Abstract: Single-channel and multi-channel ultrasonic flow meters were tested at the flow meter research facility of the Gas Research Institute in San Antonio, Texas, establishing benchmark conditions and evaluating their performance under benchmark installation conditions and other pipeline installation conditions. The aim of the tests was to determine the performance of 8-inch single-channel and multi-channel flow meters under different pipeline installation conditions and when used with a flow rectifier. The test results were compared with data under benchmark conditions for in-depth evaluation. The test results for the 8-inch multi-channel flow meter show that the rectifier has the function of improving its measurement accuracy, depending on the type of flow meter and the location relative to the disturbance source. The tests for the single-channel flow meter show that an accuracy better than 0.5% can be obtained under good flow field conditions. The sensitivity of the flow meter to simple disturbances is illustrated by the installation at a 90° bend upstream. The flow measurement error ranged from 1% to 4% relative to the benchmark test installation in these tests. I. Introduction Ultrasonic flow meters derive the gas volumetric flow rate by measuring the propagation time of ultrasonic pulses in the fluid. For an ultrasonic flow meter, determining flow accuracy is a function of the flow meter design and calculation methods, as well as upstream piping requirements, unlike many traditional measurement methods. For example, orifice plate flow meters, measuring flow, require symmetry and no vortex flow within a defined discharge coefficient. To obtain the published performance of flow measurements, AGA Report 3 recommends minimum straight pipe lengths, variations in pipe diameter, and upstream installation requirements for orifice plate flow meters. As the orifice plate flow meter specification (β ratio) increases, the orifice plate's shaping effect on the velocity profile decreases due to reduced flow blockage, but its sensitivity to fluid disturbances also increases. Turbine flow meters also improve the flow profile due to the fluid's action on the rotor. This improved inlet flow profile enhances the performance of turbine flow meters. Because turbine flow meters are sensitive to velocity profile asymmetry, AGA Report 7 requires the use of rectifiers to eliminate inlet flow vortices. Because ultrasonic flow meters have virtually no obstructions affecting population flow and do not alter the flow profile, they accurately determine flow rate based on reliable calculations using flow samples obtained by the flow meter. In the design of multi-channel ultrasonic flow meters, manufacturers attempt to optimize the flow meters to reduce their sensitivity to flow disturbances. If the flow meter can compensate for all flow disturbances, then a rectifier is unnecessary. However, due to the limited published data on how rectifiers affect the performance of multi-channel ultrasonic flow meters, there remains interest in the combined application of rectifiers and ultrasonic flow meters. Based on experience with other types of flow meters, rectifiers represent potential benefits. Using a rectifier in conjunction with an inexpensive mono-channel ultrasonic flow meter is also a viable alternative to a multi-channel flow meter. Furthermore, a limited amount of reliable experimental data confirms the performance of mono-channel flow meters using rectifiers. These results are derived from the first part of the experiments, which sought to determine the practical experience of 8-inch mono-channel and multi-channel ultrasonic flow meters installed in different piping configurations and when used in conjunction with a single rectifier. This test involves installing one or two bends upstream of the test flowmeter and operating the flowmeter with or without a rectifier. Further testing aims to evaluate the bidirectional flowmeter's measurement performance and accuracy at flow rates below 1% of the flowmeter's capacity. In locations where fluid flow can be bidirectional, such as underground gas storage facilities, those planning to install ultrasonic flowmeters are concerned with bidirectional measurement performance. For the installation of a metering station, the number and diameter of flowmeters need to be determined, as low-flow measurement performance should also be considered due to range ratios. II. Test Methods The flowmeter is installed on the test section of the high-pressure loop (HPL), and the test medium is transport-grade natural gas. Data are simultaneously collected from the ultrasonic flowmeter and the critical flow nozzle group on the HPPL, serving as flow standards. Five dual-weighted sonic nozzles are calibrated in-situ relative to the HPPL weighing tank system at different pressures. All calibrations are performed using an online gas chromatograph and AGA Report 8 equation of state to determine gas properties. The static pressure, related to the HPPL reference pressure, is measured at twice the pipe diameter downstream of one flowmeter. Gas temperature is measured three times the pipe diameter downstream of each flowmeter. The measured temperature and pressure, along with the measured gas composition and the gas volume measured by the ultrasonic flowmeter, are used to calculate the mass flow rate of the ultrasonic flowmeter. This mass flow rate is then compared with the flow rate determined by the critical flow nozzle. Depending on the manufacturer's chosen approach, the ultrasonic flowmeter can obtain the volumetric flow rate using different methods. The calibration methods within flowmeters M3 and M4 are used to calculate the total gas volume and the time of operation. The average flow rate is then calculated based on the total volume. Flowmeters M1 and M2 report (measure) the actual flow rate, providing flow rate values ​​once per second to determine the average volumetric flow rate. The status, velocity, and sound velocity data of individual channels are also recorded. A typical test system consists of recirculating gas flowing through a flow loop until the gas temperature and pressure stabilize. Different combinations of sonic nozzles are selected and switched to determine a stable flow rate. A test point consists of the average value calculated over a 90-second period of flow rate and other measurements. A test point is repeated six times to calculate the average value and standard deviation. Measurement data are also acquired by two turbine flowmeters. The turbine flowmeter data confirms the consistency of the experiment. The four flow meters used in this test were supplied by the manufacturer and are all commercially available. No flow calibration was performed prior to this test. Two of the four flow meters are multi-channel, and two are mono-channel. Table 1 shows the flow meter channel arrangement. Table 1: Test Flow Meter Geometry. All flow meters were manufactured with uniform flange-to-flange dimensions (31.5 in length) and inner diameters (within 0.005 in) to facilitate interchangeability between flow meters in different installation locations. The manufacturer provided flow meter setting parameters according to a dedicated program for mechanical, electronic, and other measurements. During testing, the profile correction parameters for the flow meters were checked by the manufacturer. Specific parameters related to the test conditions (fluid properties for the internal electronics of the flow meters) were adjusted as needed each time. III. Basic Test Installation The basic test piping installation is shown in Figure 1. All piping was made of 40# carbon steel pipe with an 8-inch inner diameter (7.981-inch inner diameter), and the internal welds were polished. The flow meters under test were installed downstream of the 90° long-diameter elbow at points 40D (D = 8in), 59D, and 97D. A 12in × 16in × 10in Sprengle jet rectifier was installed upstream of the elbow, followed by a 10in × 8in concentric reducer and a 43D 8in straight pipe section extending to the elbow. Each flow meter could be tested in two of the four axial positions (see Table 2). Table 2: Test Positions for Flow Meters under Baseline Conditions. The complete test plan for each flow meter requires testing in all four positions. When a flow meter is moved from one axial installation position to another, the upstream and downstream straight pipe sections (10D and 5D, respectively) should be disassembled as a whole with the flow meter to align the flanges connecting the upstream and downstream of the flow meter. The pressure and temperature transmitters associated with each flow meter should also be retained on the original straight pipe section when the flow meter position is changed to reduce additional errors. IV. Basic Test Results The test performance of the two multi-channel flow meters (M1 and M3) under the baseline installation conditions is shown in Figures 2 and 3. Results are expressed as percentage error (relative to the MRFHPL critical flow nozzle), as a function of the average flow velocity through the flow meter. A single point represents the average of six repeated measurements at each flow rate, and the error band represents a 95% confidence level. For flow meter M1, at velocities greater than 10 ft/s, all data fall within the 0.4% range, independent of velocity (see Figure 2). At velocities below 10 ft/s, the error curve deflects upwards, which may be due to a positive zero bias or a deviation in the correction algorithm. This cannot be entirely attributed to differences in the velocity profile at low flow rates; it may be a combination of both effects. Checking the flow meter at zero flow indicated a zero drift of 0.01 ft/s. If zero drift is removed, the upward deflection will be flattened, with a drift of 0.35% at the minimum velocity point (2.8 ft/s) and 0.18% at 5.6 ft/s. However, this indicates that the upward deflection of the curve cannot be considered entirely due to zero drift. Examining the flow calibration curve shows an approximately 0.2% difference in error between measurements taken when the flowmeter is installed at 97D and at 59D. The error at 97D is larger (more negative) than the error at 59D. Similar measurement errors exist in the 400 lb/in²(A) and 900 lb/in²(A) test cases. The measurement error of flowmeter M3 is related to the velocity flowing through it, as shown in Figure 3. The non-linear characteristics of this flowmeter's measurement error differ from those observed in similar experiments with 12-inch flowmeters previously conducted at MRF (Grimly) and elsewhere (completed by Van Bloemendaal and Van der Kam). Only when the flow meter was installed at 59D, and data was collected at a pressure of 400 lb/in²(A), did the error fall within 0.3% when the average flow velocity was still above 10 ft/s. The error curve sloped downwards as the velocity increased. In the 59D, 400 lb/in²(A) data set, individual channel status information indicated that this was because the measured propagation time failed the internal consistency check. This may be the reason for the deviation from other data sets. Data, including the questionable data set, remained within 0.5% error at velocities above 10 ft/s. These data also indicated a 0.2% difference in flow measurement when the pressure increased from 400 lb/in²(A) to 900 lb/in²(A). When comparing the flow meter at 59D and 97D, the difference in current flow results at 900 lb/in²(A) was 0.1%. A possible explanation for this difference is that the velocity profile continued to develop after 59D. By comparing the velocity ratios of individual channels, an improvement in the velocity profile shape can be observed. Table 3 shows the ratio of the center channel velocity (ultrasonic channel at the pipeline axis centerline) to the outer channel velocity of flowmeter M1 at the maximum flow point. Similar calculated values ​​for flowmeter M3 are shown in Table 4. Since each flowmeter has a unique channel location, comparisons should not be made between different flowmeter types. However, consistent differences exist between the test results of the two flowmeters at 59D and 97D. The smaller ratio in the 59D test data indicates that the velocity profile at that location has a smaller curvature than that at 97D. Differences in flowmeter responsiveness also result in a profile sensitivity. The purpose of measuring the velocity profile is to better characterize the velocity profile under baseline installation conditions. Table 3 shows the velocity from center to outside of M1. Table 4 shows the velocity from center channel to outside channel of M3. The velocity of the single-channel flowmeter M2 (see Figure 4) is approximately -1.2% offset when installed at 78D and approximately -1.8% offset when installed at 40D. These results indicate that the error in flow measurement is independent of pressure but related to axial position. The test error curves for the single-channel flowmeter M4 at two axial positions and two operating pressures are shown in Figure 5. The curves show an average deviation of 0.5%, indicating that the online pressure has a greater impact on the measurement results than the distance between the 90° bend and the flowmeter. The error curves show large deviations and inconsistencies with other results at velocities below 10 ft/s, at 400 lb/in²(A), and at 78D. Generally, the measurement data from a single-channel flowmeter are more dispersed than those from a multi-channel flowmeter. This dispersion is not surprising, as this type of flowmeter does not have the advantage of averaged airflow found in multi-channel ultrasonic flowmeters. This characteristic of mono-channel flow meters is particularly pronounced at flow rates below 10 ft/s. V. Installation Impact on Testing The installation process for the impact tests implemented so far is shown in Figure 6. Mono-channel flow meters were tested at two different locations 1 and 2 (downstream of a single 90° long-diameter L-bend at locations 10D and 19D). Multi-channel flow meters were tested downstream of two planar bends at locations 3 and 4 (downstream of a second L-bend at locations 10D and 19D). The test pipeline was bare pipe, with two different flow straighteners installed upstream of locations 3 and 4. The first flow meter was located 5D downstream of the straightener outlet, which meets the minimum requirements for turbine flow meter installation in AGA Report 7. Additionally, in the testing of the multi-channel flow meters, the two planar bends were arranged horizontally with respect to the pipe. Because this arrangement represents an installation form where the bends provide a vertical offset, the flow meter had to be rotated 90° to maintain the correct relative orientation of its measuring channel to flow interference. VI. Impact of Installation on Test Results These results are reported as deviations from test results under baseline conditions. For multi-channel flow meters M1 and M3, data obtained with the flow meters installed 97D downstream of a single 90° bend represent test results under baseline conditions. For single-channel flow meters M2 and M4, the results obtained with the flow meters installed 78D downstream of a 90° bend are used as test results under baseline conditions. Test results under baseline conditions represent the closest possible result from an optimized process installation. The results of testing multi-channel flow meter M1 10D downstream of two planar bends are shown in Figure 7. The test results show that the measurement error introduced by bare pipe installation is within 0.5% of the basic measurement result, with a relative error on the order of 0.3% to 0.5%. The relative error occurs at high gas velocities. There is no significant difference in test results with the flow meters installed 10D downstream of the second bend, equipped with a 19-tube bundle rectifier and a GFC-type rectifier. At velocities above 20 ft/s, both rectifiers reduce the relative error by at least 0.25%. For velocities below 20 ft/s, all results tend to converge, indicating that the flowmeter is either sensitive to the velocity profile resulting from the installation arrangement or that the velocity profile effect is not significantly different within this velocity range. The determination of the planned velocity profile will help interpret these results and will further interpret the data already collected related to single-channel velocity ratios. The test results for flowmeter M1 installed 19D downstream of two planar bends are shown in Figure 8. The relative measurement errors for both bare pipe and 19-tube bundle rectifier are approximately 0.5%. The error in bare pipe measurements when the flowmeter is installed 19D downstream of the bend is slightly larger than the error when the flowmeter is installed at 10D. Above velocities above 10 ft/s, the measurement results installed with GFC are within 0.1% compared to the test results under baseline conditions. Furthermore, the measurement results converge at low flow rates. The test results for multi-channel flowmeter M3 installed downstream of the double bend (see Figure 9) show that for bare pipe, when the velocity is greater than 20 ft/s, the error of M3 is less than 0.25%. The results for the 19-tube bundle rectifier and GFC rectifier, including measurements of the bare tube, show a difference of -0.30% to 25%. The variation in relative error with flow velocity is caused by the radians of the test data under the baseline conditions at 97D, 400 lb/in²(A) (see Figure 3). The results under the bare tube mounting conditions, tested by flowmeter M3 installed at 19D, are shown in Figure 10, deviating by approximately 0.2% compared to the flowmeter data at 10D, and remaining within 0.25% of the baseline test results above 20 ft/s. Except for the highest flow velocities, the data for the 19-tube bundle rectifier are nearly consistent. For flow velocities above 45 ft/s, the measurement error generated by the GFC is closer to the error under the baseline mounting conditions than that of the 19-tube bundle rectifier and the bare tube. The test results from Figures 7 to 10 indicate certain benefits of using a rectifier for multichannel ultrasonic flowmeters, depending on the type and location of the flowmeter. In some of the cases described above, the test error deviation relative to the reference conditions can cause such absolute errors, which are actually closer to zero than the results under the reference conditions. It is unclear whether this result is due to unforeseen circumstances or to a less-than-ideal velocity profile in the flowmeter study. The test results for a single-channel flowmeter installed at 10D and 19D downstream of a 90° long-diameter bend are shown in Figure 11. The measurements indicate that the error deviation relative to the reference condition test error installed at 78D downstream of the 90° bend is approximately 2% to 2.2%. An additional error deviation of 1.3% is generated by rotating the flowmeter 90° about its axis. Because in this example, the single-channel flowmeter is installed in its normal orientation, with its channel at 30° to the vertical, this deviation is likely related to the flowmeter's installation orientation. In this installation, the asymmetric main component of the flow is in the vertical plane, and the channel orientation changes relative to flow disturbances when the flowmeter rotates 90°. The flowmeter M4 installed at 19D downstream of the bend, except at the 90° position, shows an deviation of approximately 1.5% to 2% relative to the reference condition, as shown in Figure 12. The test results for this flowmeter are essentially independent of its location, likely due to the 45° angle between the measuring channel and the vertical. In this case, the channel position remains consistent relative to flow disturbances even when the flowmeter's shaft center deflects. VII. Conclusion The tests under baseline conditions show that the fluid velocity profile at 59D downstream of a 90° long-diameter bend is not fully developed, and the tested flowmeter is sensitive to the subsequently developing velocity profile. This indicates that the accuracy of the tested 8-inch flowmeter can be improved through flow calibration. The test results for the 8-inch multi-channel flowmeter demonstrate the potential benefit of using a rectifier for improved measurement accuracy. The single-channel flowmeter tests show its potential to achieve measurement accuracy better than 0.5% under favorable flow conditions. Tests using flowmeters installed 10D and 19D downstream of a single 90° bend illustrate the flowmeter's sensitivity to simple turbulence. The flow measurement error ranges from 1% to 4% relative to the baseline conditions. The experimental test results are only a preliminary part of a large test plan. Further in-depth analysis of the obtained data and other test data from this test plan will yield new insights into the performance of ultrasonic flow meters in high-pressure natural gas applications.