Flow Measurement and Instrumentation 12 (2001) 89–99 www.elsevier.com/locate/flowmeasinst The manufacturing of ultrasonic gas flow meters Jan G. Drenthen *, Geeuwke de Boer Instromet International N.V., Rijkmakerlaan 9, B-2910 Essen, Belgium Abstract From their introduction, ultrasonic flow meters have received a rapid acceptance as being one of the favored measurement methods for high accuracy custody transfer applications in high-pressure gas transmission systems. There are many benefits when using ultrasonic technology; increased rangeability and capacity over conventional measurement technology with unparalleled accuracy are near the top of the list. But in many cases, even more important are the cost savings obtained due to the decrease in maintenance costs and savings in compressor fuel cost by the reduction of the pressure drop through the station. Key elements in the success of the ultrasonic technology are the manufacturing methods and procedures that result in tight tolerances in the geometry of the meter. Whereas the accuracy of the meter is mainly dependent on the quality of the geometry and accuracy of the time measurement, the stated performance of the meter can be guaranteed based on a dry calibration only; a practice identical to the widely accepted orifice measurement. The purpose of this paper is to focus on the influence of the manufacturing tolerances on the uncertainty of the measurement, the dry calibration procedure and the final comparison with the results obtained after wet calibrations. 2001 Elsevier Science Ltd. All rights reserved. Keywords: Ultrasonic; Gas measurement; High pressure meter body manufacturing 1. Introduction The principle of an ultrasonic flow meter is illustrated in Fig. 1. Two transducers are installed in the flow line in such a way that ultrasonic sound pulses emitted from one transducer can be received by the other transducer, Fig. 1. Principle of an ultrasonic flow meter. * Corresponding author. Tel.: +32-36-667-34-40; fax: +32-3-670- 05-60. E-mail address:
[email protected] (J.G. Drenthen). 0955-5986/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S0955- 59 86 (01)00 00 3- 6 thus creating an acoustic path. The transducers alter- nately transmit and receive pulses within a few millise- conds. The ultrasonic sound pulses travel, with respect to the gas, at the speed of sound. The velocity of a sound pulse along the acoustic path traveling downstream is increased with the projection of the gas velocity onto the acoustic path. The velocity of the sound pulse traveling upstream along the acoustic path is decreased with a pro- jection of the gas velocity onto the acoustic path. This results in travel times for the upstream and downstream direction as: tdown� L C+Vm cos j (1) tup� L C−Vm cos j , (2) where: L=length of the acoustic path; C=speed of sound in the medium (gas); Vm=velocity of the moving medium (gas); and ϕ=angle between acoustic path and a vector representing the direction in which the medium moves. Using Eqs. (1) and (2) the following expression for the measured gas velocity can be derived: 90 J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 2. Flow profile correction factor. Vm� L 2 cos j� 1tdown� 1tup�. (3) It is important to notice that the speed of sound in the gas is eliminated in this expression. This means that the measurement of the gas velocity is independent of the properties of the gas such as pressure, temperature and gas composition. To measure the total volume flow through the pipe, the flow velocity measured across the ultrasonic path has to be multiplied by the cross section of the pipe. When the gas velocity is equal over the whole cross section, that is, has a uniform flow profile, the flow calculated in this way would have the exact value. As this is not the case by law of nature the measurement has to be corrected with a factor K which is related to the shape of the flow profile (see Fig. 2). In this figure, Vm rep- resents the average gas velocity as perceived by the ultrasonic flow meter, which is the linear weighted gas velocity averaged along the acoustic path. This results in the following expression for the gas flow rate: Q� L2 cos jAK� 1tdown� 1tup� (4) where A is the cross section of the pipe; and K is the flow profile correction factor. From other studies, literature and our own research, Fig. 3. Profile correction factor K. Instromet established a relationship between the Reyn- olds number and the flow profile correction factor (also referred to as Reynolds factor) K, of which an example is shown in Fig. 3. In practical situations the actual flow profile may show some variation resulting in an uncertainty in the flow calculation. Based on numerous tests, in Fig. 4 the response of a single axial path flow meter is shown using the K factor. From this graph it is clear that the uncer- tainty in the flow profile correction factor for a single path meter is not less than ca 1%. The reason for this is that the actual flow profile in many cases is affected by swirl, asymmetry and pulsations and thereby deviates from the assumed ideal profile. By adding other paths to detect the type and strength of the flow profile distor- tion, the uncertainty can be reduced considerably. For custody transfer applications the level of uncer- tainty of a single path flow meter is normally not accept- able and a multi-path meter is required. There, by the implementation of integration techniques the data of multiple acoustic paths is used to improve the accuracy of the flow profile correction. In Fig. 5, the path configuration of a five-path flow meter is shown, represented by the following equation: Q�� L2 cos jAK� 1tdown� 1tup��Mf. (5) This expression is identical to that of a single path meter except that the part between the brackets {} represents the integration using all acoustic paths. The parameter Mf is a correction factor and used only when — based on a flow calibration — the reading of the meter is adjusted. This parameter is first set to a standard default value, usually 1.000. When the meter is wet cali- brated, this parameter corrects the error due to all other parameters and variables involved. The data presented further on in this paper is based upon the results of numerous flow calibrations of ultra- sonic meters. As a result of these flow calibrations the default value of the correction factor Mf is adjusted. The variation in the correction factors as found will reflect both the uncertainty due to the geometry and dimen- 91J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 4. Single path meter performance. Fig. 5. Path configuration of a five-path Q.Sonic-5 ultrasonic flow meter. sional parameters resulting from the manufacturing pro- cedures as well as the other sources of uncertainty. The frequency distribution of this adjustment factor is a good tool for verifying the overall uncertainty for meters manufactured without flow calibration. 2. Uncertainty and the influence of the manufacturing tolerances From Eq. (5) it can be seen that the total uncertainty of the measurement is dependent on those of all parameters involved, namely: � the K-factor (the flow profile); � the time measurement; � the geometry; 2.1. Uncertainty of profile correction factor The Reynolds (profile) correction factor for a single path meter is estimated, based on the graph as presented in Fig. 4, to have an uncertainty of ±1.0%. Based upon Instromet’s research and hundreds of test results with Instromet’s multi path meters and the path configuration as implemented, the uncertainty of the Reynolds (profile) correction factor can be estimated to be ca 0.3% for a five-path meter and ca 0.4% for a three-path meter. 2.2. The uncertainty due to the time measurement The travel time measurements and therefore the flow measurement are directly proportional to the clock sig- nal. Based a highly accurate and stable quartz crystal (accuracy ±50 ppm or 0.005%), the uncertainty due to it can be ignored. The only time error that can occasion- ally be of some importance is the zero error affecting the measurement at the very the low end. 2.3. The geometry 2.3.1. Uncertainty due to the manufacturing tolerances and the connecting pipes Not only the meter body, but also the differences between the meter body and the surrounding piping can 92 J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 influence the final measurement uncertainty and have therefore to be taken into account. The factors to be taken into consideration are: � The difference between the internal diameter of the meter body and that of the connecting pipe. � The difference in roughness between the meter and the connecting pipe. � The tolerances in the manufacturing. 2.3.2. The difference between the internal diameter of the meter body and that of the connecting pipe The influence of the difference between the internal diameter of the meter body and that of the connecting pipe on the meter error is dependent on the acoustic path configuration. Acoustic paths that are located closer to the pipe wall are more vulnerable to step changes in the internal diameter. In order not to favor one single sup- plier and to try to incorporate all current designs, the present regulations and standards like the AGA-9 are much stricter than necessary and allow only a step change in the internal diameter of 1%. However, tests conducted with a Q.Sonic-5 ultrasonic flow meter showed no detectable influence of step changes up to 2%; a 5% step change resulted in a just detectable change in the error curve of ca 0.1–0.15%. To avoid the possible influence of step changes and still allowing the possibility of coping with a wide var- iety of pipe schedules a chamfering can be applied. As long as the angle of chamfering is �7° the boundary layer will not separate and the chamfering will not influence the measurement result (see also [1]). In prac- tice, a chamfering angle of ca 4° will normally be more than adequate to get the desired match. 2.3.3. The influence of roughness of the connecting pipe and the meter Rick Wilsack and Huib Dane have investigated the influence of the roughness of the pipe wall of the con- necting pipe on the measurement result (see also [2]). Tests conducted with even highly corroded and to be rejected pipes showed a hardly detectable influence on a meter performance of a Q.Sonic-5. With all other results being within the order of the short term repeat- ability range of the facility, a maximum shift in the order of 0.1–0.2% could be detected with the heavily cor- roded pipe. The influence of the wall roughness of the meter body is part of an ongoing research project. In exploring the extreme, a 12� meter has been coated on the inside with steel grid. In this far from normal situation, the cali- bration curve shifted ca 0.3% in reference to the original smooth pipe wall. With the flow profile changing gradu- ally as a function of the wall roughness, from these tests it might be concluded that in a practical situation the effects will be much smaller. In practice it is difficult to separate the influence of fouling and that of the wall roughness. A small layer of dirt, that effects the wall roughness also reduces the inside area. With a thickness of only 0.2 mm. such a layer already reduces the inside area in a 12� pipe ca 0.3%. 2.3.4. The tolerances in the manufacturing As far as the geometry and dimension of the meter body is concerned the relevant parameters that have an impact with respect to the accuracy of an ultrasonic flow meter are [see also Eq. (5)]: L acoustic path length ϕ angle of acoustic path A cross section of the pipe The acoustic path parameters are related to the position of the front side of the ultrasonic transducers; the surface that emits and receives the ultrasound pulses. For the high quality manufacturing of the meter bod- ies, a Union CBFK 150 five-axis cutting and drilling machine has been installed in the Silvolde factory (see Fig. 6); a sixth axis will be added to a new expansion table of 2000 mm capable of handling loads up to 12 000 kg. The accuracy of machining can be as high as 0.02 mm; an accuracy which is verified by the NKO, the Dutch calibration authority. The height of the meter is 4.2 m, the length is 5 m and the width is 4 m. In order to reduce the vibrations and stabilize the machine, a foundation of 10 m×10 m×1 m of concrete has been made weighing ca 300 ton. Fig. 7 shows a meter body (B) as installed on the sup- port of numerical controlled machining equipment (A). Initially the meter body is positioned so that the center of the body (1) is aligned with the reference point (2) of the machining equipment, having coordinates (0, 0). 2.3.5. The manufacturing process After the welding of the flanges to the pipe, the nozzles are positioned. This is done using the union cbfk 150 as a high accurate positioning device (see Figs. 8 and 9). After the positioning, the nozzles are welded to the spoolpiece (see Figs. 10 and 11). The next step is the machining of the meter body, starting with the pro- gramming of the coordinates into the machine (see Figs. 12 and 13). The spoolpiece is machined completely without being removed from the machine table. After that the meter is hydro-tested, non-destructively tested (X-ray and ultrasonic) and painted (see Figs. 14 and 15). For the larger meters, 12� and upward the specified dimensional accuracy’s are: 93J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 6. Union CBFK 150. Fig. 7. Set-up for the meter body machining. Fig. 8. Positioning of the transducer nozzle. Fig. 9. Positioning of the transducer nozzle. Fig. 10. Welding of the nozzles. 94 J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 11. Welding of the nozzles. Fig. 12. Programming of the coordinates into the machine. Fig. 13. Programming of the coordinates into the machine. Fig. 14. Being machined after nd-testing. Fig. 15. Being painted after nd-testing. angle ±0.05° X ±0.1 mm Y ±0.1 mm For the uncertainty of the inner diameter a practical value of 0.05% is chosen; a value which can be improved when necessary. Taking a typical example of a 16� meter, the contri- bution of each parameter’s uncertainty can be calculated according to Eq. (5) and results in: path length L ±0.06% 1/cos ϕ (path angle) ±0.15% cross sectional area A ±0.1% When all these factors add up to a worst case scenario, the uncertainty due to the meter body geometry and 95J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 16. Geometry measurement and special tools; larger sizes (42�) measure easier. dimensions would be ±0.3%. However, since each of these contributions is due to independent sources of error, a more appropriate total uncertainty calculated according to the root mean square rule, is 0.2%. 2.4. Total estimated meter uncertainty In the previous sections it has been shown that the significant sources that contribute to the total measure- ment uncertainty of an ultrasonic flow meter are: Flow profile correction factor K ±0.3% Meter body geometry (rms) ±0.2% In a worst case scenario the combination of both sources of uncertainty would result in a total uncertainty of 0.5%. Since these are independent sources of uncertainty it is Fig. 17. Geometry measurement and special tools; larger sizes (42�) measure easier. Fig. 18. Geometry measurement and special tools; larger sizes (42�) measure easier. Fig. 19. Geometry measurement and special tools; larger sizes (42�) measure easier. justified to estimate the total uncertainty using the square root rule to calculate total uncertainty as �0.32+0.22�0.36% This number is of the same order of magnitude as the uncertainty of most of the flow calibration facilities (0.25–0.3%). Dry calibration is the next step in the production. 3. Dry calibration A dry calibration of a flow meter is not the same as a flow calibration. That is, it is not a check of the meter’s result (measured gas volume/reference flow) in compari- 96 J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 20. Dry calibration of a 30� meter. son with a standard or reference device. If only dimen- sions were checked, and electronics and transducers tested for proper operation, it may be more suitable to call this procedure a verification. However, since the meter’s configuration is also adjusted as a result of this verification, the procedure currently employed may more appropriately be called a dry calibration. A dry calibration of an ultrasonic gas flow meter is performed in several major steps of which a brief sum- mary is provided in the following section. 3.1. Dry calibration procedure In the preceding sections it has been shown that geometry and dimension of the meter body and the Reynolds (profile correction) factor are the relevant sources of uncertainty. The meter body geometry and dimensions reflect in the acoustic path geometry and are calculated using the data reported in the protocol provided by the machine Fig. 21. Dry calibration of a 30� meter. Fig. 22. Ratio pattern determined during dry calibration with nitrogen and with wet gas. Fig. 23. Ratio pattern determined before and after the replacement of a ball valve on path 5. shop where the spoolpiece is manufactured. The various steps of a dry calibration must be followed to ensure meter performance and to provide an audit trail of the work performed. This section discusses the essentials of the dry calibration procedure. 3.1.1. Verification of meter geometry The relevant geometry parameters for an ultrasonic gas flow meter must be measured to verify the acoustic path angles and path lengths as presented on the printout from the machine shop. In Figs. 16–19 the geometry verification and the tools are shown. 3.1.2. Electronics and transducer function test After the electronics and ultrasonic transducers are installed on the meter body, a function test is performed. This test insures the meter electronics package, also known as a signal processing unit (SPU), and all trans- ducers are operating properly. The meter is also checked for leakage at this step. It should be noted that by using state-of-the-art elec- tronics and high quality quartz oscillators, time measure- ments can be performed with excellent accuracy and stability (better than 0.01% for both thermal and long- term). Insuring these components are all operating cor- 97J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 24. A 20� Meter calibration by the NMi, at the Gasunie facility in Westerbork, The Netherlands. rectly is part of the dry calibration procedure (Figs. 20 and 21). 3.1.3. Configuration of electronics Once the correct dimensional measurements are obtained, an ultrasonic meter is configured to accommo- date minor variations in fabrication. As discussed earlier, dimensional measurements are very important in determining the path angle and path length. These are the only configuration items that are adjusted for a spe- cific meter. All other configuration items are generic for a given meter size and do not vary from meter to meter. Minor manufacturing variations are taken into con- Fig. 25. Calibration results of 48 meters. Fig. 26. Adjustments of 48 meter calibrations. sideration by adjustments in path length and path angle. Generally the path angle for the ultrasonic signal is 60° relative to the gas flow. If the mechanical dimensions are not accurate to within a few tenths of a millimeter, this angle will change. Measuring the “as built” dimen- sion and computing the actual angles and path lengths improves meter accuracy. 3.1.4. Zero flow check A zero flow check is a test to ensure the meter does not indicate flow when none exists. Due to small differ- ences in transducers and electronics all meters exhibit a small time error, which results in a small offset error in the velocity measurement. To check this, the meter is first fitted with blind flanges and pressurized with nitrogen. The assembly must be located in a thermally stable environment. Due to the high resolution of the ultrasonic meter, even air blowing from an air conditioning system will affect the results. Since there is no gas flow (movement) through 98 J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 27. Underground gas storage in Hungary. the meter (after thermal equilibrium has been achieved), gas velocity observed on all acoustic paths should be zero. To ensure there is thermal stability, the speed of sound is monitored as an indication of temperature change. The speed of sound in gas is very sensitive to temperature. Once the temperature has stabilized, 600 s of meter gas velocity data is collected. The maximum permissible zero flow velocity changes with meter size. Larger meters having a proportionately lower value since they have a longer path length, and the offset due to timing errors becomes less significant as path length increases. It might be noted that this test can also be performed with an ultrasonic gas flow meter installed in a field application, provided the meter can be isolated from the main gas stream. However, utmost care should be taken to avoid misleading results since even the slightest iso- lation valve leakage, or thermal gradients due to sun and wind, can affect the reliability of results. 3.1.5. Speed of sound calculation Next to the flow velocity, the speed of sound is also always measured by an ultrasonic meter and calculated according to the equation: C� L 2� 1tdown� 1tup�F(v), (7) Fig. 28. A combination of ultrasonic meters and turbine meters used as reference standards in a calibration facility. where F(v) is a flow velocity dependent function which eliminates the effects of ray-bending, caused by the gradient in the speed of sound; F(v)=1 at v=0. As the speed of sound calculation is dependent on the path length, any path length error will impact the met- er’s accuracy. As with all devices, small tolerances can contribute to detectable errors that may be difficult to measure directly. This is especially true when trying to determine the exact path length when the ultrasonic signal is being “bounced” twice across the spool piece, as occurs in the Q.Sonic meter. However, if the gas composition, press- ure and temperature are known, the theoretical speed of sound can be computed very accurately (generally better than 0.05%). The computed theoretical speed of sound can then be compared to the meter’s output and used to “adjust” the path length. This is also known as acoustic path length adjustment. This not only reduces measurement uncer- tainty, but it provides also an excellent baseline for field (or laboratory) checks in the future. Currently there are several methods available for com- puting speed of sound in different gas compositions. Instromet uses it’s own program, based on AGA 8 and developed by the Instromet Systems group. Another popular computer program is called SonicWare, a pro- gram that is developed by Lomic, Inc. with the support of GRI. All these programs compute the speed of sound in pure gases such as nitrogen as well as in those of various gas compositions. Whereas this method is excellent for use under test conditions, applying this method on a live pipeline can be problematic when the gas composition is not accu- rately known or when the pressure and temperature are fluctuating. To avoid these problems, Instromet has developed a robust method that is even applicable on drilling plat- forms measuring wet gas and can be used for ultrasonic meters with a minimum of three paths. This method is based on the elimination of the absolute speed of sound by using the ratios of the speed of sound. The advan- tages are: � There is no need for any knowledge about the gas flowing through the pipe. � The measurement can be done even under flowing conditions; only at very high velocities will ray-bend- ing cause a larger uncertainty. � The calculation can automatically be done as part of a diagnostic package. As part of the dry calibration procedure both the absolute as well as the pattern of ratios of the speed of sound measured by the various paths are determined. This pat- tern is used as the basic pattern to which all later patterns have to be compared. 99J.G. Drenthen, G. de Boer / Flow Measurement and Instrumentation 12 (2001) 89–99 Fig. 22 provides results from a five-path ultrasonic flow meter, showing the ratios measured at the dry cali- bration and off-shore with wet gas. In these figures all the different ratios between the speed of sound from the various paths are shown. The ratios are numbered according to the path numbers; 5/1 means the speed of sound from path 5 divided by that of path 1 etc. In Fig. 23 the same method is applied for checking the system after replacing a ball valve, positioned between the transducer and the spoolpiece. The thickness of the ball valve was 0.2 mm smaller than the original valve. Although well within the acceptance limits (the resulting effect is �0.01%), the change of the ball valve is clearly visible and shows the sensitivity of the system of 0.2 mm on an acoustic path length of ca 300 mm. 4. Flow calibration results The calibrations were obtained from the following three different facilities: SwRI in San Antonio, TX; Gasunie in Westerbork, The Netherlands (Fig. 24); and Pigsar in Dorsten, Germany. As a typical example, the results are shown of data obtained from 48 flow calibrations done on behalf of Instromet Inc. in Houston, all plotted together. These results were not linearized in any way. The only meter configuration change was adjustment of the multiplier F (also called the adjust factor). The meter sizes range from 8� to 24� (Fig. 25). Fig. 26 shows the adjustment factor, or F-factor for all meters included in this study. This graph clearly shows, that the reproducibility of all the calibrated met- ers is within ±0.3%; a factor that is in line with the value predicted under Section 2.4, based on the production tol- erances. 5. Installations In Figs. 27 and 28 some examples are shown from actual field installations. 6. Conclusion The combination of high quality fabrication with advanced dry calibration procedures has led to a gener- ation of flow meters whose versatility and accuracy are setting a new standard. Based on the uncertainty analysis in this paper and supported by the results of flow cali- brations due to the tight manufacturing tolerances, the resulting uncertainty is of the same order of magnitude of that of the best test facilities. Acknowledgements The authors would like to express their thanks to John Lansing and Marcel Vermeulen for their valuable contri- butions. References [1] J.O. Hinze, Turbulence, McGraw–Hill, New York, 2nd edition, 1975. [2] H. Dane, R. Wilsack, Upstream pipe wall roughness influence on ultrasonic flow measurement, in: AGA Operations Conference, Cleveland, 1999.