Laboratory Tests on Compensation Grouting, The Influence of Grout Bleeding

1 INTRODUCTION Compensation grouting or fracture grouting has been successfully applied in several projects to compen- sate for surface settlements induced by for example tunnelling. Fracture grouting is planned to compen- sate possible settlements of buildings in Amsterdam during the construction of the tunnels for the North/South Metro Line. Fracture grouting uses hydraulic fracturing to get a heave that can compensate the settlement. Hydrau- lic fracturing of competent material has been studied extensively in the oil industry where it is used suc- cessfully to enhance the capacity of oil wells. Hy- draulic fracturing can also be an unwanted pheno- menon in tunnelling or horizontal drilling when high pressures can cause a blow-out that starts with a hy- draulic fracture (Bezuijen & Brassinga 2001). The area where heave can be expected is deter- mined by the shape of the fracture. Some pictures taken near Antwerp Station and a theoretical model by Te Grotenhuis (2004) showed thin fractures propagating over a large distance. However, experi- ments on fracture grouting in sand performed by Ga- far & Soga (2006) and experiments reported by Kleinlugtenbelt (Kleinlugtenbelt 2005 and Kleinlug- tenbelt et al. 2006) showed no fractures but a shape that looks as a perfect ball or an irregular ball with something that looked like the beginning of a frac- ture, see Figure 1. That sand can be fractured was proven by expe- riments with different fracture materials (Khodave- rdian & McElfresh 2000, Pater & Bohloli, 2003 and Bezuijen, 2003iii). This paper starts with a conceptual idea how frac- turing of sand can occur. This idea shows what pa- rameters are of importance to determine the shape of the fracture. It describes briefly the set-up and the results of the model tests on fracturing that has been performed and the tests to characterize the injection material and will end up with conclusions. Figure 1. Grout body formed in sand during a compensation grouting test. 2 CONCEPTUAL MODEL 2.1 fracturing When a grout is injected into an opening of sand, as is done by compensation grouting, there will be an Laboratory tests on compensation grouting, the influence of grout bleeding A. Bezuijen GeoDelft, and Delft University of Technology, Delft, the Netherlands M.P.M. Sanders Delft University of Technology, Delft, Netherlands D. den Hamer GeoDelft A.F. van Tol GeoDelft and Delft University of Technology, Delft, Netherlands ABSTRACT: The process of compensation grouting has been investigated by a series of laboratory tests. In previous test series all tests resulted in an injection body of grout that had an irregular shape relatively close to the injection point. Shape of the fractures and necessary injection pressure indicate that the process was more comparable to compac- tion grouting than to fracture grouting. It was assumed that this was caused by bleeding of the grout mortar. Therefore a new test series was performed with larger variations in permeabilities and amounts of solid material between the injec- tion fluids of different tests. Here it appeared that the injected ‘grout’ sometimes created fractures. Calculation models are described to calculate the interaction between soil and the injection fluid and a conceptual model what parameters determine the shape of the injected volume. elastic deformation of the opening, the opening will increase. Assuming the injection fluid (the grout) forms a plastering layer we can describe this first de- formation with cavity expansion theory. The radial stress will increase and the tangential stress will de- crease. This will go on until, according to the Mohr- Coulomb criterion for cohesionless soil, the stress ratio (σr/σθ) in the sand at the boundary of the injec- tion hole fulfils the relation: φ φ σ σ θ sin1 sin1 − + =r (1) Where φ is the friction angle of the sand. For sand with a friction angle of for example 35 degrees this ratio is 3.7. If the radial stress increases further, the stress ratio will remain constant and plastic deforma- tion will occur. Classical cavity expansion theory assumes that there will be a radial expansion, which is uniform in all directions. However, in reality the sand is never perfectly homogeneous. Looking at the scale of the grains at the boundary of the opening there will be some grains that are in closer contact and some be- tween which there is some space, see Figure 2. Pf σθ sand grains injection fluid Pf σθ sand grains injection fluid deformation according to cavity expansion localized deformation Pf Pf Pf Pf before deformation after deformation Figure 2. Sketch with possible deformation modes of the injec- tion hole. If the radial pressure is exerted by the injection fluid, the pressure in the fluid will be the same in all direc- tions. As a result the pressure working on two grains on the boundary of the opening, which are not in di- rect contact will lead to a tangential stress that wants to remove the grains form each other and that is higher than the opposing tangential stress from the soil. As a result the grains will be pushed apart. A micro mechanical calculation has shown that this situation, in which 2 grains are pushed apart, is sta- ble in case the soil has a plastering function and most forces work on the grains at the boundary of the opening. From hereon this process can go further and more grains can be pushed apart by the injection fluid until a full fracture occurs. The model as described above, explains why the injection pressure is difficult to predict. There is no clear boundary where a fracture starts. The injection pressure must be high enough to create plastic de- formation in the sand, but it is always possible that some plastic deformation occurs before a fracture starts. Literature presents the injection pressure in most cases as a ratio of the vertical stress on the soil sample (Bezuijen et al. (2003i). It appears from our tests and literature that this ratio can vary between 3 and 20. However, the highest values do not produce real fractures. The reason for that is, that there will always be cavity expansion in case pressures are high enough. Fracturing is an additional failure me- chanism that can only occur at a lower injection pressure than cavity expansion. The conceptual model described above also ex- plains why the properties of the injection fluid are of importance. The injection pressure has to work close to the boundary. When there is too much plastering than the injection pressure is not applied on the sand, but on the plaster, see Figure 3. In that case the tan- gential force that is exerted on the grains at the boundary of the opening is much less and depends on the mechanical properties of the plastered materi- al. Pf σθ sand grains injection fluid granular material plastered material Figure 3. Influence of plastering. The mechanism described here for injecting in loose granular material is different for cohesive soils and rock where tension cracks can occur. In sand there will never be tension, because there is no cohe- sion in the material and therefore plastic deforma- tion will occur in the material before tension stresses develop. 2.2 Injection fluid The injection fluid used in compensation grouting is a mixture of water, bentonite and cement. Plastering can occur by 2 different mechanisms: 1 Bleeding of the injection fluid, as described by Bolton and McKinley (1997). The amount of bleeding depends on the injection pressure, the duration and the permeability of the plastered ma- terial. 2 Leak-off of the injection fluid. This phenomenon occurred when most of the particles in the fluid are very small. In that case the injection fluid will penetrate into the sand. In the situation that there are also larger particles, these will not penetrate, but will end up at the boundary between the fluid and the soil, see Figure 4. The injection fluid for compensation grouting has 3 different components. Looking at the particle size only, the cement particles will not penetrate into the sand, but the very small bentonite particles will. However, from literature (Mitchell, 1976) it is known that these particles cannot be dealt with sepa- rately. The calcium in the cement leads to floccula- tion of the bentonite, which results effectively in larger particles and therefore in a larger permeability of the cake. Due to this the bleeding will increase but the leak-off will decrease. sand larger grains from injection liquid P x injection liquid finer grains that penetrate into the sand Figure 4: Plastering that can occur during leak off due to larger grains in the liquid that cannot penetrate into the sand and sketch of the pressure distribution over the injection liquid. 3 MODEL EXPERIMENTS 3.1 Test set-up A circular container with a diameter of 0.9 m was used for the tests (see Figure 1). This container was filled with saturated sand up to 0.84 m height. A PVC plate was placed on top of the sand sample. There was an air tight connection between the plate and the container by means of a rubber ring. It was therefore possible to pressurize the sand sample, us- ing air pressure, simulating sand at a larger depth. The injection nozzle is comparable to the system de- veloped for compensation grouting. A pipe with a rubber sleeve has been used (see Figure 2). The sleeve will only allow outflow of the grout when the grout pressure is higher than the soil pressure. The injection nozzle was located 0.37 m above the bot- tom of the tank. Tests were performed with Baskarp sand (d50 = 130 μm). The sand was wet pluviated into water in the container. The loose sand was densified by dropping the whole container over a few centimetres several times depending on to the required density. The sand was ‘pre-stressed’ by applying a high ver- tical pressure before the test to achieve a higher K0. This pre-stressing was only partly successful. The K0 achieved with this procedure was around 1. Dur- ing grout injection K0 rises rapidly to values well above 1 and even to 4.5 in the case of dense sand. Pore pressure transducers and total stress transducers were installed at various locations in the container, see Figure 7. air pressure 100 kPa grout reservoir grout pump plunger filter sand filter chamber filled with water water level drainage water level TAM 480 360 900 Figure 5. Set-up of the experiments. Note that changes in pore volume and sand volume can be measured as changes in the water levels. tube rubberring ring rubberinjection holes Figure 6. Injection system with rubber sleeve. The steel rings were placed on the tube to prevent grout flow along the tube. -50 0 Z (m m ) -300 -200 -100 0 100 Y (mm) P V H inj 123 4 Figure 7. Position of the instruments with respect to the injec- tion tube (inj). P are the pore pressure gauges, V measures the vertical pressure and H the horizontal. Special attention was paid to the measurement of the volumes. The increase in volume of the sample due to the injection and the drainage of pore water was measured continuously during the tests, more details on the measurements of the volume can be found in (Kleinlugtenbelt, 2005). Grout was injected by means of a plunger pump. This pump pumped water and by means of a bladder the grout was pumped into the system (the sketch in Figure 5 does not show the bladder). The bladder system was installed to avoid damage to the pump by the granular particles in the grout. The maximum injection pressure of the injection pump was 25 bars. The grout was allowed to harden for one day af- ter the test before the sand was washed away and the shape of the injected grout became visible. There was no hardening of the grout in the test without cement. After these tests the capillary forces in the sand were used to see the shape of the fracture. 3.2 Classification experiments The properties of the injection fluids were deter- mined for each test. Viscosity and yield point were determined using a Fann V.G. viscometer that was operated at 300 and 600 rpm. Leak-off and bleeding were determined using a modified consolidation test, see Figure 8. Spalte 2 air pressure supply grout sand 39 50 78 collection and weighing of liquid membrane end test membrane and plate start test plastic plate end test wire mesh Figure 8. Sketch of the apparatus used for the bleeding tests, see also text (dimensions in mm). A cylindrical pressure vessel with an inward diame- ter of 78 mm and a height of 89 mm can be pressu- rized on one side and has drainage capabilities on the other side. A fine wire mesh allows water to flow through, but prevents sand particles to pass. In an experiment 50 mm of medium fine sand (d50=130 μm) is applied on the wire mesh and densified until the maximum density. 39 mm of injection fluid was applied and the pressure vessel was sealed off with a rubber membrane and a plastic plate. The latter is to achieve an even settlement of the grout. An air pres- sure of 4 bar is applied on top of the membrane and the amount of liquid expelled is continuously meas- ured with a balance. The membrane and plastic plate will settle as is indicated in the figure during the ex- periment. The experiment is continued until the amount of expelled water remains constant. The result of the test depends on the type of injec- tion fluid. For a grout without any leak- off the amount of expelled water is proportional with the square root of time, as in agreement with the find- ings of Bolton and McKinley (1997) and can be de- scribed with the formula: t n nn kx i ei φΔ − − = 1 2 (2) Where x is the thickness of the consolidated layer, ni the initial porosity, ne the porosity after consolida- tion, k the permeability of the consolidated layer,Δφ the difference in piezometric head over the column and t the time. The formula was used to describe bleeding tests on tail void grout and appeared to de- scribe the bleeding process quite well (Bezuijen & Talmon, 2003ii). In case there is leak-off, there is a quick dis- charge in the first few seconds when leak-off is do- minant, after that the curve follows again the curve that is expected from the bleeding. 3.3 Tests results The typical conditions of the tests performed with the set-up of Figure 5 are listed in Table 1. The idea for this test series was that the grout properties are determining the result of the test, more than varia- tions in other parameters (changing the sand into silt of clay will also have an influence, but the aim was to create compensation grouting in sand). For that reason most parameters were kept constant but a wide variety of ‘grout mixtures’ was used in the tests as an injection fluid. To compare with other re- sults, injection fluids without cement were used and in one case we used X-linked gel, a fracture fluid that is used in the oil industry to create fractures. Table 1. Parameters of Tests performed. No Yield stress (Pa) Vis c (Pas) σv peak (kPa) σh peak (kPa) Perm. x10-12 (m/s) Peak press. (kPa) Frac. 1 39 7.5 145 250 211 1600 No 2 35 17 150 270 17 1200 No 3- 1 98 11 195 300 35 1200 No 3- 39 7.5 210 400 470 1060 No 2 4 n.d. n.d. 160 225 5.6 750 Yes 5 8 9 150 170 6 1200 Yes 6 51 21 170 200 50 1700 Yes Notes: Yield stress = the yield stress of the grout, Visc = the viscosity of the grout, σv = the vertical total stress, σh = the ho- rizontal total stress, Perm = the permeability of the grout filter cake, peak pressure = the maximum pressure measured and Frac = the fracturing of the sand. n.d = not determined. The injection rate in all tests was 167 ml/s. All tests were run with a relative density of the sand sample of around 65%. The confining stress on the sand was 100 kPa on all tests and the pore pressure in the sand was only a few kPa. Table 2. Additives injection fluid No Bentonite (%) Cement (%) Silica flower (%) Fly ash (%) Retarder (%) 1 7 - - 33.3 5.3 2 7 - 33.3 - 5.3 3- 1 12 - 0.5 33.3 5.5 3- 2 7 - 0.5 33.3 5.3 4 - - 0.5 - - 5 6.2 0.5 - - 6 6.2 0.5 - - - 7 6.2 5 - - Note: All values are related to the real water content of the grout mortar. The first two tests resulted from the conclusions of the test series described before (Kleinlugtenbelt et al. 2006). A very impermeable grout mortar was used, but this did not lead to fractures, probably due to the relative high percentage of solids in the injec- tion fluids. Even with a limited amount of leak-off this leads already to a plastering layer that prevents fractures, see also the discussion section. Two injections were performed with the sand model of test 3. It appeared that the second injection was capable to fracture the grout body that was formed by the first injection, but was not able to fracture the sand any further, see Figure 9. Figure 9. Injection results test 3. The fourth test was performed with X-linked gel. This test was performed to test the model set-up. X- linked gel produces fractures in a comparable model set-up (De Pater & Bohloli, 2003). We also found fractures in our test set-up for this injection fluid, see Figure 10. The last two tests also resulted in fractures, but only to a limited extent, because in both cases there was a considerable leak-off, see Figure 11. Figure 10. Fractures obtained using X-linked gel.. Figure 11. Results test 6 with some fractures and a considera- ble leak-off. 4 DISCUSSION The results of these tests, together with the results of Kleinlugtenbelt et al. (2006) and De Pater & Bohloli (2003) show that it is not easy to get long hydraulic fractures in sand. Fractures as predicted by the mod- el of Grotenhuis (2004) could not be produced in the tests using mixtures of bentonite slurry and cement or silica flour. The reason is that the traditional grout mixtures (5% bentonite and a WCR of 3 or less) are too per- meable, when pressurized the consolidated material will form a cake, which will prevent a fracture as de- scribed in Section 2. This can be shown by using Equation (2) that is plotted in Figure 12. Calcula- tions were run for 0.5 s, approximately the time ne- cessary to pressurize the injection hole. It can be seen that in this period the consolidated layer is al- ready significant compared with the grain size di- ameter (d50=130 μm). 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 th ic kn es s co ns ol id at ed la ye r( m m ) 0.0 0.1 0.2 0.3 0.4 0.5 time (s) WCR 1 WCR 3 k =5*10 m/sg -8 Figure 12: Plot of equation (1) for relevant permeabilities and injection times. Results show that with WCR 3 a filter cake of several tenths of mm is formed with one second. The mixture in test 1 and 2 did not fracture be- cause of the leak-off. This leak-off causes the fly ash and the silica flour to gather at the interface between the injection fluid and the sand. The leak-off after the test was approximately 10 mm. The ratio (with respect to the weight) between water and fly ash or silica flour was 3. Neglecting the weight of the ben- tonite (only 7%) the ratio in volume for silica flour is 3*2650/1000=7.95 (the correction for the differ- ence in density), for fly ash this is 3*2900/1000=8.70. Assuming that the porosity of the silica flour after sedimentation is roughly com- parable to the porosity of the sand this means that a leak-off of 10 mm into the soil will result in a layer of silica flour of 1.3 mm, which is again much more than the grain size of the sand. Leak-off itself does not hamper fractures, as long as no filter cake is formed, but leads to less fluid that can contribute to the fractures. This is shown from the results from Test 5 and 6. The results show that fracture grouting in clean sand will be more an exception than the normal re- sult of a compensation grouting process. Projects where, according to the contractor, settlement in sand is corrected by fracture grouting will in most cases be examples of compaction grouting, unless there is cohesion in the soil or the soil is much more impermeable than the soil tested here (with a per- meability of 10-4 m/s). Since compensation grouting projects are executed deep under the soil surface it is seldom known what is created, only in few cases where there was the opportunity to investigate the fracture afterwards because some soil has to be re- moved. Some projects report high injection pres- sures, which indicate that there is a kind of compac- tion due to cavity expansion. How compensation is achieved, is of importance for the field situation, because it indicates how loca- lized the compensation will be. Fracture grouting will result in heave over a wider area than compac- tion grouting. 5 CONCLUSIONS From the tests performed we came to the following conclusions: 1. The grout as used in practice formed a dehydrated cake between the injection fluid and the sand where through a fracture could not propagate. 2. More cement in the grout mortar results in a thicker and more permeable filter cake. 3. Tests implemented with a low amount of particles and impermeable filter cake gave small fractures. From this we concluded that the dry mass entity of the hy- draulic fracture fluid also has a major influence on the mechanism of fracturing. 4. Too much leak-off decreases the amount of injection fluid that can really contribute to the fracturing, where through the fracture stays small. (Cement can counte- ract the leak-off through the acting reactions of cal- cium.) REFERENCES Bezuijen, A. & Brassinga, H.E. 2001. Blow-out pressures measured in a centrifuge and in the field Proc. XIII ECSM- GE 2001, Istanbul. Bezuijen, A., Pruiksma, J.P. & Pater, C.J. 2003i. Maximum pressures in tunnelling limited by hydraulic fractures, (Re)claiming the underground Space, Proc ITA 2003, Sa- veur, Amsterdam: 1007-1008. Bezuijen, A. & Talmon, A.M. 2003ii. Grout the foundation of a bored tunnel. Proc ICOF. Dundee: Thomas Telford Bezuijen A. 2003iii, Hydraulic fracturing experiments ant low stresses in sand. rep. Delft University of Technology TA/TG/02-03. Bolton M.D. & McKinley J.D., 1997, Geotechnical properties of fresh cement grout – pressure filtration and consolidation tests, Géotechnique 47, No 2, 347-352. Gafar K. & Soga K. 2006. Fundamental Investigation of soil- grout interaction in sandy soils, Un. of Cambridge report, August. Grotenhuis te, R. 2004. Fracture grouting in theory, MSc the- sis, Delft University of Technology. Khodaverdian M. & McElfresh P. 2000. Hydraulic fracturing simulation in poorly consolidated sand: Mechanism and consequences, Proc. Conf. Soc. of Petroleum Engineers, Dallas no. 63233. Kleinlugtenbelt, R. 2005. Compensation grouting, laboratory experiments in sand. MSc thesis, Delft Un. of Technology. Kleinlugtenbelt, R., Bezuijen, A. & Tol A.F. van. 2006. Model tests on compensation grouting. Proc. ITA 2006, Seoul. Mitchell, J.K. 1976. Fundamentals of soil Behavior, University of California, Berkely ISBN 0-471-61168-9. Pater de, H.J. & Bohloli, B. 2003. Fracturing unconsolidated rock, rep. Delft University of Technology TA/TG/02-03.