ + [[ 0.5 0 0 1.5 ] + [ 0.4 λ λ 0.4 ] cos τ ] x = 0, (10) first studied by (Szemplinska-Stupnicka, 1978). That problem has already been considered by the author (Turhan, 1998), but the developments in computer capac- ities made it possible finer approximations to be obtained today. Accordingly, the generalized Bolotin method is implemented by truncating the Fourier series of Eq. (3) at K=30 when determining the parametric resonance boundaries through Eqs. (6) and (7), and at K=6 when determining the combination resonance boundaries through Eq. (9). The result is presented in Fig. 1a as a stability chart constructed on the ω−λ parameter plane, where different shadings are applied for different kinds of instability zones. Fig. 1b depicts the same chart as obtained via the monodromy matrix method. It can be seen that the two charts are in perfect accordance. Notice in Fig. 1a the existance of three summation type combination resonance regions 502 TURHAN located at the vicinites of (ω1+ω2), (ω1+ω2)/2 and (ω1+ω2)/3 where ω1 = √ 0.5 and ω2 = √ 1.5 are the natural frequencies of the system. As a second example we consider the dynamic stability of the viscoelastic coupler of an otherwise rigid, offset slider-crank mechanism with a crank of radius r2 rotating at constant rate ω2, and a slider with mass m4. The coupler is a uniform rod with cross section A, length �, flexural rigidity EJ and mass m3 made of a Kelvin-Voigt material so that the stress (σ)-strain (ε) relationship reads σ = E(ε+ ηε̇). The system is characterized by the dimensionless parameters ω = ω2/ω∗, ζ = ηω∗, λ = m4/m3, µ2 = r2/�, µ = e/� where ω∗ = √ EJ/m3�3 and e is the offset of the mechanism. When discretisized through Galerkin’s method, the system dynamics can be shown to be governed by a set of coupled, damped Hill’s equation whose homogeneous part has the form g̈ + ζ ω Eġ + [ 1 ω2 E + P(λ, µ2, µ, �; τ) ] g = 0 (11) where the matrix P is 2π-periodic in τ = ωt. See (Turhan, 1996) for the details of the formulation. Here, two Galerkin terms are used in discretization so that the system is reduced to a 2-d.o.f one. Figure 2a and b show the stability analysis results on the crank speed-offset plane for a mechanism with � = 0.3 m, µ2 = 0.3, λ = 0.5 and ζ = 0.01. Eight harmonics are considered in expanding the matrix P, and the problem is truncated at K=16. One sees that the results of the two methods agree well. No combination resonance is encountered in this problem, but interesting overlapping of different resonance regions exist. Figure 1. The Szemplinska-Stupnicka problem (a) Generalized Bolotin method (b) Monodromy matrix method GENERALIZED BOLOTIN METHOD AND APPLICATION EXAMPLES 503 Figure 2. Stability of the coupler of a slider-crank mechanism (a) Generalized Bolotin method (\\\ : harmonic, /// : subharmonic resonance region) (b) Monodromy matrix method A number of further examples can be found in (Turhan and Bulut, 2005) where the set of coupled, damped Hill’s equations g̈ + ζ ω Λ4ġ + 1 ω2 [ Λ4 − β20(1 + δ 2 sinτ)2 · (αA + B + I) ] g = 0 (12) is considered. Figs. 3a,b are reproduced from Fig. 8 of that work. In Fig. 3a, a 5-d.o.f model and K=10 are considered for parametric stability analyses of Eqs. (6) and (7), while a 3-d.o.f model and K=3 are taken in the determination of combination resonance regions through Eq. (9). The monodromy matrix method Figure 3. Stability of a rotating beam (a) Generalized Bolotin method (\\\ : harmonic, /// : subharmonic, ≡ : combination resonance region) (b) Monodromy matrix method 504 TURHAN is applied to the 3-d.o.f model. Two strong combination resonance regions, which can be shown to correspond to ω1 + ω2 and ω1 + ω3, are visible on Fig. 3a, and the results of the two methods are seen to be in perfect harmony. 4. Conclusions The Generalized Bolotin method is applied to 2, 3 and partially 5-d.o.f system examples. It is seen that its results check well with those of the monodromy matrix method. The method, possess the obvious advantages of being a boundary tracing method: It is less time consuming and less exposed to the risk of missing narrow regions of a stability chart as compared to scatter plot methods. Also, it provides direct information on the bifurcation values of the considered parame- ters, whose accurate knowledge is important in most applications. As it treats different types of resonance regions separately, the method has also the advantage of pre-classifying the instability regions; a feature which facilitates getting insight into the problems. The limitations of the method come from the high dimension- ality of the involved matrices and of the elimination procedure mentioned in Sec. 2. These make of the combination resonance boundaries calculation problem of Eq. (9) a time-consuming one, although not comparable with the excessive time consumption (and sometimes failure) of the monodromy matrix method. References Bolotin, V. V. (1964) The Dynamic Stability Of Elastic Systems, Holden-Day, San Francisco. Cesari, L. (1959) Asymptotic Behavior and Stability Problems in Ordinary Differential Equations, Springer-Verlag, Berlin. Fuller, A. T. (1968) Conditions for a matrix to have only characteristic roots with negative real parts, Journal of Mathematical Analysis and Applications 23, 71–98. Guttalu, R. S. and Flashner, H. (1996) Stability study of a periodic system by a period-to-period mapping, Applied Mathematics and Computation 78, 123–135. Nayfeh, A. H. and Mook, D. T. (1979) Nonlinear Oscillations, Wiley, New York. Pernot, S. and Lamarque, C. (2001) Journal of Sound and Vibration, A wavelet-Galerkin procedure to investigate time-periodic systems: Transient vibration and stability analysis 245, 845–875. Seyranian, A. P. and Mailybaev, A. A. (2003) Multiparameter Stability Theory With Mechanical Applications, World Scientific, New Jersey. Szemplinska-Stupnicka, W. (1978) The generalized harmonic balance method for determining the combination resonance in the parametric dynamic systems, Journal of Sound and Vibration 79, 581–588. Turhan, Ö. (1996) Dynamic stability of four-bar and slider-crank mechanisms with viscoelastic (Kelvin-Voigt model) coupler, Mechanism and Machine Theory 31, 77–89. Turhan, Ö. (1998) A generalized Bolotin’s method for stability limit determination of parametrically excited systems, Journal of Sound and Vibration 216, 851–863. Turhan, Ö. and Bulut, G. (2005) Dynamic stability of rotating blades (beams) eccentrically clamped to a shaft with fluctuating speed, Journal of Sound and Vibration 280, 945–964. 505 505–510. COUPLING EFFECTS BETWEEN SHAFT-TORSION AND BLADE-BENDING VIBRATIONS IN ROTOR-BLADE SYSTEMS Özgür Turhan and Gökhan Bulut Faculty of Mechanical Engineering, İstanbul Technical University, Gümüşsuyu 34437, İstanbul, Turkey Abstract. Coupling effects between shaft-torsion and blade-bending vibrations in rotor-blade sys- tems are studied. Certain features of the related eigen-analysis problem are underlined. Eigen- analyses are performed for multi-stage rotor-blade system examples. Eigen value-loci-veering phe- nomena are shown to occur, causing substantial departure of the eigen-characteristics from those obtained via an uncoupled analysis. Key words: rotor-blade systems, turbomachinery, coupled vibrations 1. Introduction Due to their very important applications, accurate prediction of the vibration char- acteristics of rotor-blade systems is of crucial importance. The common practice is to consider the vibrations of the shaft and those of the blading separately through uncoupled models. Some exceptions are (Loewy and Khader, 1984), (Crawley et al., 1986), (Chun and Lee, 1996), (Okabe et al., 1991), (Huang and Ho, 1996) and (Chatelet et al., 2005). All these studies imply that coupling between shaft and blade vibrations is of substantial consequences on the vibration characteristics of rotor-blade systems. This study focuses on the coupling between shaft-torsional and blade-(in plane) bending vibrations through an idealized model consisting in a torsionally elastic shaft carrying a number of rigid discs, which in their turn, carry a number of identical blades modelled as Euler-Bernouilli beams (Fig. 1a,b). The equations of motion of this multi-continuous-body system is obtained via a mixed finite element-Galerkin method. The method is introduced in (Turhan and Bulut, 2005) and will briefly be outlined below. The related eigen-analysis problem reveals that the system has two different kinds of normal mode motions. These are referred to as “coupled shaft-torsion-blade-bending modes” and “rigid shaft modes” by the authors. Numerical examples show that the eigen-frequencies pertaining to the coupled shaft-torsion-blade-bending modes are subject to eigenvalue-loci-veering Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 506 TURHAN, BULUT Figure 1. The system phenomena, which causes the vibration characteristics of the system to depart considerably from those obtained via an uncoupled analysis. 2. Modelling and analysis Consider Fig. 1, let the shaft carry p rigid disks, the i-th of which carry, in its turn, ki identical blades undergoing bending vibrations in the rotation plane. Let the left end of the shaft rotate at constant rate Ω0. The equation of the torsional motion of the shaft acted upon by the reaction torque Ti of the i-th disk, can be written via finite element method as ρ0Ip�0 · M · d2θ dt2 + GIp �0 · K · θ = p∑ i=1 esi ·Ti, (1) where ρ0=mass density, G=shear modulus, Ip=polar area moment of inertia, �0= length of the shaft, θ is the 2n-dimensional nodal coordinates vector, e j is the j-th 2n-dimensional unit vector, si is the attachment station of the i-th disk, and M and K are the numerical parts of the 2nx2n, global mass and stiffness matrices. Next consider the i-th rigid disk acted upon by the reaction torque −Ti of the shaft and the reaction forces −Fi j and torques −Mi j; j = 1, 2, . . . , ki of the ki blades it carries. Its equation of motion reads Ji · (eTsi · d2θ dt2 ) = −Ti − ki∑ j=1 (riFi j + Mi j), (2) where ri is the radius and Ji the mass moment of inertia of the disk. Finally consider the j-th blade (beam) of the i-th disk. It can be shown that a linearized and non-dimensionalized equation of motion of its in-plane, transverse vibrations COUPLING EFFECTS IN ROTOR-BLADE SYSTEMS 507 relative to the frame B; xi jyi jzi j attached to the i-th rigid disk can be given as v̈i j +η 2 i ·vivi j −β 2{[αi(1−ui j)+0.5(1−u2i j)]v ′′ i j − (αi +ui j)v ′ i j + vi j}+ (αi +ui j) · eTsi · θ̈ = 0 (3) where ρi= mass density, (EI)i= flexural rigidity, Ai= cross section area and �i= length of the beam and ui j = xi j �i , vi j = yi j �i , τ = ω∗1t, αi = ri �i , β = Ω0ω∗1 , ω∗i =√ (EI)i ρiAi�4i , ηi = ω∗i ω∗1 with overdots and primes representing differentiation w.r.t. τ and ui j, respectively. Equation (3) may be approximated by a finite set of or- dinary differential equations by means of Galerkin’s method. To this end, in- troduce vi j(ui j, τ) = m∑ k=1 gi jk(τ) · ϕk(ui j), use the eigen-functions of a stationary cantilever as the set of comparison functions ϕk(ui j), follow the usual procedures of Galerkin’s method to obtain in matrix-vector form g̈i j(τ) + [η2i · Λ4 − β2 · (αi · A + B + I)] · gij(τ) + (αi · c + d) · eTsi · θ̈(τ) = 0, (4) where Λ, A and B are mxm matrices and c and d are m-dimensional vectors whose elements are defined as Ars = ∫ 1 0 [(1−ui j)ϕ ′′ r −ϕ ′ r]·ϕs·dui j, Brs = ∫ 1 0 [0.5(1−u2i j)ϕ ′′ r − ui jϕ ′ r] · ϕs·dui j, Λrs = λrδrs, cr = ∫ 1 0 ϕr·dui j, dr = ∫ 1 0 ui jϕr·dui j, where λr’s are the dimensionless eigenfrequencies of a cantilever, δrs is the Kronecker delta, and gi j(τ) is the m-dimensional Galerkin coordinates vector of the related blade. Let one also note that the shearing force Fi j and bending moment Mi j exerted to the j-th beam by the i-th disk can be calculated as Fi j(τ) = −2 (EI)i �2i eT gi j(τ), Mi j(τ) = −2 (EI)i �i fT gi j(τ), (5) where e and f are m-dimensional vectors with elements er = κrλ3r , fr = λ 2 r with κk = (cosh λk + cos λk)/(sinh λk + sin λk). To synthesize the equation of motion of the whole system, eliminate first Fi j, Mi j and Ti between Eqs. (1), (2) and (5) to obtain, after amening to the time scale τ and non-dimensionalizing M + p∑ i=1 γi · esi eTsi · θ̈(τ) + µ2Kθ(τ) − p∑ i=1 2δi·esi ( αi·eT + fT ) · ki∑ j=1 gi j(τ) = 0 (6) where γi = Ji ρ0Ip�0 , δi = ρiAi�3i ρ0Ip�0 , µ = ω∗0 ω∗1 , ω∗0 = √ G ρ0� 2 0 . Combining now Eqs. (4) and (6) into a single hyper-matrix-vector equation one obtains finally M00 0 0 · · · 0 M10 I 0 · · · 0 M20 0 I · · · 0 ... ... ... . . . ... Mp0 0 0 · · · I θ̈ g̈1 g̈2 ... g̈p + K00 δ1K01 δ2K02 · · · δpK0p 0 K11 0 · · · 0 0 0 K22 · · · 0 ... ... ... . . . ... 0 0 0 · · · Kpp θ g1 g2 ... gp = 0 0 0 ... 0 (7) 508 TURHAN, BULUT where gi = {gTi1 g T i2 · · · g T iki }T , Kii = blockdiag[Kii], Mi0 = [MTi0 M T i0 · · · MTi0] T , K0 j = [K0 j K0 j · · · K0 j], with M00 = M + p∑ i=1 γi · esi eTsi , Mi0 = (αi · c + d) · e T si , K00 = µ 2K, K0 j = −2es j ( α j · eT + fT ) , Kii = η2i · Λ4 − β2 · (αi · A + B + I) . (8) It can be shown that Eq. (7) leads to the factorized frequency equation det A00 ∆1B01 ∆2B02 · · · ∆pB0p C10 ∆1D11 + K11 ∆2B12 · · · ∆pB1p C20 ∆1D21 ∆2D22 + K22 · · · ∆pB2p ... ... ... . . . ... Cp0 ∆1Dp1 ∆2Dp2 · · · ∆pDpp + Kpp − σ2I · p∏ i=1 { det [ Kii − σ2I ]}ki−1 = 0 (9) where ∆i = kiδi and A00 = M−100 K00, B0 j = M −1 00 K0 j, Ci0 = −Mi0A00, Di j = −Mi0B0 j. Eq. (9) shows that the eigen-analysis problem splits into p+1 indepen- dent sub-problems corresponding to two different class of normal mode motions of the system. These are referred to as coupled shaft-torsion-blade-bending modes (first determinant in Eq. (9)) and rigid shaft modes (the remaining p determinants) by the authors. It can be shown that, in the rigid shaft modes, the entire shaft and the blades of all the disks, except one, are at rest and the blades of the remaining disk vibrate in such a manner that their total effect on the disk vanishes. 3. Numerical examples Let first a rotor-blade system with four identical and equally spaced stages be considered where µ = 5, αi = 0.5, γi = 0.1, ∆i = 0.4, ηi = 1; i = 1, 2, . . . , 4. Fig. 2a shows the variation of the eigenfrequencies with the dimensionless rotation speed β as obtained from an uncoupled analysis, while Fig. 2b depicting the results of the coupled analysis outlined above. The horizontal lines in Fig. 2a are the eigen-frequencies of the shaft and the rising lines are those of the blades. Comparing Fig. 2a and b one notes that the uncoupled blade eigen-frequencies also subsist in the coupled case. These frequencies correspond to the rigid shaft modes and are p(ki − 1)-fold, as implied by Eq. (9). The other frequencies on Fig. 2b are those of the coupled shaft-torsion-blade-bending modes. It can be seen that these are subject to eigenvalue-loci-veering phenomena, which makes the coupled frequencies to depart considerably from the uncoupled ones. Next consider a rotor-blade system with four different,equally spaced stages where µ = 5, α1 = 0.4762, α2 = 0.5, α3 = 0.4545, α4 = 0.4348, ∆1 = 0.4, COUPLING EFFECTS IN ROTOR-BLADE SYSTEMS 509 Figure 2. Eigen-frequencies versus rotation speed:Four identical stages (a, uncoupled; b, coupled) ∆2 = 0.4631, ∆3 = 0.5324, ∆4 = 0.6083, η1 = 1, η2 = 0.9070, η3 = 0.8264, η4 = 0.7561, γ1 = γ2 = γ3 = γ4 = 0.1. The variation of the eigen-frequencies with β is given in Fig. 3a and b. It can be seen that the figures are similar to Figs. 2a and b except there are now four different blade frequencies at each mode, this being a result of the fact that the blades of different stages are now differing. Finally consider an example with four identical stages where µ = 5, β = 10, γi = 0.1, ∆i = 0.4, ηi = 1; i = 1, 2, . . . , 4 and study the variation of the eigen-frequencies with the dimensionless disk radii α. The results are presented in Figs. 4a and b where negative ranges of α, which correspond to systems where the blades are oriented towards the centre of rotation, are also considered. It can be seen that the blade frequencies vanish at a certain negative value of α. This means buckling of the blade, due to compressive effect of the centrifugal forces. Furthermore, the shaft and blade frequencies are again seen to be involved in loci- veering phenomena, which makes of the coupled frequency loci something very different than the uncoupled ones. Figure 3. Eigen-frequencies versus rotation speed:Four different stages (a, uncoupled; b, coupled) 510 TURHAN, BULUT Figure 4. Eigen-frequencies versus disk radius:Four identical stages (a, uncoupled; b, coupled) 4. Conclusions Coupling effects between shaft-torsional and blade-bending vibrations of multi- stage rotor-blade systems are studied. It is shown that the system has two different kind of normal mode motions. These are referred to as rigid shaft modes and coupled shaft-torsion-blade-bending modes by the authors. The coupled shaft- torsion-blade-bending modes are shown to be subject to eigenvalue-loci-veering phenomena. It is concluded that coupling must be taken into account if accuracy is required in the calculations. References Chatelet, E., D’Ambrosio, F., and Jacquet-Richardet, G. (2005) Toward global modelling ap- proaches for dynamic analyses of rotating assemblies of turbomachines, Journal of Sound and Vibration 282, 163–178. Chun, S. B. and Lee, C. W. (1996) Vibration analysis of shaft-bladed disk system by using substructure synthesis and assumed modes method, Journal of Sound and Vibration 189, 587–608. Crawley, E. F., Ducharme, E. H., and Mokadam, D. R. (1986) Analytical and experimental investi- gation of the coupled bladed disk/shaft whirl of a cantilevered turbofan, Journal of Engineering for Gas Turbines and Power 108, 567–576. Huang, S. C. and Ho, K. B. (1996) Coupled shaft-torsion and blade-bending vibrations of a rotating shaft-disk-blade unit, Journal of Engineering for Gas Turbines and Power 118, 100–106. Loewy, R. G. and Khader, N. (1984) Structural Dynamics of Rotating Bladed-Disk Assemblies Coupled with Flexible Shaft Motions, AIAA Journal 22, 1319–1327. Okabe, A., Otawara, Y., Kaneko, R., Matsushita, O., and Namura, K. (1991) An equivalent reduced modelling method and its application to shaft-blade coupled torsional vibration analysis of a turbine-generator set, Journal of Power and Energy 205, 173–181. Turhan, Ö. and Bulut, G. (2005) Linearly coupled shaft-torsional and blade-bending vibrations in multi-stage rotor-blade systems, Journal of Sound and Vibration (Submitted). 511 511–518. FORCED VIBRATION OF THE PRE-STRETCHED SIMPLY SUPPORTED STRIP CONTAINING TWO NEIGHBOURING CIRCULAR HOLES Nazmiye Yahnioğlu and Surkay D. Akbarov Yıldız Technical University, Department of Mathematical Engineering, Davutpaşa Campus, Topkapı, İstanbul, Turkey Abstract. Within the framework of the Three-Dimensional Linearized Theory of the Elastic Waves in Initially Stressed Bodies under plane-strain state the influence of the initial stretching of the sim- ply supported plate strip containing two circular holes on the stress concentration around the holes caused by the action of the additional uniformly distributed normal dynamical (time-harmonic) forces on the plane of its upper face is studied. For the solution of the problem the FEM is employed. The numerical results on the stress concentration and the influence of the statical initial stresses and the frequency of the additional dynamical external forces frequency to this concentration are presented. Key words: dynamical (time-harmonic) stress field, initial stresses, plate-strip, forced vibration 1. Introduction Investigation on the stress concentration around holes has a wide range of applica- tions in almost all branches of modern industry. There are many monographs, such as Savin (1951) and Savin (1968), that contain the results of these investigations. A review of the above-mentioned research has been given in references Kosmodumi- anskii (2002), Maksimyuk et al. (2003), Lei et al. (2001) and Hyde et al. (2003). It should be noted that these investigations are being studied continuously by many researches at present. It follows from the analysis of the above-mentioned investigations that among those there are no studies on the influence of the initial stresses arising as a result of an initial stretching or another type of initial loading, on the stress concentration caused by an additional loading in the case in which the superposition principle is not applicable. Here under non-applicability of the superposition principle it is understood that the stress field caused by additional loading depends signif- icantly on the initial loading. The theoretical investigation of the phenomenon corresponding to these cases requires the use of complicated geometrical non- linear equations of mechanics of the deformable body. However, according to well-known mechanical considerations for the cases where the magnitude of the Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 512 YAHNİOĞLU, AKBAROV Figure 1. The geometry of the considered strip. initial loading is greater than that of the additional loading, these investigations can be carried out within the framework of the Three-Dimensional Linearized Theory (TDLT) (Biot (1965) and Guz (1999)) of the deformable body. In the paper Akbarov et al. (2005), an attempt was made in this field and the influence of the initial tension of a simply supported strip containing a rectangular hole on the stress concentration around the hole caused by bending of the strip under the action of the uniformly distributed normal forces on the plane of its upper face. However, in Akbarov et al. (2005) it was assumed that the initial and additional loading are the statical ones. In the present paper the investigation Akbarov et al. (2005) is developed for the case where the additional loading is dynamical (time-harmonic) one and the strip contains two neigbouring circular holes. Throughout the investigations, repeated indices are summed over their ranges. 2. Formulation of the problem Consider a plate strip containing two neighboring circular holes. The geometry of the plate strip is shown in Fig. 1. The Cartesian coordinate system Ox1x2x3 is associated with the strip so as to give Lagrange coordinates in the initial state. As- sume that the plate strip occupies the region, {0 ≤ x1 ≤ �, 0 ≤ x2 ≤ h, −∞ < x3 < +∞}, and the axis Ox3 is directed normal to the plane of Fig. 1. Suppose that the material of the strip is orthotropic with principal axes Ox1,Ox2 and Ox3. Moreover, assume that the strip is simply supported at the ends and in the initial state the uniformly distributed normal stretching forces with intensity q act on these ends. The additional uniformly distributed dynamical (time-harmonic) normal forces with amplitude p (� q) act on the plane of the strip’s upper face. The influence of the initial stretching of the strip the stress concentration around the holes caused by the additional dynamical forces with amplitude p is investi- gated below. Henceforth all the quantities referred the initial state will be labelled by the superscript (0). FORCED VIBRATION OF THE PRE-STRETCHED STRIP 513 According to the above, the initial stress state can be determined by the solu- tion of the boundary value problem ∂σ(0)i j ∂x j = 0, σ(0)11 = A11ε (0) 11 + A12ε (0) 22 , σ(0)22 = A12ε (0) 11 + A22ε (0) 22 , σ (0) 12 = 2A66ε (0) 12 , ε(0)i j = 1 2 ∂u(0)i ∂x j + ∂u(0)j ∂xi , u (0) 2 ∣∣∣∣ x1=0;� = 0, σ(0)11 ∣∣∣∣ x1=0 = σ(0)11 ∣∣∣∣ x1=� = q, σ(0)i1 ∣∣∣∣ x2=0;h = 0, σ(0)i j n j ∣∣∣∣ Lk = 0, i; j = 1.2; k = 1, 2 L1 = { (x1, x2)| (x1 − (�E + R))2 + (x2 − (HA + R))2 = R2 } L2 = { (x1, x2)| (x1 − (� − (�E + R)))2 + (x2 − (HA + R))2 = R2 } (1) In equation (1) the conventional notation is used. Moreover, in equation (1) L1 and L2 denote the contour of the holes, and n j are the component of the unite normal vector to the contours Lk . To determine the stress state caused by additional dynamical (time-harmonic) loading the following boundary value problem must be solved: ∂ ∂x j ( σ ji + σ (0) in ∂ui ∂xn ) = ρ0 ∂2ui ∂t2 σ11 = A11ε11 + A12ε22, σ22 = A12ε22 + A22ε22, σ12 = 2A66ε12, εi j = 1 2 ( ∂ui ∂x j + ∂u j ∂xi ) , u2|x1=0;� = 0, ( σ11 + σ (0) 1n ∂u1 ∂xn )∣∣∣∣∣∣ x1=0;� = 0, σi2|x2=h = p e iωtδ2i , σi2|x2=0 = 0, δ21 = 0, δ 2 2 = 1, ( σ ji + σ (0) in ∂ui ∂xn )∣∣∣∣∣∣ Lk n j = 0, i; j; n = 1.2; k = 1, 2 (2) In (2) the corresponding equations and relations of the Three-Dimensional Lin- earized Theory of Elastic Waves in Initially Stressed Bodies (TLTEWISB) are written. 514 YAHNİOĞLU, AKBAROV 3. FEM modelling For the FEM modelling of the boundary value problem (1) the functional Π(0) = 1 2 � Ω− ( ΩL1 ∪ΩL2 ) σ(0)i j ε (0) i j dx1 dx2 − ∫ h 0 q u(0)1 ∣∣∣∣ x1=0 dx2 − ∫ h 0 q u(0)1 ∣∣∣∣ x1=� dx2 (3) is used, where Ω = {0 � x1 � �, 0 � x2 � h} and ΩLk is the region occupied by the holes. Using the virtual work principle and employing the well-known Ritz technique, FEM modelling of the problem (1) is obtained from the equation δΠ(0) = 0. In this case the region Ω − (ΩL1 ∪ ΩL2 ) is divided into certain num- ber rectangular Lagrange family quadratic elements. For around of the holes, the curvilinear triangular finite elements with six nodes are used. The selection of NDOF values follows from the requirements that the boundary conditions should be satisfied with very high accuracy and the numerical results obtained for various NDOFs should converge. For the FEM modelling of the problem (2) the variational principle of TDLT of elasticity given in Akbarov et al. (2005) is used. In this case, first, the sought val- ues are presented as g(x1, x2, t) = g(x1, x2) eiωt and for computed values g(x1, x2) the functional Π = 1 2 � Ω− ( ΩL1 ∪ΩL2 ) ( Ti j ∂u j ∂xi + ρ0 ω 2 u j u j ) dx1 dx2 − ∫ � 0 p u2|x2=h dx1 (4) is used where Ti j = σi j + σ (0) in ∂ui ∂xn (5) With the same finite elements and with the same arrangements as used for the FEM modelling of the problem (1), the FEM modelling for the problem (2) is formulated from the equation δΠ = 0 . The validity of the algorithm and PC programs are tested by the corresponding known results. 4. Numerical results and discussion Consider the case where the geometry of the strip has a symmetry with respect to x1 = �/2 and x2 = h/2 . Assume that the material of the strip is a composite consisting of a large number of alternating layers of two materials. Suppose that the material of each layer is isotropic and these layers are located on the planes x2 = costant. In the following, the values related to the matrix and to the reinforc- ing material will be indicated by the superscripts (1) and (2) respectively.E(k) are FORCED VIBRATION OF THE PRE-STRETCHED STRIP 515 TABLE I. The values of the fundamental frequency Ω2 = ω 2 ρ �2 A22 for various and E(2)/E(1) E(2)/E(1) q/E(1) 0.000 0.0001 0.001 0.002 0.003 0.005 0.007 0.010 1 0.0367 0.0375 0.0442 0.0516 0.0590 0.0739 0.0888 0.1111 5 0.0653 0.0657 0.0698 0.0742 0.0786 0.0875 0.0964 0.1098 10 0.1077 0.1082 0.1118 0.1159 0.1200 0.1281 0.1362 0.1484 20 0.1893 0.1897 0.1932 0.1970 0.2009 0.2087 0.2164 0.2281 Figure 2. The influence of the initial stretching to the stress concentration Figure 3. The influence of the frequency on the stress concentration 516 YAHNİOĞLU, AKBAROV Young moduli, are ν(k) are Poisson’s ratios and η(k) are the concentrations of the components in the representative pack. It is known that in the considered case the material of the representative pack (or the material of the strip) within the frame- work of the continuum approach can be modelled as homogeneous transversal isotropic and its effective mechanical constants Ai j are calculated according, for example, to references Akbarov et al. (2000) and Cristensen (1979). Assume that ν(1) = ν(2) = 0.3, η(1) = η(2) = 0.5, h/� = 0.075, R/� = 0.00625. Introduce the dimensionless frequency Ω2 = ω 2 ρ �2 A22 and consider the fundamental values of this frequency obtained for various E(2)/E(1) and q/E(1). Note that the values of q/E(1) characterize the magnitude of the initial stretching. The values of the fundamental frequency are given in Table I and these values does not depend on the distance between the holes, i.e. on the c/R (Fig. 1). According to Table I, the values of the fundamental frequency (denote it by Ω∗ ) increase with q/E(1) and E(2)/E(1). Below we will assume that Ω < Ω∗ and will analyse the concentration of the stress σθθ(in the corresponding Orθx3 cylindrical coordinate system (Fig. 1)). Figure 2 shows the influence of the initial stretching, i.e. of the q/E(1) to the values of σθθ/p at r = R for the case where c/R = 2, E(2)/E(1) = 5, Ω2 = 0.008 . It follows from the graphs that the absolute maximum values of σθθ/p decrease with q/E(1). Figure 3 shows the influence of the frequency Ω to the values of σθθ/p which increase monotonically with Ω. The numerical results given in Table II shows the influence of the interaction between the holes to the values of σθθ/p. According to these results, it can be concluded that the influence of the initial stretching of the strip to the values of the σθθ/p becomes more significantly with decreasing distance between the holes. 5. Conclusion In the present paper within the framework of the TLTEWISB under plane-strain state the influence of the initial stretching of the simply supported plate strip containing two circular holes on the stress-concentration around the holes caused by the action of the additional uniformly distributed normal dynamical (time- harmonic) forces on the plane of its upper face has been studied. For the solution of the problem the FEM was employed. As a result of the numerical investigations the following were established: 1. The absolute maximum values of the σθθ/p decrease with initial stretching. 2. An increase in the values of the frequency of the external forces causes to increase in the values of the |σθθ|/p. FORCED VIBRATION OF THE PRE-STRETCHED STRIP 517 3. The influence of the pre-stretching of the stripto the dynamical(time-harmonic) stress concentration σθθ/p becomes more significantly with decreasing of the distance between the holes. TABLE II. The values of σθθ/p obtained under E(2)/E(1) = 5, Ω2 = 0.008 for various c/R, q/E(1) and θ θ c/R q/E(1) 0.00 π/8 π/4 3π/8 π/2 5π/8 3π/4 7π/8 π 18 0 -3.0404 21.8993 24.6854 -62.1360 -94.3030 -82.4453 -7.5757 5.2369 -3.0433 0.005 -3.0793 14.8341 16.6586 -44.8140 -67.1678 -58.4656 -5.0526 3.0875 -1.8390 12 0.000 -3.0194 19.4773 19.8129 -65.8340 -95.0808 -79.9433 -2.5784 7.9169 -3.0311 0.005 -3.0443 13.0943 13.3289 -47.2814 -67.6756 -56.7577 -1.7103 4.9548 -2.1469 6 0.000 -3.0014 16.3132 13.9976 -69.3742 -95.1827 -76.6646 2.8040 10.7257 -3.0183 0.005 -3.013 10.8343 9.34482 -49.6476 -67.7292 -54.5314 1.8823 6.9045 -2.4642 4 0.000 -3.0014 16.3132 13.9976 -69.3742 -95.1827 -76.6646 2.8040 10.7257 -3.0183 0.005 -3.0130 10.8343 9.34482 -49.6476 -67.7292 -54.5314 1.8823 6.9045 -2.4642 2 0.000 -3.2190 9.4705 7.5990 -68.8237 -92.2121 -72.0739 7.2436 12.8491 -3.1550 0.005 -3.2209 5.8890 4.8470 -49.1649 -65.6003 -51.3561 4.8437 8.3373 -2.8182 1 0.000 -3.7280 4.1388 6.6393 -65.1413 -89.8233 -70.2884 8.1412 13.2205 -3.3056 0.005 -3.7218 1.9806 4.1412 -46.5000 -63.8640 -50.0935 5.4089 8.5467 -3.0206 References Akbarov, S.D., Guz, A.N. Mechanics of Curved Composites, (Kluwer, Dordrecht, The Netherlands 2000). Akbarov, S.D., Yahnioglu, N., Yucel, A.M. (2005) On the influence of the initial tension of a strip with a rectangular hole on the stress concentration caused by additional loading. J. Strain Analysis, 39(6), 615-624. 518 YAHNİOĞLU, AKBAROV Biot, M.A. Mechanics of Incremental Deformations, (John Wiley, New York 1965). Cristensen, R.M. Mechanics of Composite Materials, (John Wiley, New York 1979). Guz, A.N. Fundamentals of the Three-dimensional Theory of Stability of Deformable Bodies, (Springer-Verlag, Berlin, Germany 1999). Hyde, T.H., Notarangelo, G., Pappalettero, C., Sun, W. (2003) Stress concentration factors in square and circular cross-sectioned bars with axial slots. J. Strain Analysis, 38(3), 247-259. Kosmodumianskii, A.S. (2002) Accumulation of internal energy in multiply connected bodies. Int. Appl. Mechanics, 38(4), 399-422. Maksimyuk, V., Mulyar, V.P., Chernyshenko, I.S. (2003) Stress state of thin spherical shells with an off-centre elliptic hole. Int. Appl. Mechanics, 39(5), 595-598. Lei, G.H., Cherles, W.W. Ng., Rigby, D.B. (2001) Stress and displacement around an elastic artificial rectangular hole. J. Engng. Mechanics, 127(9), 880-890. Savin, G.N., Stress Concentration Around Holes, (Pergomon, Oxford 1951). Savin, G.N. Stress Distribution Around Holes (in Russian), (Naukova Dumka, Kiev, Ukraine 1968). 519 519–523. FREE VIBRATION OF CURVED LAYERED COMPOSITE BEAMS Mustafa Yavuz and M. Ertaç Ergüven Faculty of Civil Engineering, İstanbul Technical University, İstanbul, Turkey Abstract. In practice, fibrous and layered composite beams have periodically and locally curved layers because of the design considerations and manufacturing processes. In this study, the effect of these curvatures and composite material properties to the natural frequencies of the beams is in- vestigated. The periodically curved layered composite material of the considered beam is modelled with the use of the continuum theory proposed by Akbarov and Guz. The free vibration problems are solved by employing the finite element method. Obtained natural frequencies of the beams are presented for the different parameters of the curvature, modulus of elasticity and support condition of the beams. For the case that the ratio of the modulus of elasticity of the layers equals to one and the parameter of the curvature equals to zero, the results converge to natural frequencies of a classical Euler-Bernoulli beam. Results are in good agreement with the literature. Key words: composite beam, finite element method, natural frequency, periodically curved layers 1. Introduction Free vibration analysis of the periodically curved layered unidirectional fibrous composite beams is performed. Model of the imperfection of the layers is based on the continuum theory presented in (Akbarov and Guz, 1991). In literature, Akbarov determined normalized nonlinear mechanical properties of composite materials with periodically curved layers (Akbarov, 1995). Kutug studied free vibrations and buckling loads of composite plates with curved layers (Kutug, 1997). Free vibration analysis and buckling loads of cracked beams resting on an elastic foundation is presented by (Kocer, 1997). Yokoyama studied parametric instability of Timoshenko beams resting on an elastic foundation (Yokoyama, 1988). 2. Mechanical properties of composite materials with curved layers The normalized constants are as follows: A011 = µ1η1+µ2η2+(µ1+λ1)η1+(µ2+λ2)η2−η1η2 (λ1 − λ2)2 (λ1 + 2µ1)η2 + (λ2 + 2µ2)η1 (1) Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 520 YAVUZ, ERGÜVEN A012 = λ1η1 + λ2η2 − (λ1 − λ2)η1η2 [(λ1 + 2µ1) − (λ2 + 2µ2)] [(λ1 + 2µ1)η2 + (λ2 + 2µ2)η1] (2) A022 = (λ1 + 2µ1)η1 + (λ2 + 2µ2)η2 − η1η2 [(λ1 + 2µ1) − (λ2 + 2µ2)]2 [(λ1 + 2µ1)η2 + (λ2 + 2µ2)η1] (3) A066 = µ1µ2 µ1η2 + µ2η1 (4) The main constants are A11(x) = A 0 11Φ 4 1 + 2(A 0 12 + 2A 0 66)(Φ1Φ2) 2 + A022Φ 4 2 (5) A12(x) = (A 0 11 + A 0 22 − 4A 0 66)(Φ1Φ2) 2 + A012(Φ 4 1 + Φ 4 2) (6) A22(x) = A 0 11Φ 4 2 + 2(A 0 12 + 2A 0 66)(Φ1Φ2) 2 + A022Φ 4 1 (7) Functions representing periodically curved layers are given below. Φ1(x) = 1√ 1 + [a(x)]2 (8) Φ2(x) = Φ1(x1)a(x) (9) a(x) = � ∂ f (x) ∂x (10) � f (x) = �sin(γπx + δ) (11) γ = 10 δ = π 2 ν = 0.2 (12) The stress-strain relationship for transversely isotropic material (Razaqpur, 1991) is σ11 σ22 σ12 = A11 A12 0 A21 A22 0 0 0 2A66 �11 �22 �12 (13) E(x) = A11(x) − A211(x) A22(x) (14) where E(x) is moduli of elasticity. FREE VIBRATION OF COMPOSITE BEAMS 521 3. Finite element formulation Euler-Bernoulli beam theory was used. Cubic displacement functions are selected. Five elements used during the calculations. Element rigidity matrix and element mass matrix respectively are as follows: [K]e = E(x)(I/l3) 12 6l −12 6l 4l2 −6l 2l2 12 −6l sym. 4l2 (15) [M]e = ρAl 13/35 11l/210 9/70 −13l/420 l2/105 13l/420 −l2/140 13/35 −11l/210 sym. l2/105 (16) Well-known equation for free vibration is ([K] − w2[M])q = 0 (17) where w represents natural frequency. TABLE I. Classical Euler-Bernoulli beam theory m (mode) w 1 0.395 2 1.582 TABLE II. Simply supported beam with small curvatures (mode 1) w � E2/E1 0 0.01 0.02 0.03 0.04 0.05 0.06 1 0.403 0.403 0.403 0.403 0.403 0.403 0.403 10 0.945 0.943 0.939 0.931 0.921 0.908 0.893 20 1.306 1.303 1.295 1.282 1.263 1.240 1.212 50 2.035 2.030 2.016 1.990 1.953 1.902 1.841 100 2.864 2.857 2.834 2.792 2.725 2.632 2.520 522 YAVUZ, ERGÜVEN TABLE III. Simply supported beam with small curvatures (mode 2) w � E2/E1 0 0.01 0.02 0.03 0.04 0.05 0.06 1 1.614 1.614 1.614 1.614 1.614 1.614 1.614 10 3.786 3.780 3.762 3.732 3.691 3.640 3.582 20 5.231 5.221 5.190 5.137 5.064 4.971 4.861 50 8.152 8.134 8.077 7.976 7.827 7.629 7.388 100 11.470 11.440 11.360 11.190 10.920 10.560 10.120 TABLE IV. Simply supported beam with large curvatures (mode 1) w � E2/E1 0 0.1 0.2 0.3 0.4 0.5 0.6 1 0.403 0.403 0.403 0.403 0.403 0.403 0.403 10 0.945 0.824 0.697 0.644 0.619 0.605 0.596 20 1.306 1.080 0.848 0.755 0.709 0.681 0.663 50 2.035 1.560 1.140 0.974 0.887 0.832 0.795 100 2.864 2.051 1.452 1.216 1.088 1.005 0.947 4. Numerical results Natural frequencies for modes I and II of classical Euler-Bernoulli beam are presented in Table 1. It is given that natural frequencies for modes I and II of simply supported periodically curved layered composite beam with small and large curvatures, respectively, in Table 2 to 5. 5. Conclusions In this study, natural frequencies of the periodically curved layered unidirectional fibrous composite beams are obtained. Periodically curved structure of the composite material cause the loss of rigidity, therefore frequencies become smaller. In the case that � = 0 and E2/E1 = 1, the results converge to natural frequencies of classical Euler-Bernoulli beam. FREE VIBRATION OF COMPOSITE BEAMS 523 TABLE V. Simply supported beam with large curvatures (mode 2) w � E2/E1 0 0.1 0.2 0.3 0.4 0.5 0.6 1 1.614 1.614 1.614 1.614 1.614 1.614 1.614 10 3.786 3.308 2.801 2.590 2.490 2.434 2.397 20 5.231 4.338 3.417 3.042 2.855 2.744 2.670 50 8.152 6.275 4.602 3.935 3.583 3.363 3.211 100 11.470 8.259 5.870 4.922 4.403 4.069 3.833 It is shown that Akbarov and Guz Continuum Theory works well even with five elements via FEM. References Akbarov S. D., Guz A. N. (1991) On the continual theory in mechanics of composite materials with curved layers, Prikl. Mech. (in Russian), 20 3-9. Akbarov S. D. (1995) it On the determination of normalized nonlinear mechanical properties of composite materials with periodically curved layers, Int. J. Solids and Structures 32 3129-3143. Kocer M. (1997) Free vibration analysis and buckling loads of cracked beams resting on an elastic foundation, (in Turkish) Ph.D. Thesis, Istanbul Technical University, Istanbul, Turkey. Kutug Z. (1997) Free vibrations and stability of composite plates with curved layers (in Turkish), Ph. D. Thesis, Yildiz Technical University, Istanbul, Turkey. Razaqpur A. G. (1991) Method of analysis for advanced composite structures, Specialty Conference on Advanced Composites Materials in Civil Engineering Structures, Las Vegas, Nevada, USA Yokoyama T. (1988) Parametric instability of Timoshenko beams resting on an elastic foundation, Computers & Structures 28 207-216. 525 525–530. VIBRATIONS OF BEAM CONSTRUCTIONS SUBMERGED INTO A LIQUID Vladimir V. Yeliseyev and Tatiana V. Zinovieva St. Petersburg State Polytechnic University, Russia Abstract. Coupled hydroelasticity problems for beam systems in compressible viscosity liquid are considered in connection with designing of offshore drilling rigs. The use of singular perturbation method to take effects of stretch, shear, initial stress and damping into account is discussed. The main term of asymptotics is found by the solvability requirements for the small correction terms on the following steps. Non-stationary coupled hydroelasticity problem of cantilever beam resonance oscillations in a liquid is considered. The resistance force is determined by solving the two-dimensional problem of hydromechanics for a cylinder in boundless liquid medium. The Navier-Stokes equations are corrected for a case of compressibility. Key words: rods, perturbation method, viscosity liquid, coupled problem, resonance oscillations 1. The equations of linear theory of rods The one-dimensional model of Cosserat rod represents the material line consisting of elementary solid bodies. An angular position s at reference configuration is attached to each particle. The position of each particle is given by the position vec- tor r. The particles displacements are defined by the vector of small displacement u(s, t) and the vector of small rotational displacement ϑ(s, t). The element of material line ds has a mass ρ(s)ds, eccentricity ε(s) (sets the center of mass position to a pole) and tensor of inertia Ids. External force qds and moment mds act on it. Internal interactions are defined by force Q(s, t) and moment M (s, t). The linear balance equations of forces and moments are given by: Q′ + q = ρ ( ü + ϑ̈ × ε ) , M ′ + r′ ×Q +m = ρε × ü + I · ϑ̈, (1) see (Yeliseyev, 2003), elastic stress-strain relations are as follows M = a ·ϑ′ + c · γ, γ ≡ u′ −ϑ × r′, Q = b · γ +ϑ′ · c. (2) All coefficients here are functions of angular position, they are prescribed by initial state. Rigidity of the rod is characterized by the second-rank tensors: a — Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 526 YELISEYEV, ZINOVIEVA tensor of bending and torsion rigidity, b — tensor of tension and shear rigidity, c — tensor of cross relations. The reciprocity theorem of works is valid in a linear statics of rods l∫ 0 (q1 ·u2 +m1 ·ϑ2)ds + (Q1 ·u2 +M 1 ·ϑ2) ∣∣∣l 0 ≡ A12 = A21, (3) where two different states of a rod are denotes by subscripts 1 and 2, q and m are external distributed force and moment, Q and M — loads on the ends of rod. 2. Perturbation method The asymptotic methods are very effective at presence of many factors influencing the solution. Usually formulation of the problem is reduced to dimensionless form for introduction a small parameter λ → 0. Then a small dimensionless combina- tion is denoted as λ. But another way exists: if it is known, that a some parameter p changes the solution slightly, then it is possible to rename it λp and to begin the asymptotic analysis at λ → 0. Let us consider the problem of natural frequencies and natural modes defin- ition of the linear elastic discrete system with the small perturbations of masses and rigidities: ( C − ω2A ) U = 0, A = A0 + λA1, C = C0 + λC1, λ → 0, (4) here A — matrix of masses, C — matrix of rigidities. Let us search the solution in the form ω = ω0 + λω1 + . . . , U = U0 + λU1 + . . . (5) We receive a sequence of problems, substituting (5) in (4) and equating the coef- ficients at identical powers of λ: ( C0 − ω20A0 ) U0 = 0, ( C0 − ω20A0 ) U1 = ( −C1 + ω20A1 + 2ω0ω1A0 ) U0, . . . (6) On the first of these steps we find the natural frequencies ω0 and the modes U0 of unperturbed system. On the second step we have non-uniform system with a matrix whose determinant equals zero; it is solvable only in a case of orthogonality of its right member to the solutions of conjugate uniform system. Since C0 and A0 are symmetric, conjugate system coincides with initial. Therefore the requirement of solvability takes the following form UT0 ( −C1 + ω20A1 + 2ω0ω1A0 ) U0 = 0 ⇒ ω1 = UT0 ( C1 − ω20A1 ) U0 2ω0UT0 A0U0 . BEAM CONSTRUCTIONS SUBMERGED INTO A LIQUID 527 Similarly, on the following steps we obtain the allowances ω2, ω3 etc. The requirement of solvability is expressed by the reciprocity theorem of works (Yeliseyev, 2003) in the rod systems. This approach is used in (Yeliseyev and Zinovieva, 2004) for taking the effects of tension, shear and initial perturba- tions in rod models into account. Nextly it is used to solve the problem of beam resonance oscillations in a liquid. 3. Resonance oscillations of an elastic beam in a compressible viscosity liquid The calculations of beam systems oscillations in a liquid are necessary for an estimation of rigidity of offshore drilling rigs constructions. The hydrodynamic forces change inertial and damping properties of system. They can cause cross oscillations of body when viscosity liquid flows about it. 3.1. LINEAR VIBRATIONS OF A LIQUID Usually it is assumed that a liquid is ideal and incompressible when the motion and the deformation of a body in the liquid are considered. However a preanalysis has shown the necessity of taking viscosity and compressibility into account. The stress tensor in a compressible viscosity liquid is given by τ = −pE + 2η ( ∇vS − 1 3 ∇ · vE ) , (7) where p — pressure, v — velocity vector, E — unit tensor, η — viscosity. In a compressible liquid at constant temperature p = p (ρ) — density function. The balance equations of forces and mass should be satisfied: ρ (v̇ + v · ∇v) = ∇ · τ = −∇p + η ( ∆v + 1 3 ∇∇ · v ) , (8) ρ̇ + ∇ · (ρv) = 0 (9) ((. . .)· — local derivative with respect to time t). Further the perturbations of pressure p̃ = p − p0, density ρ̃ = ρ − ρ0 and velocity v are considered as small quantities. The linearized set of equations is as follows ∇ · τ̃ = ρ0v̇, (10) ˙̃ρ + ρ0∇ · v = 0, (11) p̃ = c2ρ̃, c2 ≡ dp dρ ∣∣∣∣∣ 0 . (12) 528 YELISEYEV, ZINOVIEVA Let us introduce a displacement field u (r, t) (r — position vector of loca- tion): du/dt ≡ u̇ + v · ∇u = v ⇒ v = u̇ (13) after a linearization. Then from (11) at zero initial conditions ρ̃ = −ρ0∇ · u. Varying (7) and taking account of (12), we obtain τ̃ = ( c2ρ0 − 2 3 η∂t ) ∇ ·uE + 2η∂t∇uS . (14) It looks as a relation of the classical theory of elasticity τ = λ∇ ·uE + 2µ∇uS , only Lame coefficients became the operators. At harmonic oscillations u (r, t) = u (r) eiωt etc. We obtain for complex amplitudes (λ + µ)∇∇ ·u + µ∆u + ω2ρ0u = 0, λ = c2ρ0 − 2 3 i ωη, µ = iωη. Usually the scalar and the vector potentials are introduced to solve this equation (Slepyan, 1972): u = ∇ϕ + ∇ ×ψ, ∇ ·ψ = 0; c21∆ϕ + ω 2ϕ = 0, c22∆ψ + ω 2ψ = 0, c21 = (λ + 2µ)/ρ0, c 2 2 = µ/ρ0. (15) Let us search a liquid reaction on the solid cylinder of radius R, which oscil- lates in a plane under the law û (t) = weiωti. Using relations (15), we obtain the expression of force per unit of cylinder length in Hankel functions: F = R 2π∫ 0 er · τ (R, θ) dθ = −Fi, F = πR2ρ0wω 2 β1α2 − 2β1β2 + β2α1 α1α2 − β1β2 , (16) αk ≡ Ḣ(2)1 (γk) γk, γk ≡ ωR ck , βk ≡ H(2)1 (γk) , k = 1, 2. The calculations by formula (16) shows that the liquid gives an additional inertia (associated mass) and a damping. BEAM CONSTRUCTIONS SUBMERGED INTO A LIQUID 529 3.2. RESONANCE OSCILLATIONS OF BEAM The resonance oscillations are very dangerous conditions, the dynamic rigidity of a construction becomes minimal. The resistance forces restrict a vibration ampli- tude on a resonance. Let us find a resonance amplitude and phase of bending oscillations of can- tilever beam simulated by the Bernoulli-Euler’s beam. Let us presume small the inducing and damping forces and use the asymptotic method. The frequency ω of an external force coincides with one of natural frequencies of a “dry” beam. The statement of problem for the beam deflection u (x, t) is as follows auIV + λ f (u̇) + ρ̂ü = λq (x) sinωt, u (0) = u′ (0) = u′′ (l) = u′′′ (l) = 0. (17) Here λ → 0 — formal small parameter, a — bending rigidity, ρ̂ — beam mass per unit of length. The natural modes and frequencies of “dry” beam are found from the equation aUIV − ω2ρ̂U = 0 (18) with boundary conditions (17). The modes are orthonormal: l∫ 0 ρ̂UnUkdx = δnk, δnk = { 1, n = k 0, n � k (19) The periodic solution of problem (17) we search as Poincare expansion (Nayfe, 1976) u (x, t) = u0 (x, t) + λu1 (x, t) + . . . We have for a main term auIV0 + ρ̂ü0 = 0, u0 = ΘU sin (ωt − α) , (20) as the frequency ω = ωi — coincides with one of natural one. Amplitude factor Θ and phase shift α are found from an equation for u1 auIV1 + ρ̂ü1 = q (x) sinωt − f (u̇0) ≡ β (x, t) (21) with boundary conditions (17). The solution is searched as expansion in terms of natural modes u1 = ∑ Θk (t) Uk (x). 530 YELISEYEV, ZINOVIEVA Considering (18) and (21) as two problems of statics and applying the theorem (Yeliseyev, 2003), we can determine the principal coordinates Θk l∫ 0 ω2nρ̂Unu1dx = l∫ 0 (β − ρ̂ü1) Undx. (22) Whence, allowing for a normality condition (19), it is follows Θ̈k + ω 2 kΘk = l∫ 0 βUkdx. (23) The right side of (23) has a period T = 2π/ω. The absence of the first har- monics in the right side of the equation for a resonating mode U (corresponding ωi = ω) is necessary for periodicity of u1. So for Θ and α we obtain a set of equations l∫ 0 qUdx − 2 T T∫ 0 sinωt l∫ 0 f (u̇0)Udxdt = 0, T∫ 0 cosωt l∫ 0 f (u̇0)Udxdt = 0. (24) The used asymptotic method reminds a principle of balance of works. References Nayfe A. (1976) Perturbation methods, M. Mir. Slepyan L. I. (1972) Non-steady elastic waves, L. Sudostroenie. Yeliseyev V. V. (2003) Mechanics of elastic bodies, St.-Petersburg State Polytech. Univ. Yeliseyev V. V., Zinovieva T. V. (2004) Perturbation method in problems of oscillations of beam systems, Mechanics of materials and strength of constructions, SPbSPU papers, St. Petersburg, 489 200-209. 531 531–536. VIBRATION BEHAVIOR OF COMPOSITE BEAMS WITH RECTANGULAR SECTIONS CONSIDERING THE DIFFERENT SHEAR CORRECTION FACTORS Vebil Yıldırım Department of Mechanical Engineering, University of Çukurova, 01330 Adana, Turkey Abstract. As is well known, there are the first and higher order shear deformation theories that involve the shear correction factor (k- factor), which appears as a coefficient in the expression for the transverse shear stress resultant, to consider the shear deformation effects with a good approx- imation as a result of non-uniform distribution of the shear stresses over the cross-section of the beam. Timoshenko’s beam theory (TBT) accounts both the shear and rotatory inertia effects based upon the first order shear deformation theory which offers the simple and acceptable solutions. The numerical value of the k- factor which was originally proposed by Timoshenko depends upon gener- ally both the Poisson’s ratio of the material and the shape of the cross-section. Recently, especially the numerical value of the k-factor for rectangular sections is examined by both theoretical and experimental manners. Although there are no large numerical differences among the most of the theories, a few of them says that the k-factor varies obviously with the aspect ratio of rectangular sections while Timoshenko’s k-factor is applicable for small aspect ratios. In this study, the effect of the different k-factors developed by Timoshenko, Cowper and Hutchinson on the in-plane free vibration of the orthotropic beams with different boundary conditions and different aspect ratios are studied numerically based on the transfer matrix method. For the first six frequencies, the relative differences of among the theories are presented by charts. Key words: shear correction factor, Timoshenko’s theory, vibration, rectangular section, first-order 1. Introduction The forth-order differential equation for the flexural motion of bars in y − z plane based on the TBT by two coefficients is given by (Craig, 1981) (out-of-plane bending) (EIx) ∂4v ∂z4 + (ρA) ∂2v ∂t2 − ρIx(1 + E k1G ) ∂4v ∂z2∂t2 + ρ2Ix( 1 k2G ) ∂4v ∂t4 = 0 (1) where k1 and k2 are the shear correction factors (k−factors), t is the time, v is the transverse displacement, ρ is the density, A is the cross-sectional area, Ix is Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 532 YILDIRIM the moment of inertia, E and G are Young’s and shear modulus, respectively. The roots of the frequency equation, which are obtained from the solution of equation (1) with the boundary conditions, give the natural frequencies. As seen in eqn. (1) when k → ∞ the effect of the shear deformation is negligible, just the rotational inertia effect remains. If the rotatory inertia is also neglected (h/L → 0), the following well-known Bernoulli-Euler equation is obtained. EI ∂4v ∂z4 + (ρA) ∂2v ∂t2 = 0 (2) In most cases, the effects of the shearing deformations for isotropic materials, k = k1 = k2 = 5/6 = 0.8333 and υ = 0.3 are observed as remarkable significant than the rotatory inertia effects. In general, as the mode number increases the effects of both the rotatory inertia and shearing deformations increase. Those effects decrease with the increasing slenderness ratio. Apart from these, those effects are affected from the boundary conditions. On the contrary to the isotropic beams, the effect of shear deformations becomes very important for even slender composite beams. The E/G ratio varies in the range of 20 and 50 for modern com- posites while that ratio is 2, 5 ÷ 3 for isotropic materials. For sandwich structures those effects take also a great deal significance. The range of the validity of the first-order shear deformation theory is strongly dependent on the shear correction factors used, which exactly depend upon both the geometry of the cross-section and material type of the beam. In this respect, the main problem in solving the problems by the first order shear deformation theory is that the decision to the best k− factor. Timoshenko (1921, 1922) offers the k-factor dependent on just Poisson’s ratio as follows by solving 2-D elasticity bending problem kTimoshenko(υ) = k1 = k2 = 5 + 5υ 6 + 5υ (3) (Cowper, 1966) recommended the following k-factor which depends upon Pois- son’s ratio for the 3-D integral solution of the elasticity problem kCowper(υ) = k1 = k2 = 10 + 10υ 12 + 11υ (4) In both beam and plate analyses, the most used factor is given as k = k1 = k2 = 5/6 ≈ 0.833333 (Reissner, 1947; Timoshenko, 1991; Stephen, 2002, Renton, 1991) which is obtained by parabolic shear distribution assumption. The corre- sponding factor in plate dynamics is taken as k = k1 = k2 = π2/12 ≈ 0.822467 (Mindlin et al., 1956). The other k-factor used in the plate dynamics was offered by (Uflyand, 1948) as k = 2/3 = 0.666667. (Kaneko, 1975) reviewed the devel- opments on k-factor up to that time. (Stephen and Levinson, 1979) proposed the beam theory with two coefficients k1 = kTimoshenko(υ), k2 = kCowper(υ) (5) VIBRATION BEHAVIOR OF COMPOSITE BEAMS WITH . . . 533 The earliest study about the relationship between k-factor and the aspect ratio was performed by (Stephen, 1980). For this relation, (Hutchinson and Zilmer, 1986) offered a serial solution includes the aspect ratio. (Stephen, 1997) outlined the k−factors used in plate dynamics. For isotropic beams, (Pai and Schulz, 1999) suggested a k−factor formulation called energy-consistent without solving the elasticity equations. kPai/Schulz = h4(1 + υ)2 36 ( h4(1+υ)2 30 + υ2b4 180 − υ2b5 2π5h ∞∑ n=1 tanh πhb n5 ) (6) The above yields k = 5/6 for h/b → 0 and/or υ → 0 and verifies the third order theory. (Hutchinson, 2001) formulated the k−factor consists of both υ and (b/h) effects based on the basic dynamic beam theory. kHutchinson(υ, b h ) = k1 = k2 = − 2(1 + υ)[ 9 4h5b C4 + υ(1 − b 2 h2 ) ] C4 = 4 45 h3b(−12h2 − 15h2υ + 5b2υ) + ∞∑ n=1 16υ2b5 [nπh − b tanh(nπh)] n5π5(1 + υ) (7) (Stephen, 2001) showed that, the results in his previous study (Stephen, 1980), which was obtained by using the different method, are the same as (Hutchin- son, 2001). (Renton, 2001) studied the elasto-dynamic Timoshenko’s solution of plane stress problem. (Stephen, 2002) expanded the Renton’s work. For or- thotropic Bickford beam and plates, (Soldatos and Sophocleous, 2001), proposed the k−factor as k = k1 = k2 = 14/17 ≈ 0.823529 based on the parabolic shear distribution. For both two-skin as well as multi-skin sandwich structures, (Birman, 2002) states that the k−factor should be taken as unity. (Puchegger et al., 2003) have performed several experiments with the ”ideal” computer-generated material specimens with free-free ends. They show that Hutchinson’s k−factor is the only one to describe the clear dependence on the aspect ratio correctly. While Hutchinson’s k−factor is used for large aspect ratios (b/h) and Poisson’ ratios, Timoshenko’s k−factor is valid for just small aspect ratios. (Puchegger et al., 2003) also expressed that the effect of the slenderness ratio (L/h) on the formulation of k-factor is negligible except for very short and very long beams. As seen from the above, there are not enough experimental studies to verify the true value of k−factor in the literature except the particular work of (Puchegger et al., 2003). Numerical investigation of the effect of k−factors frequently used on the free vibration of beams will be very useful to the readers. 534 YILDIRIM Figure 1. The variation of the shear coefficient reciprocal wrt the aspect ratio Figure 2. The geometry of the beam and coordinate axes 2. Formulation and examples The in-plane (axial and flexural) linear free vibration equations of the orthotropic beams with rectangular sections are given as follows (Yıldırım et al., 1999a-b) dUz dz = A ′ 11Tz, dUx dz = Ωy + A ′ 22kTx, dΩy dz = D ′ 33My, dTz dz = −ρAω 2Uz, dTx dz = −ρAω 2Ux, dMy dz = −ρIyω 2Ωy − Tx (8) Where ω is the natural frequency, Tz is the axial force, Tx is the shear force, My is the bending moment, Ωy is the rotation, Ux and Uz are displacements. The other terms are computed with the help of the elements of the stiffness and compliance matrices, Ci j and S i j (Yıldırım et al., 1999a-b). Equation (8) corre- sponds to a set of six linear ordinary differential equations. In order to obtain its exact solution, the transfer matrix method together with an effective numerical algorithm developed by the author to obtain the element transfer matrix is used in this work (Yıldırım et al., 1999a-b). Two boundary conditions are considered with two aspect ratios in this study for the slenderness ratio, L/h = 10. Figure 3 VIBRATION BEHAVIOR OF COMPOSITE BEAMS WITH . . . 535 Figure 3. Variation of the six first natural frequencies with the boundary conditions, k−factors and aspect ratios (Dimensionless frequency= ω̄ = ωL2 √ ρ E1h2 ; E1 = 144.8GPa, E2 = 9.65GPa, G12 = G13 = 4.14GPa, G23 = 3.45GPa, ρ = 1389.23kg/m3, ν = 0.3) shows the variation of the first six natural frequencies with the boundary condi- tions, k-factors and aspect ratios. As seen from Figure (3a), for higher frequencies, Cowper’s k−factor gives a little bit higher frequencies than k = 5/6. In general, the same results are obtained with both Timoshenko’s and Hutchinson’s k−factors for higher frequencies of square sections (b/h = 1). For higher aspect ratios, e.g. b/h = 5, it is obviously seen that the relative differences between Hutchin- son’s k−factor and the other k−factors becomes very significant. It may be over 100% for some frequencies, e.g. for the second frequency of fixed-fixed ends and b/h = 5. This ratio may be greater for the other high aspect ratios. 3. Conclusions The effect of the k−factors offered by different researchers on the free vibration of orthotropic beams is investigated numerically. It is observed that the results are very close to each other except Hutchinson’s results. To get rid of the expected confusion in choosing the k-factors to be used in the analysis, some experiments related the k−factors should be performed. 536 YILDIRIM References Birman, V. (2002) On the choice of shear correction factor in sandwich structures, Journal of Sandwich Structures & Materials 4, 83-95. Cowper, G. R. (1966) The shear coefficient in Timeshenko’s beam theory, ASME JAM 33, 335-340. Craig, R. R. (1981) Structural Dynamics, John Wiley & Sons. Hutchinson, J. R., Zilmer, S. D. (1986) On the transverse vibration of beams of rectangular cross- section, ASME JAM 53, 39-44. Hutchinson, J. R. (2001) Shear coefficient for Timeshenko beam theory, ASME JAM 68, 87-92. Kaneko, T. (1975) On Timoshenko’s correction for shear in vibrating beams, Journal of Physics D 8, 1927-1936. Mindlin, R. D., Schacknow, A., Deresiewicz, H. (1956) Flexural vibrations of rectangular plates, ASME JAM 23, 430-436. Pai, P. F., Schulz, M. J. (1999) Shear correction factors and an energy-consistent beam theory, Int. J. Sol. & Str. 36, 1523-1540. Puchegger, S., Bauer, S., Loidl, D., Kromp, K., Peterlik, H. (2003) Experimental validation of the shear correction factor, JSV 261, 177-184. Reissner, E. (1947) On bending of elastic plates, Quarterly of Applied Mathematics 5, 55-68. Renton, J. D. (1991) Generalized beam theory applied to shear stiffness, Int. J. Solid. Struct. 27, 1955-1967. Renton, J. D. (2001) A check on the accuracy of Timoshenko’s beam theory, JSV 245, 559-561. Soldatos, K. P., Sophocleous, C. (2001) On shear deformable beam theories: The frequency and normal mode equations of the homogeneous orthotropic Bickford beam, JSV 242, 215-245. Stephen, N. G., Levinson, M. (1979) A second order beam theory, JSV 67, 293-305. Stephen, N. G. (1980) Timoshenko’s shear coefficient from a beam subjected to gravity loading, ASME JAM 47, 121-127. Stephen, N. G. (1997) Mindlin plate theory: Best shear coefficient and higher spectra validity, JSV 202, 539-553. Stephen, N. G. (2001) On “Shear coefficient for Timoshenko beam theory”, ASME, JAM 68, 959- 960. Stephen, N. G. (2002) On “A check on the accuracy of Timoshenko’s beam theory”, JSV 257, 809-812. Timoshenko, S. P. (1921) On the correction for shear of the differential equation for transverse vibrations of prismatic bars, Philosophical Magazine 41, 744-746. Timoshenko, S. P. (1922) On the transverse vibrations of bars of uniform cross-section, Philosoph- ical Magazine 43, 125-131. Timoshenko, S. P. (1991) Mechanics of Materials, Chapman & Hall. Uflyand, Y. S. (1948) The propagation of waves in the transverse vibrations of bars and plates, Prikladnaia Mathematika Mekhanica 12, 287-300. Yıldırım, V., Sancaktar, E., Kiral, E. (1999a) Free vibration analysis of symmetric cross-ply lami- nated composite beams with the help of the transfer matrix approach, Com. Num. Meth. Eng. 15, 651-660. Yıldırım, V., Sancaktar, E., Kiral, E. (1999b) Comparison of the in-plane natural frequencies of symmetric cross-ply laminated beams based on the Bernoulli-Euler and Timoshenko beam theories, ASME, JAM 66, 410-417. 537 537–542. AIR BLAST-INDUCED VIBRATION OF A LAMINATED SPHERICAL SHELL Hüseyin Murat Yüksel and Halit S. Türkmen Faculty of Aeronautics and Astronautics, İstanbul Technical University, Maslak, İstanbul 34469, Turkey Abstract. The scope of this study is to investigate the dynamic behavior of a laminated spherical shell subjected to air blast load. The shell structure considered here is a hemisphere in shape and made of a glass/epoxy laminated composite material. The blast experiments are performed on the spherical shell. The strain-time history of the center of the spherical shell panel is obtained experi- mentally. The blast loaded spherical shell is also modeled and analyzed using ANSYS finite element software. The static analysis is performed to characterize the material. The dynamic response of the spherical shell panel obtained numerically is compared to the experimental results. It is observed that the response frequency corresponds to the higher vibration modes of the panel. The qualitative agreement is found between the numerical and experimental results. Key words: air blast, vibration, spherical shell, laminated composite 1. Introduction The dynamic response of laminated composite structures to the blast load is an extremely important subject in engineering. The suddenly applied loading such as a blast load is known as a critical loading condition. They may cause the fiber brekage, delamination and finally failure of the structure. Theoretical studies on this area give many useful information about the response of laminated structures to the blast load. However, it is necessary to support all these studies with exper- iments. It is difficult to perform real blast tests on the structures. In this study it is aimed to make a simple experimental setup and perform the blast tests. There are many studies reported on the response of laminated structures to the blast load. Many of them are related to the laminated plates ((Librescu,1990), (Wier- nicki, 1990), (Turkmen, 1999), (Turkmen, 1999)). Some of them are related to the laminated cylindrical shell panels subjected to the blast load ((Mukhopadhyay, 1996), (Zhu, 1997), (Turkmen, 2002), (Stein, 1986)). There are very few studies reported on the spherical shell panels subjected to the blast load. The spherical, or in general doubly curved, shell panels have been found many application areas particularly in the aerospace industry. The radom, external storage tank, leading Vibration Problems ICOVP 2005, c© 2007 Springer. (eds.), . şE. Inan and A. Kırı 538 YÜKSEL AND TÜRKMEN edge of a missile, wheel pant can be given as examples for the use of doubly curved shell panels. These panels usually are made of laminated composites to save weight. Therefore, a spherical shell panel made of laminated composites is investigated under the effect of the blast load, experimentally and theoretically. The strain at the center of the shell is measured during the effect of the blast load. The displacement, strain and stress are calculated using the finite element method. The vibration frequencies are obtained from both experimental and numerical data. 2. Experimental procedure The static and blast tests are carried out to investigate the dynamic behavior of the laminated spherical shell. The shell structure considered in this study has a hemispherical shape with a radius of 101.5 mm. The number of layers are 3 and each layer is made of a glass/epoxy laminated composite material with 0.16 mm thickness. The material properties are given in Table 1. TABLE I. Mechanical properties of the glass/epoxy laminate Ex = Ey(GPa) Ez(GPa) Gxy(GPa) Gyz = Gxz(GPa) νxy νyz = νxz ρ(kg/m3) 24.14 3.0 3.79 1.076 0.11 0.4 1800 The experimental setup is composed of a pressure tube, a main frame, a com- pressor, a digital scope, a dynamic strain meter, a multimeter, and a computer (Yuksel, 2005). A schematic of the experimental setup is shown in Figure 1. The static tests are carried out to characterize the material. The metal objects with known weights are put on the top face at the center of the clamped panel and the strain on the bottom face at the center of the panel is measured using a dynamic strainmeter and a multimeter. The static test results is shown in Table 2. The measured strains are proportional to the loads linearly. This indicates that the material response stays in the elastic regime for the loads applied in the tests. The air blast load is obtained as a result of a sudden charge from a compressor. The procedure to obtain this sudden charge is as follows: 1) The air is fed into a steel tube from one end and the other end of the tube is closed using a thin membrane. 2) A manometer mounted on the tube is used to read the pressure value inside the tube. 3) The membrane is torn when the air pressure reaches a critical value for the membrane and the sudden discharge occurs. The strain- time history on the bottom face at the center of the panel are measured using strain gauges and a dynamic strain meter. The signals obtained from the strain gauges and the dynamic strainmeter are digitized using a digital scope and transferred AIR BLAST-INDUCED VIBRATION 539 Figure 1. The schematic of the experimental setup TABLE II. Static test and analysis results (microstrain) Weight(gr) Test Analysis 56.69 73.0 77.8 134.08 131.5 184.0 190.77 184.5 262.0 236.39 229.0 325.0 366.03 374.5 503.0 to a personal computer using an RS232C serial interface. The tests are repeated to decrease the experimental uncertainties. The strain-time history obtained by four different tests with a time duration of 20 ms are shown in Figure 2. In these tests, the pressure read on manometer is 1.7, 1.6, 1.7, 1.8 bar, respectively. The observations in these tests are as follows: 1) The first peak is negative in each test. 2) Damping becomes effective after 2 ms. 3) The response frequency is about 4500 Hz (18 cycles per 4 ms). 4) The highest peak value is negative in each test. Although the tube pressure is almost same in each test, there are also some differences observed in the strain-time history. The highest peak strain values are -1269, -2187, -1542, -1542, respectively. 3. Finite element model of the blast loaded shell structure On the numerical side of the study, the static and blast tests are modeled using a finite element software (ANSYS). The mesh structure used to model the static tests is shown in Figure 3a. In this model, the mesh is dense at the center which the load is applied and the stress is concentrated (Figure 3b). Because of the concentrated load at this point it is needed finer mesh. The strain is calculated on the bottom face at the center of the panel for each weight. The results of the 540 YÜKSEL AND TÜRKMEN 1200 800 400 0 -400 -800 -1200 -1600 0 0.5 1 1.5 2 2.5 3 3.5 4 M ic ro st ra in Time (ms) Test 1 1200 800 400 0 -400 -800 -1200 -1600 -2000 -2400 0 0.5 1 1.5 2 2.5 3 3.5 4 M ic ro st ra in Time (ms) Test 2 1200 800 400 0 -400 -800 -1200 -1600 0 0.5 1 1.5 2 2.5 3 3.5 4 M ic ro st ra in Time (ms) Test 3 1200 800 400 0 -400 -800 -1200 -1600 0 0.5 1 1.5 2 2.5 3 3.5 4 M ic ro st ra in Time (ms) Test 4 Figure 2. The experimental strain-time history on the bottom face at the center of the panel Figure 3. a) The finite element model used in the static analysis b) The detail of the refined mesh at the center of the sphere c) The finite element model of the hemispherical shell panel used in the transient analysis static analyses are shown in Table 2. The model used in the static analysis is not used in the dynamic analysis, because of the distributed nature of the blast load. The laminated spherical shell subjected to air blast load is modeled using the finite element method. For this purpose a finite element software (ANSYS) is used. The modal analysis is achieved to obtain the free vibration frequencies. The free vibration frequencies of the first twenty modes ranged from 2373 Hz to 4216 Hz. The effect of the mesh density on the free vibration frequencies is investigated to decide the number of elements of the final model. The increase in the mesh density slightly effected the free vibration frequencies. This effect is most pronounced for the higher vibration modes. It is decided that using a moderate number of elements is enough for the accuracy during the transient AIR BLAST-INDUCED VIBRATION 541 -60000 -40000 -20000 0 20000 40000 60000 0 5 10 15 20 25 30 35 40 P re ss ur e (N /m 2 ) Time (ms) Pm=57812 N/m 2(a) -25 -20 -15 -10 -5 0 5 10 15 0 0.5 1 1.5 2 2.5 3 3.5 4 S tr es s (M P a) Time (ms) (c) -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0 0.5 1 1.5 2 2.5 3 3.5 4 D is pl ac em en t ( m m ) Time (ms) (b) -800 -600 -400 -200 0 200 400 0 0.5 1 1.5 2 2.5 3 3.5 4 M ic ro st ra in Time (ms) (d) Figure 4. a) The variation of air blast pressure b) The displacement-time history during the blast loading c) The stress-time history during the blast loading d) The strain-time history during the blast loading analysis. The model which has 1200 elements is chosen for the transient analysis (Figure 3c). The transient analysis is achieved using the final model to obtain the strain-time histories. The air blast load is represented by using a Friedlander decay function shown in Figure 4a. This load is applied to the circular area around the center of the shell panel. The displacements in the perpendicular direction to the shell panel, stresses, and strains on the bottom face at the center are obtained numerically using the finite element method. These are shown in Figure 4b,c,d, respectively. The results of the analysis are listed below: 1) The peak displacement is -0.18 mm. This is in the linear elastic range. 2) The vibration follows the blast load. The vibration amplitude takes its minimum value when the load is zero. 3) The vibration frequency is about 4500 Hz. 4) The peak stress is -21.5 Mpa. 5) The peak strain is -794.8 microstrain. 4. Comparisons of the experimental and numerical results The strains and vibration frequencies obtained experimentally and numerically are compared. A qualitative agreement is found between the results. The response frequency is predicted very well using the finite element method. The peak strain obtained numerically differs from the experimental ones. In this study the air pres- sure on the panel is not measured. The air pressure is predicted from the pressure read on the manometer and the decrease in the pressure with the distance from the open end of the tube (Turkmen, 1999). Therefore, the predicted air blast pressure 542 YÜKSEL AND TÜRKMEN may differ from the exact pressure value on the panel. This mainly effects the peak strains. 5. Conclusion The dynamic behavior of a hemispherical laminated composite panel is investi- gated under the air blast load. The experiments indicate the higher vibration modes are stimulated under the air blast load. The response frequency obtained numer- ically is in an agreement with the experimentally obtained response frequencies. There is a qualitative agreement between the numerical and experimental results. However the amplitudes are not predicted correctly because the air blast pressure is not measured. The numerical errors, experimental errors are the other reasons for the discrepancies. This study presents an experimental method to perform the blast experiments and to investigate the dynamic behavior of panels. The panels made of different materials can be tested using this method. So the structural optimization can be achieved for the blast loaded panels. The experimental setup can be modified to investigate the impact of a subject on the panels. These will be the subject of future studies. References Librescu L., Nosier A. (1990) Response of Laminated Composite Flat Panels to Sonic Boom and Explosive Blast Loadings, AIAA Journal 28 345-352. Mukhopadhyay M., Goswami S. (1996) Transient finite element dynamic response of laminated composite stiffened shell, Aeronautical Journal 100 223-233. Stein M. (1986) Nonlinear Theory for Plates and Shells Including the Effects of Transverse Shearing, AIAA Journal 24 1537-1544. Turkmen H. S., Mecitoglu Z. (1999) Dynamic response of a stiffened laminated composite plate subjected to blast load, Journal of Sound and Vibration 3 371-389. Turkmen H. S., Mecitoglu Z. (1999) Nonlinear structural response of laminated composite plates subjected to blast loading, AIAA Journal 12 1639-1647. Turkmen H. S. (2002) Structural response of laminated composite shells subjected to blast loading: Comparison of experimental and theoretical methods, Journal of Sound and Vibration 249 663- 678. Wiernicki C. J., Liem, F., Woods, G. D., Furio, A. J. (1990) Structural Analysis Methods for Light- weight Metallic Corrugated Core Sandwich Panels Subjected to Blast Loads, Naval Engineers Journal 28 192-203. Yuksel H. M. (2005) Dynamic Behavior of a Laminated Hemispherical Shell Under the Blast Loading, M. Sc. Thesis, Istanbul Technical University, Istanbul, Turkey. Zhu W. H., Xue H. L., Zhou G. Q., Schleyer G. K. (1997) Dynamic response of cylindrical explosive chambers to internal blast loading produced by a concentrated charge, International Journal of Impact Engineering 19 831-845. AUTHOR INDEX (*: not personally presented) Adanur S. RC 1 Bikçe M.∗ RC 15 RC 9 Birlik G. RC 85 Akbarov S. D. RC 225 Biswas P. RC 265 RC 469 RC 271 RC 511 Bonadies M.∗ RC 243 Akkaş N. ST 7 Bonifasi-Lista C.∗ RC 91 Akköse M. RC 9 Bontcheva N.∗ RC 243 Aksoğan O. RC 15 Boyacı H. RC 97 RC 21 Bulut G. RC 505 RC 371 Akyüz U. RC 27 Celep Z.∗ RC 231 Aldemir Ü.∗ RC 33 Chakrabarti B. K. GL 103 RC 39 Chatterjee A.∗ 103 Alışverişçi F. RC 389 Cherkaev A. GL 111 Apaydın N.∗ 181 Cherkaev E.∗ RC 91 Askes H.∗ RC 317 111 Aslan Ö. RC 377 Choo B. S.∗ RC 15 Aslanyan A. G.∗ RC 45 Aşçı N. RC 51 Çelebi M. GL 123 Aydın E. RC 33 Çırak İ. RC 137 Aydoğdu M. RC 57 Demir F.∗ RC 137 Baltacı A. RC 63 Demiray H. GL 143 Banerjee M. M. GL 69 Deniz A. RC 437 Banks S. P.∗ RC 79 Doğan V. RC 151 Bayraktar A.∗ RC 1 RC 157 RC 9 Dronka J.∗ RC 45 543 544 AUTHOR INDEX Dumanoğlu A. A.∗ RC 1 Kahali R. RC 265 RC 9 RC 271 Kan C. RC 249 Ecker H. GL 163 Kaplunov S. M.∗ RC 195 RC 487 RC 255 Emiroğlu İ∗ RC 469 Karanji S.∗ RC 265 Emsen E.∗ RC 15 RC 271 Erdem A. Ü. RC 175 Karmakar B. RC 265 Erdik M. GL 181 RC 271 Ergüven M. E.∗ RC 519 Kaya M. O.∗ RC 279 Ertepınar A.∗ RC 27 Kırca M.∗ RC 157 Kırış A. RC 285 Fedorov V. A.∗ RC 195 Korfalı Ö. RC 291 Fesenko T. N.∗ RC 205 Kouzov D. P.∗ RC 213 Filippenko G. V. RC 213 Kumbasar N. RC 297 Fomenkov A. V.∗ RC 195 Foursov V. N.∗ RC 205 Levin V. M.∗ RC 317 Gögüş M. T. RC 219 Makhutov N. A.∗ RC 255 RC 475 Matĕjec R.∗ RC 455 Güler C. RC 225 Mehta A. GL 305 Güler K. RC 231 Michelitsch T. M. RC 317 Güney D. RC 39 Mishuris G. S. RC 45 Mizirin A. V.∗ RC 195 Haq S.∗ 327 Mondal S. RC 323 Movchan A. B. GL 327 Ivanova J. RC 243 Movchan N. V. RC 327 Mukherjee A. RC 337 İnan E. RC 285 RC 383 Nabergoj R.∗ RC 487 İtik M. RC 237 Nath Y. GL 415 Nordmann R. GL 345 AUTHOR INDEX 545 Okrouhlik M. GL 357 Singh S.∗ 415 Onbaşlı Ü. RC 377 Sivc̆ák M.∗ RC 429 Oskouei A. V. RC 371 Skarolek A.∗ RC 429 S̆kliba J. RC 429 Özakça M. RC 219 RC 455 RC 475 Slepyan L.∗ 111 Özdemir Ö. RC 279 Smolskiy S. M.∗ RC 195 Özdemir Z. P. RC 377 ∗ RC 21 Özer A. Ö. RC 383 RC 21 Özer M. RC 389 RC 437 Özeren M. S.∗ RC 395 RC 443 Solonin V. I.∗ RC 255 Parlak B. RC 291 Soyluk K.∗ RC 1 Pastrone F.∗ RC 243 Sönmez Ü. RC 449 Patel B. P.∗ 415 Svoboda R. RC 455 Pes̆ek L. RC 403 Postacıoğlu N. RC 395 Şenocak E. RC 249 Pu ◦ st L. RC 403 RC 463 Ramachandran J. GL 409 Tanrıöver H. RC 463 Reşatoğlu N. RC 15 Taşçı F. RC 469 Taşkın V.∗ RC 57 Salamcı M. U. RC 79 Tayşi N.∗ RC 219 RC 237 RC 475 Samadhiya R.∗ RC 337 Tekeli H. RC 137 Sarıkanat M.∗ RC 63 Teymür M. RC 481 Schnack E.∗ RC 443 Tondl A. RC 487 Selsil Ö. RC 45 Topal U. RC 493 Sezgin Ö. C.∗ RC 85 Turhan Ö. RC 499 Shariy N. V.∗ RC 255 RC 505 Sharma A.∗ 415 Türkmen H. S. RC 537 Shtykov V. V.∗ RC 195 Türkmen M.∗ RC 137 Shukla K. K.∗ 415 Sofiyev A. Sofiyev A. H. 546 AUTHOR INDEX Urey H.∗ RC 249 Yavuz M. RC 519 Uysal H.∗ RC 51 Yeliseyev V. V.∗ RC 525 Uzman Ü. RC 51 Yıldırım V. RC 531 RC 493 Yıldız H.∗ RC 63 Yüksel H. M.∗ RC 537 Ülker F. D.∗ RC 237 Zinovieva T. V.∗ RC 525 Wang J.∗ RC 317 Zora B.∗ RC 395 Yahnioğlu N. RC 511 LIST OF PARTICIPANTS (with mailing addresses) Süleyman ADANUR Karadeniz Technical University Department of Civil Engineering 61080 Trabzon Turkey e-mail:
[email protected] Surkay D. AKBAROV Yıldız Technical University Faculty of Chemistry and Metallurgy Department of Mathematical Engineering Davutpasa Campus No:127 34010 Topkapı Istanbul Turkey e-mail:
[email protected] Nuri AKKAŞ Ankara, Turkey e-mail:
[email protected] Mehmet AKKÖSE Karadeniz Technical University Department of Civil Engineering 61080 Trabzon Turkey e-mail:
[email protected] Orhan AKSOĞAN Çukurova University Department of Civil Engineering 01330 Adana Turkey e-mail:
[email protected] Uğurhan AKYÜZ Middle East Technical University Department of Civil Engineering 06531 Ankara Turkey e-mail:
[email protected] Ünal ALDEMİR Istanbul Technical University Department of Civil Engineering Division of Mechanics 34469 Maslak Istanbul Turkey e-mail:
[email protected] G. Füsun ALIŞVERİŞÇİ Yıldız Technical University Department of Mechanical Engineering Division of Mechanics 80750 Yıldız Istanbul Turkey e-mail:
[email protected] Nurdan APAYDIN General Directory of State Highways Istanbul Turkey 547 Nurcan AŞÇI Karadeniz Technical University Gümüşhane Engineering Faculty Department of Civil Engineering 29000 Gümüşhane Turkey e-mail:
[email protected] ¨TUBITAK . . . . . . 548 LIST OF PARTICIPANTS Özden ASLAN Yıldız Technical University Faculty of Science and Art Department of Physics Davutpasa 34010 Istanbul Turkey e-mail:
[email protected] A. G. ASLANYAN Druzhba St. 5 kv. 355 Khimki Moskow Region 141400 Russia Ersin AYDIN Istanbul Technical University Department of Civil Engineering Division of Mechanics 34469 Maslak Istanbul Turkey e-mail:
[email protected] Metin AYDOĞDU Trakya University Department of Mechanical Engineering 22030 Edirne Turkey e-mail:
[email protected] Aysun BALTACI Ege University Department of Mechanical Engineering Bornova Izmir Turkey e-mail:
[email protected] M. Muralimohan BANERJEE Retired Reader Department of Mathematics A. C. College Jalpaiguri 735101 W.B India e-mail: dgp
[email protected] Stephan P. BANKS The University of Sheffield Department of Automatic Control and System Engineering Sheffield UK Alemdar BAYRAKTAR Karadeniz Technical University Department of Civil Engineering 61080 Trabzon Turkey e-mail:
[email protected] M. BİKÇE University of Mustafa Kemal Department of Civil Engineering 31024 Hatay Turkey Gülin BİRLİK Middle East Technical Univeristy Department of Engineering Sciences Ankara Turkey e-mail:
[email protected] Paritosh BISWAS Executive Secretary Von Karman Society for Advanced Study and Research& Founder Member ICOVP (India) old Police Line Jalpaiguri-735101 Bengal India e-mail: biswas
[email protected] M. BONADIES University of Torino Department of Mathematics v.C. Alberto 10 10123 Torino Italy C. BONIFASI-LISTA University of Utah Department of Mathematics Salt Lake City UT 84112 USA Harm ASKES The University of Sheffield Department of Civil and Structural Engineering Sir Frederick Mappin Building Mappin Street Sheffield S1 3JD UK . . . . LIST OF PARTICIPANTS 549 N. BONTCHEVA Bulgarian Academy of Sciences Institute of Mechanics Acad. G. Bonchev Str. bl. 4 113 Sofia Bulgaria Hakan BOYACI Celal Bayar University Department of Mechanical Engineering 45140 Muradiye Manisa Turkey e-mail:
[email protected] Gökhan BULUT Istanbul Technical University Faculty of Mechanical Engineering 34439 Gümüşsuyu Istanbul Turkey e-mail:
[email protected] Zekai CELEP Istanbul Technical University Faculty of Civil Engineering 34469 Maslak Istanbul Turkey e-mail:
[email protected] Bikas K. CHAKRABARTI Saha Institute of Nuclear Physics Bidhan Nagar Kolkata 700064 India e-mail:
[email protected] A. CHATTERJEE Saha Institute of Nuclear Physics Bidhan Nagar Kolkata 700064 India Andrej CHERKAEV University of Utah Utah Salt Lake City Utah USA e-mail:
[email protected] Elena CHERKAEV University of Utah Utah Salt Lake City Utah USA e-mail:
[email protected] B. S. CHOO 29 Shirley Road Nottingham NG35 DA UK Mehmet ÇELEBİ USGS (MS977) 345 Middlefield Rd. Menlo Park CA. 94025 e-mail:
[email protected] İffet ÇIRAK Suleyman Demirel University Department of Civil Engineering Isparta Turkey e-mail:
[email protected] Fuat DEMİR Süleyman Demiral University Department of Civil Engineering Isparta Turkey e-mail:
[email protected] Hilmi DEMİRAY Işık University Faculty of Art & Sciences Maslak Istanbul Turkey e-mail:
[email protected] Ali DENİZ Karlsruhe University Institute of Solid Mechanics Karlsruhe Germany e-mail:
[email protected] Vedat DOĞAN Istanbul Technical University Department of Aeronautical Engineering Maslak Istanbul Turkey e-mail:
[email protected] J. DRONKA Rzeszow University of Technology Department of Mathematics W. Pola 2. 35-959 Rzeszow Poland . . . . . . . 550 LIST OF PARTICIPANTS A. Aydın DUMANOĞLU Grand National Assembly of Turkey Ankara Turkey Horst ECKER Vienna University of Technology Institute of Mech. and Mechatronics Wiedner Hauptstrasse 8-10/E325/A3 A-1040 Vienna Austria e-mail:
[email protected] İbrahim EMİROĞLU Yıldız Technical University Faculty of Chemistry and Metallurgy Department of Mathematical Engineering Davutpaşa Campus No:127 34210 Topkapı stanbul Turkey E. EMSEN Çukurova University Department of Civil Engineering 01330 Adana Turkey Ali Ünal ERDEM Gazi University Architecture of Engineering Faculty Ankara Turkey e-mail:
[email protected] Mustafa ERDİK Bogazici University Bebek stanbul Turkey e-mail:
[email protected] M. Ertaç ERGÜVEN Istanbul Technical University Faculty of Civil Engineering 34469 Maslak stanbul Turkey e-mail:
[email protected] Aybar ERTEPINAR Middle East Technical University Department of Civil Engineering 06531 Ankara Turkey e-mail:
[email protected] V. A. FEDOROV Moscow Power Engineering Institute (Technical University) Moscow Russia Tatiana N. FESENKO Mechanical Engineering Research Institute of Russian Academy of Sciences Moskow Russia George V. FILIPPENKO Institute of Mechanical Engineering Vasilievsky Ostrov Bolshoy Prospect 61 St. Petersburg 199178 Russia e-mail:
[email protected] A. V. FOMENKOV Moscow Power Engineering Institute (Technical University) Moscow Russia Valeriv N. FOURSOV Mechanical Engineering Research Institute of Russian Academy of Sciences Moskow Russia M. Tolga GÖGÜŞ University of Gaziantep Department of Civil Engineering 27310 Gaziantep Turkey e-mail:
[email protected] I . I . I . . LIST OF PARTICIPANTS 551 Coşkun GÜLER Yıldız Technical University Faculty of Chemistry and Metallurgy Department of Mathematical Engineering Davutpasa Campus No: 127 34210 Topkapi Istanbul Turkey e-mail:
[email protected] Kadir GÜLER Istanbul Technical University Faculty of Civil Engineering 34469 Maslak Istanbul Turkey e-mail:
[email protected] Deniz GÜNEY Istanbul Technical University Department of Engineering 34469 Maslak Istanbul Turkey e-mail:
[email protected] S. HAQ University of Liverpool Department of Mathematical Sciences Liverpool UK Jordanka IVANOVA Bulgarian Academy of Sciences Institute of Mechanics Acad. G. Bonchev Str. bl. 4 113 Sofia Bulgaria e-mail:
[email protected] Esin İNAN Işık University Faculty of Arts and Sciences Department of Mathematics Kumbaba Mevkii Sile 34980 Istanbul Turkey e-mail:
[email protected] Mehmet İTİK Erciyes University Engineering and Architecture Faculty Mechanical Engineering Department Yozgat Turkey e-mail:
[email protected] R. KAHALI Department of Physics P. D. Women’s College Jalpaiguri 735101 Bengal India Cihan KAN Istanbul Technical University Department of Mechanical Engineering Maslak Istanbul Turkey e-mail:
[email protected] S. M. KAPLUNOV Imash ran MGTU by N. Bauman OKB GIDROPRESS e-mail:
[email protected] Bapi KARMAKAR Executive Secretary Von Karman Society for Advanced Study and Research& Founder Member ICOVP(India)Old Police Line Jalpaiguri-735101West Bengal India S. KARANJAI University of North Bengal Department of Mathematics North Bengal India . . . . . . . . 552 LIST OF PARTICIPANTS Metin Orhan KAYA Istanbul Technical University Faculty of Aeronautics and Astronautics 34469 Maslak Istanbul Turkey e-mail:
[email protected] Mesut KIRCA Istanbul Technical University Department of Aeronautical Engineering Maslak Istanbul Turkey e-mail:
[email protected] Istanbul Technical University Faculty of Science and Letters Department of Engineering Science Maslak 34469 Istanbul Turkey e-mail:
[email protected] Önder KORFALI Galatasaray University Faculty of Engineering and Tecnology Ortakoy Istanbul Turkey Danill P. KOUZOV Institute of Mechanical Engineering Vasilievsky Ostrov Bolshoy Prospect 61 St. Petersburg 199178 Russia Nahit KUMBASAR Istanbul Technical University Faculty of Science and Letters 34469 Maslak Istanbul Turkey e-mail:
[email protected] Valery M. LEVIN Istituto Mexicano del Petroleo Eje Central Lazaro Cardenas No.152 Col. San Bartolo Atepehuacan C. P. 07730 D. F. Mexico N. A. MAKHUTOV IMASH RAN MGTU by N. Bauman OKB GIDROPRESS Russian Federation Radek MATĔJEC Technical University of Lberec Hálkova 6 461 17 Liberec Czech Republic Anita MEHTA 2H Cornfield Road 3rd Floor Calcutta 700 019 India e-mail:
[email protected] Thomas M. MICHELITSCH The University of Sheffield Department of Civil and Structural Engineering Sir Frederick Mappin Building Mappin Street Sheffield S1 3JD UK e-mail:
[email protected] G. S. MISHURIS Rzeszow University of Technology Department of Mathematics W. Pola 2. 35-959 Rzeszow Poland A. V. MIZIRIN Moscos Power Engineering Institute (Technical University) Moscow Russia Subrata MONDAL Faculty Member of Mechanical Engineering Department Jalpaiguri Govt. Engg College Jalpaiguri West Bengal India e-mail:
[email protected] . . . . . . Ahmet KIRIŞ . . . LIST OF PARTICIPANTS 553 Alexander B. MOVCHAN University of Liverpool Department of Mathematical Sciences Liverpool England UK e-mail:
[email protected] Natasha V. MOVCHAN University of Liverpool Department of Mathematical Sciences Liverpool England UK e-mail:
[email protected] Abhijit MUKHERJEE Department of Civil Engineering IIT Bombay 400076 India e-mail:
[email protected] Radoslav NABERGOJ University of Trieste Via Valerio Department of Naval Architecture Ocean and Environmental Engineering 10 34127 Trieste Italy Y. NATH Indian Institute of Technology Delhi Applied Mechanics Department Hauz Khas New Delhi 110 016 India e-mail:
[email protected] Ranier NORDMANN Arbeitsgruppe Mechatronik-TU Darmstadt Petersenstr. 30 D-64287 Lichtwiese Germany e-mail:
[email protected] Miloslav OKROUHLIK Institute of Thermomechanics Acad. Sci. Czech Rep. Dolejskova 5 CZ-182 00 Czech Republic e-mail:
[email protected] Ülker ONBAŞLI Yıldız Technical University Faculty of Science and Art Department of Physics Davutpasa 34010 Istanbul Turkey e-mail:
[email protected] A. Vatani OSKOUEI Shahid Rajaee University Buildings and Housing Research Center P. O. Box 13145-1696 Tehran IRAN e-mail:
[email protected] Mustafa ÖZAKÇA University of Gaziantep Department of Civil Engineering 27310 Gaziantep Turkey e-mail:
[email protected] Özge ÖZDEMİR Istanbul Technical University Faculty of Aeronautics and Astronautics 34469 Maslak Istanbul Turkey e-mail:
[email protected] Zeynep Güven ÖZDEMİR Yıldız Technical University Faculty of Science and Art Department of Physics Davutpaşa 34010 Istanbul Turkey e-mail:
[email protected] Mutlu ÖZER 205 Commodore Dr. California USA e-mail:
[email protected] Ahmet Özkan ÖZER Istanbul Technical University Faculty of Science and Letters Maslak 34469 Istanbul Turkey e-mail:
[email protected] . . . . . . 554 LIST OF PARTICIPANTS M. Sinan ÖZEREN Istanbul Technical University Faculty of Mines 34469 Maslak Istanbul Turkey e-mail:
[email protected] Burak PARLAK Galatasaray University Faculty of Engineering and Tecnology Ortakoy Istanbul Turkey e-mail:
[email protected] F. PASTRONE University of Torino Department of Mathematics v. C. Alberto 10 10123 Torino Italy Ludĕk PES̆EK Institute of Thermomechanics Dolej̆skova 5 Prague 8 CZ 182 00 Czech Republic e-mail:
[email protected] Nazmi POSTACIOĞLU Istanbul Technical University Faculty of Mines Main Campus 80630 Maslak Istanbul Turkey Ladislav PU ◦ ST Institute of Thermomechanics Dolej̆skova 5 Prague 8 CZ 182 00 Czech Republic e-mail:
[email protected] J. RAMACHANDRAN Department of Applied Mechanics Indian Institute of Technology Chennai 600 036 Tamil Nadu India e-mail:
[email protected] Rıfat REŞATOĞLU Çukurova University Department of Civil Engineering 01330 Adana Turkey Metin U. SALAMCI Gazi University Engineering and Architecture Faculty Mechanical Engineering Department Maltepe Ankara Turkey e-mail:
[email protected] Ritesh SAMADHIYA Department of Civil Engineering IIT Bombay 400076 India e-mail:
[email protected] Mehmet SARIKANAT Ege University Department of Mechanical Engineering Bornova Izmir Turkey e-mail:
[email protected] ECKART SCHNACK Karlsrube University Institute of Solid Mechanics Karlsrube Germany e-mail:
[email protected] Özgür SELSİL The University of Liverpool Department of Mathematical Sciences L69 7ZL Liverpool UK e-mail:
[email protected] Erol ŞENOCAK Istanbul Technical University Department of Mechanical Engineering 34437 Gümüşsuyu Istanbul Turkey e-mail:
[email protected] . . . . . . . . LIST OF PARTICIPANTS 555 N. V. SHARIY IMASH RAN MGTU by N. Bauman OKB GIDROPRESS Russian Federation A. SHARMA Thapar Institute of Engineering and Technology Patiala 1470004 India V. V. SHTYKOV Moscow Power Engineering Institute (Technical University) Moscow Russia K. K. SHUKLA Motilal Nehru National Institute of Technology Allahabad 211 004 India S. SINGH Applied Mechanics Department I.I.T. Delhi New Delhi 110 016 India Michal SIVC̆ÀK Technical University of Liberec Department of Mechanics and Stres Analysis Halkova 6 CZ-460 17 Liberec 1 Czech Republic Antonin SKAROLEK Technical University of Liberec Department of Mechanics and Stres Analysis Halkova 6 CZ-460 17 Liberec 1 Czech Republic Jan S̆KLIBA Technical University of Liberec Department of Mechanics and Stres Analysis Halkova 6 CZ-460 17 Liberec 1 Czech Republic e-mail:
[email protected] Lenoid SLEPYAN Tel Aviv University Depatment of Solid Mechanics Materials and Structures Ramat Aviv 69978 Israel S. M. SMOLSKIY Moscow Power Engineering Institute (Technical University) Moscow Russia e-mail:
[email protected] Abdullah H. SOFIYEV Süleyman Demirel University Department of Civil Engineering Isparta Turkey e-mail:
[email protected] Ali SOFIYEV Çukurova University Department of Civil Engineering 01330 Adana Turkey Önder Cem SEZGİN Middle East Technical University Healt and Counselling Center Ankara Turkey e-mail:
[email protected] V. I. SOLONIN IMASH RAN MGTU by N. Bauman OKB GIDROPRESS Russian Federation 556 LIST OF PARTICIPANTS Ümit SÖNMEZ Istanbul Technical University Department of Mechanical Engineering 80191 Gümüşsuyu Istanbul Turkey e-mail:
[email protected] Rudolf SVOBODA Techlab Ltd Sokolovsk 207 190 00 Praha 9 Czech Republic e-mail:
[email protected] Hakan TANRIÖVER Istanbul Technical University Department of Mechanical Engineering Gümüşsuyu 34439 Istanbul Turkey e-mail:
[email protected] Fatih TAŞÇI Yıldız Technical University Faculty of Chemistry and Metallurgy Department of Mathematical Engineering Davutpasa Campus No:127 34210 Topkapi Istanbul Turkey e-mail:
[email protected] Vedat TAŞKIN Trakya University Department of Mechanical Engineering 22030 Edirne Turkey N. TAYŞİ University of Gaziantep Department of Civil Engineering 27310 Gaziantep Turkey e-mail:
[email protected] Hamide TEKELİ Suleyman Demirel University Department of Civil Engineering Isparta Turkey e-mail:
[email protected] Mevlüt TEYMÜR Istanbul Technical University Department of Mathematics Faculty of Sciences and Letters 80626 Maslak Istanbul Turkey e-mail:
[email protected] Ales TONDL Zborovska 41 Prague 5 CZ-150 00 CZECH REPUBLIC e-mail:
[email protected] Umut TOPAL Karadeniz Technical University Department of Civil Engineering 61080 Trabzon Turkey e-mail:
[email protected] Özgür TURHAN Istanbul Technical University Faculty of Mechanical Engineering 34439 Gümüşsuyu Istanbul Turkey e-mail:
[email protected] Halit S. TÜRKMEN Istanbul Technical University Faculty of Aeronautics and Astronautics 34469 Maslak Istanbul Turkey e-mail:
[email protected] Kurtuluş SOYLUK Gazi University Department of Civil Engineering 06570 Maltepe Ankara Turkey e-mail:
[email protected] F. Demet ÜLKER Middle East Technical University Aeronautical Engineering Department Ankara Turkey . . . . . . . . . . . LIST OF PARTICIPANTS 557 Habib UYSAL Atatürk University Engineering Faculty Department of Civil Engineering Erzurum Turkey e-mail:
[email protected] Ümit UZMAN Karadeniz Technical University Gümüşhane Engineering Faculty Department of Civil Engineering 29000 Gümüşhane Turkey e-mail:
[email protected] Jizeng WANG Max Planck Institute For Metals Research Heisenbergstrasse 3 D-70569 Stuttgart Germany Nazmiye YAHNİOĞLU Yıldız Technical University Department of Mathematical Engineering Faculty of Chemistry and Metallurgy Davutpasa Campus 34210 Topkapi Istanbul Turkey e-mail:
[email protected] Mustafa YAVUZ Istanbul Technical University Faculty of Civil Engineering 34469 Maslak Istanbul Turkey e-mail:
[email protected] Vladimir V. YELISEYEV St. Petersburg State Polytechnical Univer- sity Computer Technologies in Engineering De- partment Polytechnicheskaya st. 29 St. Petersburg 195251 Russia Vebil YILDIRIM Çukurova University Faculty of Engineering and Architecture Department of Mechanical Engineering 01330-Balcal Adana Turkey e-mail:
[email protected] Hasan YILDIZ Ege University Department of Mechanical Engineering Bornova Izmir Turkey e-mail:
[email protected] Hüseyin Murat YÜKSEL Istanbul Technical University Faculty of Aeronautics and Astronautics 34469 Maslak Istanbul Turkey Tatiana V. ZINOVIEVA St. Petersburg State Polytechnical University Computer Technologies in Engineering Department Polytechnicheskaya st. 29 St. Petersburg 195251 Russia e-mail:
[email protected] Bahar ZORA Istanbul Technical University Faculty of Mines 34469 Maslak Istanbul Turkey Hakan UREY Koç University Department of Electrical Engineering Istanbul Turkey . . . . . . . . ı . springer proceedings in physics 64 Superconducting Devices and Their Applications Editors: H. Koch and H. Lübbing 65 Present and Future of High-Energy Physics Editors: K.-I. Aoki and M. Kobayashi 66 The Structure and Conformation of Amphiphilic Membranes Editors: R. Lipowsky, D. Richter, and K. Kremer 67 Nonlinearity with Disorder Editors: F. Abdullaev, A.R. Bishop, and S. Pnevmatikos 68 Time-Resolved Vibrational Spectroscopy V Editor: H. Takahashi 69 Evolution of Dynamical Structures in Complex Systems Editors: R. Friedrich and A. Wunderlin 70 Computational Approaches in Condensed-Matter Physics Editors: S. Miyashita, M. Imada, and H. Takayama 71 Amorphous and Crystalline Silicon Carbide IV Editors: C.Y. Yang, M.M. Rahman, and G.L. Harris 72 Computer Simulation Studies in Condensed-Matter Physics IV Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 73 Surface Science Principles and Applications Editors: R.F. Howe, R.N: Lamb, and K. Wandelt 74 Time-Resolved Vibrational Spectroscopy VI Editors: A. Lau, F. Siebert, andW.Werncke 75 Computer Simulation Studies in Condensed-Matter Physics V Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 76 Computer Simulation Studies in Condensed-Matter Physics VI Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 77 Quantum Optics VI Editors: D.F. Walls and J.D. Harvey 78 Computer Simulation Studies in Condensed-Matter Physics VII Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 79 Nonlinear Dynamics and Pattern Formation in Semiconductors and Devices Editor: F.-J. Niedernostheide 80 Computer Simulation Studies in Condensed-Matter Physics VIII Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 81 Materials and Measurements in Molecular Electronics Editors: K. Kajimura and S. Kuroda 82 Computer Simulation Studies in Condensed-Matter Physics IX Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 83 Computer Simulation Studies in Condensed-Matter Physics X Editors: D.P. Landau, K.K. Mon, and H.-B. Schüttler 84 Computer Simulation Studies in Condensed-Matter Physics XI Editors: D.P. Landau and H.-B. Schüttler 85 Computer Simulation Studies in Condensed-Matter Physics XII Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 86 Computer Simulation Studies in Condensed-Matter Physics XIII Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler 87 Proceedings of the 25th International Conference on the Physics of Semiconductors Editors: N. Miura and T. Ando 88 Starburst Galaxies Near and Far Editors: L. Tacconi and D. Lutz 89 Computer Simulation Studies in Condensed-Matter Physics XIV Editors: D.P. Landau, S.P. Lewis, and H.-B. Schüttler CONTENTS Preface Asynchronous and antisynchronous effects of ground motion on the stochastic response of suspension bridges European union framework programme 7 building the Europe of knowledge Dynamic response of rock-fill dams to asynchronous ground motion A comparative study on the dynamic analysis of multi-bay stiffened coupled shear walls with semi-rigid connections Free vibrations of cross-ply laminated non-homogeneous composite truncated conical shells Symmetric and asymmetric vibrations of cylindrical shells An active control algorithm to prevent the pounding of adjacent structures Vibration control of non-linear buildings under seismic loads Vibrations of damaged 1D-3D multi-structures Sizing of a spherical shell as variable thickness under dynamic loads Vibration analysis of simply supported functionally graded beams Damping effects of rubber layer in laminated composite circular plate during forced vibration A critical study on the application of constant deflection contour method to nonlinear vibration of plates of arbitrary shapes Nonlinear wave equations and boundary control using visco elastic dampers Effect of vibrations on transportation system Identification of bone microstructure from effective complex modulus Beam vibrations with non-ideal boundary conditions A two-fractal overlap model of earthquakes Dynamics of structures with bistable links Real-time seismic monitoring of the new Cape Girardeau (MO) bridge and recorded earthquake response Earthquake response of masonry infilled precast concrete structures Nonlinear waves in fluid-filled elastic tubes: a model to large arteries The nonlinear axisymmetric vibrations of circular plates with linearly varying thickness under random excitation Dynamic analysis of a helicopter rotor by Dymore program Parametric stiffness excitation as a means for vibration suppression A formulation of magnetic-spin and elasto-plastic waves in saturated ferrimagnetic media Earthquake response of suspension bridges The microwave sensor of small moving for the active control of vibrations and chaotic oscillations modes Forced oscillations of tube bundles in liquid cross-flow The exact and approximate models for the vibrating plate partially submerged into a liquid A computational tool based on genetic algorithm for determining optimum shapes of vibrating planar and space trusses On the dynamical stress field in the pre-stretched bilayered strip resting on the rigid foundation Dynamic response of a rectangular plate-column system on a tensionless elastic foundation Vibration suppression of an elastic beam via sliding mode control Thermoelastic stress analysis for linear elastic bodies Discrete and continuous mathematical models for torsional vibration of micromechanical scanners Physical modeling of stationary and impulsive processes for large-scaled constructions of fluid elastic systems Thermal stresses and nonlinear thermal deformation analysis of shallow shell panel Nonlinear thermal vibrations of a circular plate under elevated temperature Flexural-torsional coupled vibration analysis of a thin-walled closed section composite Timoshenko beam by using the differential transform method Estimation of microstretch elastic moduli by the use of vibrational data Vibrations of a circular membrane subjected to a pulse A method of discrete time integration using Betti's reciprocal theorem Bridges in vibrated granular media Solutions for dynamic variants of Eshelby's inclusion problem Note on large deflections of clamped elliptic plates under uniform load Localised defect modes and a macro-cell analysis for dynamic lattice structures with defects Acoustic wave propagation through discretely graded materials Use of mechatronic components in rotating machinery Computational limits of FE transient analysis The effect of non-linear behavior of concrete on the seismic response of concrete gravity dams Calculation of microwave plasma oscillations in high temperature superconductors One-dimensional wave propagation problem in a nonlocal finite medium with finite difference method Dynamic response analysis of rocking rigid blocks subjected to half-sine pulse type base excitations A new spectral algorithm for 3-D wave field in deep water Experimental and numerical assessment of vibro-acoustic behavior of rubber-damped railway wheels Charge simulation method applied to vibration problems Nonlinear stability and dynamics of laminated composite plates and shells On the stability of a vibroisolation system with more degrees of freedom The stability and vibration of conical shells composed of SI[sub(3)]N[sub(4)] and SUS304 under axial compressive load Dynamic buckling of elasto-plastic cylindrical shells under axial load Dynamic buckling analysis of imperfect elastica Specification of flow conditions in the mathematical model of hydraulic damper Nonlinear transient analysis of rectangular composite plates On the "resonance" values of the dynamical stress in the system comprises two-axially pre-stretched layer and half-space Optimization of vibrating arches based on genetic algorithm Propagation of long extensional nonlinear waves in a hyper-elastic layer Quenching of self-excited vibrations in a system with two unstable vibration modes Free vibration analysis of laminated plates using first-order shear deformation theory The generalized Bolotin method as an alternative tool for complete dynamic stability analysis of parametrically excited systems: application examples Coupling effects between shaft-torsion and blade-bending vibrations in rotor-blade systems Forced vibration of the pre-stretched simply supported strip containing two neighbouring circular holes Free vibration of curved layered composite beams Vibrations of beam constructions submerged into a liquid Vibration behavior of composite beams with rectangular sections considering the different shear correction factors Air blast-induced vibration of a laminated spherical shell Author Index A B C D E F G H I K L M N O P R S T U W Y Z List of Participants