The Evaluation of Dry Gas Meters in Wet Gas Conditions A Report for

The Evaluation of Dry Gas Meters in Wet Gas Conditions A Report for National Measurement System Directorate Department of Trade & Industry 151 Buckingham Palace Road London, SW1W 9SS Report No: 2002/100 Date: November 2002

The work described in this report was carried out under contract to the Department of Trade & Industry ( the Department ) as part of the National Measurement System s 1999-2002 Flow Programme. The Department has a free licence to copy, circulate and use the contents of this report within any United Kingdom Government Department, and to issue or copy the contents of the report to a supplier or potential supplier to the United Kingdom Government for a contract for the services of the Crown. For all other use, the prior written consent of TÜV NEL Ltd shall be obtained before reproducing all or any part of this report. Applications for permission to publish should be made to: Contracts Manager TÜV NEL Ltd Scottish Enterprise Technology Park East Kilbride G75 0QU E-mail: jduff@nel.uk Tel: 01355-272096 TÜV NEL Ltd 2002

Flow Centre National Engineering Laboratory East Kilbride Glasgow G75 OQU Tel: 01355 220222 Fax: 01355 272999 The Evaluation of Dry Gas Meters in Wet Gas Conditions A Report for National Measurement System Directorate Department of Trade & Industry 151 Buckingham Palace Road, London SW1W 9SS Prepared by: Dr D G Stewart.. Approved by: Mr D Boam.. Date: November 2002 for Michael Valente Managing Director Report No: 2002/100 Page 1 of 82

SUMMARY Wet gas flow measurement has become increasingly important in the oil and gas industry in recent years. It is often required in the development of small or remote gas fields where several fields are processed together in common platform facilities. In such cases the individual unprocessed streams must be metered before mixing. In some circumstances gas meters may also be subjected to small amounts of liquid in the gas. This can happen on the gas output of a separator as a result of unexpected well conditions or liquid slugging. One approach to wet gas metering is to use a standard dry gas meter and apply corrections to the measurements based on knowledge of how his type of meter is affected by the presence of liquid in the gas stream. This method requires prior knowledge of the liquid flow, which has to be obtained by another means. This report describes tests conducted on under the 1999 2002 DTI Flow Programme Project MU07. In this project five types of test meter were evaluated in wet gas conditions turbine, vortex, Venturi, V-Cone and Coriolis. Of these, the vortex, Venturi and V-Cone would be used intentionally in wet gas metering, whereas the turbine and Coriolis would not be designed into a wet gas metering system but may experience occasional liquid in the gas stream. The tests were conducted on NEL s high-pressure wet gas test facility at pressures from 15 bar to 60 bar across a range of gas velocities and liquid fractions. The results from each test are presented in this report and, where available, compared with previous data in the literature. The turbine meter was tested with only small liquid fractions to avoid any damage that would be expected at higher fractions. The maximum error induced by the liquid was 0.75% at a liquid mass fraction of 2%. This error occurred at a low flowrate and the error varied significantly with flowrate, with little error at the highest flowrates. The results from the vortex meter tests show that this meter was significantly affected by the liquid presence. The errors range from 0% to 30%, with the error generally increasing with increasing liquid fraction. The meter error reduced with increasing pressure and, for liquid fractions less than 1.0%, was relatively unaffected by the superficial gas velocity. The most extensive set of tests carried out in this project were those on three Venturi meters. Three 4-inch meters were chosen with different beta values and tested over a range of pressures, gas velocities and liquid fractions. The results show that the liquid causes the meter to overread the gas rate. This overreading is affected by the liquid fraction, gas velocity, pressure and Venturi beta value. The new results are compared with existing correlations for Venturi overreading. The V-Cone meter is a differential pressure meter, like the Venturi, and behaves in a similar manner qualitatively, although the actual overreadings are slightly different. The results of tests on two V-Cones are presented here, again tested across a range of pressures, gas velocities and liquid fractions. In general the V-Cone overreads slightly less than the Venturi. The Coriolis meter was tested at 30 bar and 60 bar, at liquid volume fractions up to 5%. The Coriolis meter was significantly affected by the presence of liquid even at the lowest level tested, 0.1% LVF. The errors do not show predictable trends with respect to gas velocity and liquid fraction. Report No: 2002/100 Page 2 of 82

The results presented in this report will help engineers to make better-informed decisions as to the suitability of these meter types. In some cases, turbine and Coriolis, the results are not suitable for deriving correction factors but may be used to estimate limits of operability. The Venturi, V-Cone and vortex meter results may be used to correct the measurements from such meters when used in wet gas. Such corrections should be applied with care, as use of any correction factor outside the range of the underlying data is not recommended. Report No: 2002/100 Page 3 of 82

CONTENTS Page 1 INTRODUCTION 5 2 TEST PROGRAMME 6 2.1 Turbine Meter. 6 2.2 Vortex Meter.. 7 2.3 Venturi Meter. 8 2.4 V-Cone Meter 9 2.5 Coriolis Meter 10 3 NEL WET GAS FACILITY.. 11 4 TEST PROGRAMME. 12 4.1 Turbine Meter.. 12 5 VORTEX METER 19 5.1 Test Programme. 19 5.2 Results and Discussions.. 20 5.3 Comparison with Previous Work 23 5.4 Conclusions.. 31 6 VENTURI METERS 33 6.1 Test Programme. 36 6.2 Calculation of Over-reading.. 36 6.3 Results and Discussion. 37 6.4 Comparison with Previous Work. 39 6.5 Development of an Improved Wet Gas Venturi Correlation 42 6.6 Conclusions. 43 7 V-CONES. 59 7.1 Test Programme. 59 7.2 Calculation of Over-reading.. 60 7.3 Results and Discussion. 60 7.4 Comparison with Previous Work.. 62 7.5 Conclusions. 62 8 CORIOLIS METER 70 8.1 Test Programme 70 8.2 Results and Discussions.. 71 8.3 Conclusions 81 REFERENCES.. 82 Report No: 2002/100 Page 4 of 82

1 INTRODUCTION Multiphase metering has become increasingly important in the oil and gas industry in recent years, particularly in the offshore sector. In many cases the development of small or remote oil and gas fields can only be made economically viable if they can be tied back to existing platform infrastructure, reducing the initial capital expenditure required by significant margins. In such cases several fields are often tied back into common production facilities, requiring each unprocessed stream to be metered prior to co-mingling. Multiphase metering can also be a valuable technology in well management, where it can provide on-line information on the production flows. In such cases the absolute accuracy of the multiphase meter is less important than for allocation purposes, as the meter is used more for detecting changes in flow conditions. Another key area for multiphase metering is well testing, again due to the reduced capital expenditure required for testing new wells. In such cases a multiphase metering skid can offer greater flexibility and improved turnaround times when compared with traditional approaches. Wet gas flow is essentially the high gas fraction end of multiphase flow, typically with a gas volume fraction (GVF) above 90%, and mainly above 95%. Most multiphase meters do not perform satisfactorily in such conditions, and indeed many manufacturers will not quote an expected accuracy in this range. The development of accurate and reliable wet gas metering technology is therefore a key requirement of the offshore oil and gas industry. The ultimate goal is the development of wet gas meters or metering systems that can measure both the liquid and gas phases (and also water cut in the liquid component) in real time. There are two principal approaches to wet gas metering. The first is to use a dedicated wet gas meter which has been designed to measure the flow rates of both the liquid and gas phases. The second is to use some standard dry gas meter and applying corrections to the measurements based on knowledge of how the meter in question is affected by the presence of a liquid phase in the gas stream. The second method requires prior knowledge of the liquid flow to be able to correct for the gas flow. Obviously the first method is desirable from the point of view of continuous measurement and well management. However, if the liquid flow is known to remain reasonably constant, or change slowly then the second method can be used, subject to a suitable means of determining the liquid flow by another method. This report presents the results of wet gas tests on several different types of commercially available dry gas meters. This work increases the available knowledge on how dry gas meters are affected by the presence of liquid in the gas stream allowing the second method above to be employed with greater confidence. This work is also of fundamental importance to the development of any true wet gas meter based on any of the metering types tested here. The results provide more information on the performance of such meter types and should allow more informed development of wet gas meters based around one or more of these meter types. Report No: 2002/100 Page 5 of 82

2 TEST PROGRAMME The meters tested in this work were selected to represent the likely scenarios encountered in oil and gas production situations. The following meter types were tested: Turbine Vortex Venturi V-Cone Coriolis The following sub-sections give a brief introduction and background to each meter type, operating principle, potential wet gas applications, and any previous test work in wet gas flows. More detail on all meters is given in the relevant section of this report. 2.1 Turbine Meter The turbine meter is a volumetric flow device, operating on the principle that the momentum of the flowing fluid drives a centrally mounted turbine rotor. The rotational speed of the rotor is proportional to the volumetric flow. The turbine meter would seem an unlikely choice to place in a wet gas environment, due to the potential damage to the blades caused by liquid impacting on them. However, gas turbine meters may experience gas flows with some liquid entrained during process upsets or on the gas output line from separators, where total separation is not often achieved. Consequently, this meter was selected for testing at relatively low liquid fractions. Fig. 1: A turbine meter for gas Report No: 2002/100 Page 6 of 82

2.2 Vortex Meter The vortex meter is principally a velocity measuring device. The meter operates on the principle of vortex shedding from a bluff body. In practice, this involves placing a square or triangular shaped block in the path of the flow as shown in Fig. 2. As the fluid flows past this body, vortices are generated alternately at each side of the obstruction at a frequency proportional to the fluid velocity (generally a linear relationship). Detectors placed on the pipe wall measure the frequency of the vortex shedding and determine the velocity (or volumetric flowrate) from the frequency by applying a suitable meter factor. Vortex meters are commonly used in process gas and steam flow measurement and could potentially be subjected to the same adverse conditions described above for the turbine meter, particularly in respect to the condensation of water in a steam line when insulation is less than perfect. A typical quoted uncertainty for an uncalibrated meter is of the order of 1.5% of reading, with calibrated meters providing uncertainties of less than 1%. Vortex meters also have turndown ratios of the order of 20:1, similar to turbine meters. Fig. 2: The principle of operation of a vortex meter The vortex meter is typically not a type of meter that would intentionally be installed in a wet gas flow line, however line conditions may exist where an unexpected liquid presence can occur. In this case a lack of knowledge of the behaviour of the meter under those conditions could produce a significant error in the meter reading. However, one known commercial wet gas meter currently in production, manufactured by Agar Corporation, uses a vortex meter in conjunction with two Venturi meters to measure the gas and liquid flowrates. This meter is to be tested as part of an ongoing wet gas Joint Industry Project (JIP) at NEL. There are only three papers in the literature that provide wet gas test data for a vortex meter. These were published in the late 1980s and early 1990s and were studies of multiple types of meter in wet gas conditions. The first two papers were published by Nederveen, Washington and Batstra [1] +, and Washington [2] in 1989 and dealt with tests on both Venturi and vortex meters in natural gas/water mixtures at low and high pressure gas. The third paper was published by Hussein and Owen in 1991 [3] who looked at the effect of both superheated and wet steam at relatively low pressure on three meter types, a variable area meter, a vortex meter and an orifice plate meter. The vortex meter results presented in the above mentioned papers are discussed in detail and compared with the wet gas results obtained at NEL in Section 5. + Numbers in parenthesis [ ] denotes references given at the end of this report Report No: 2002/100 Page 7 of 82

2.3 Venturi Meter The Venturi meter is a differential pressure meter where the flow is accelerated through a restriction (throat), thereby producing a measurable pressure drop. The flowrate is proportional to this pressure drop and is given by the following equation: m C d E At 2 p (1) where E is the velocity of approach factor 4 1/ 1, is the gas expansibility factor, A t is the throat area, is the gas density and p is the measured differential pressure. Pipe ID (D) Direction of flow Fig. 3: Schematic of standard Venturi meter Venturi meters have become increasingly popular in the measurement of wet gas flows, particularly for allocation purposes and well management. Venturi meters are less susceptible to damage from liquid slugs than orifice plates and due to the convergent inlet section, the hold up of liquid is less pronounced than in an orifice plate. When Venturi meters, or any other differential pressure meters, are used in a wet gas flow the measured differential pressure is higher than it would be if the gas phase were flowing alone. It is believed that this is caused by energy losses at the gas liquid interface(s) as the gas drives the liquid along the pipe. The exact amount of additional pressure loss will depend on several parameters, including the amount of liquid present, the pressure, gas velocity, liquid density, viscosity and surface tension, and the flow regime in the pipe (stratified or annular-mist). This additional pressure drop produces an over-reading in the gas mass flowrate, compared with what would have been measured without any liquid present. This difference must be corrected for by using some form of over-reading correlation. Several correlations have been proposed over the years, e.g. Murdock [4], Chisholm [5,6], Lin [7], de Leeuw [8], and Steven [9]. These correlations and wet gas Venturi calculations are dealt with in more detail in Section 6. Report No: 2002/100 Page 8 of 82

At present the only way to ensure an accurate correction for wet gas operation is to test a meter prior to installation in a wet gas test facility. Whilst de Leeuw and Steven both published improved correction factors based on their own data, it is clear that further work is required in this area to provide a larger, more reliable data set on which to base further improved correction factors. This report presents results from wet gas Venturi tests at NEL, which expand the available data set considerably. 2.4 V-Cone Meter The V-Cone meter is a proprietary differential pressure flowmeter manufactured by M c Crometer. The V-Cone features a cone located in the center of the pipe, with the flow passing round the outside of the cone. The differential pressure is measured between the upstream pipe tapping and a tapping on the downstream end of the cone as shown in Fig. 4. Fig. 4: V-Cone differential flowmeter The flowrate through a V-Cone is calculated in the same manner as that for an orifice plate or Venturi, although the discharge coefficient and expansibility factor are different for both of these meters. M c Crometer performed a limited amount of wet gas testing on the V-Cone several years ago [10]. These tests only featured low liquid fractions and relatively low gas flowrates. From these tests it was concluded that the V-Cone over-reads in wet gas in a similar manner to the orifice plate and Venturi meter, although not by the same amount. One possible advantage over the orifice plate or Venturi is that the cone should not hold up the liquid in the same way as the orifice plate or, to a lesser extent, the Venturi convergent section. Report No: 2002/100 Page 9 of 82

2.5 Coriolis Meter The Coriolis meter is a true mass flow meter, operating on the principle of the Coriolis force. In a Coriolis meter this works by measuring the effect that the flowing fluid has on the vibration of one or two tubes through which the fluid flows. The Coriolis meter is also an unlikely choice for wet gas metering, as there is little information on the performance of such meters in wet gas situations, and furthermore there are so many different designs making transfer of experiences from one meter to another difficult. However, as Coriolis meters are becoming increasingly accepted for gas measurement, it may encounter the same conditions as the turbine or vortex meter. Because the Coriolis meter is a true mass meter there is particular interest in how it will be affected by the presence of liquid. Fig. 5: A Coriolis mass flow meter Report No: 2002/100 Page 10 of 82

3 NEL WET GAS FACILITY The NEL Wet Gas Test Facility has been operational since 1999, and has been heavily used for research, testing and calibration work since. The facility, which was developed under the 1996-1999 DTI Flow Programme, is a 6-inch (NB) recirculating loop based around a 12 m 3 gas/liquid separator. Although nominally 6 inches in diameter, the test section can accommodate line sizes ranging from 2 inches through to 8 inches. The fluids used are oxygen-free nitrogen (density range 2 to 70 kg/m 3 ) and a kerosene substitute (Exxsol D80; approximate density of 800 kg/m 3 ). The facility generally operates at an ambient temperature of 18 C (range 15 to 20 o C) at pressures up to 63 barg. A schematic of the facility is shown in Fig. 6. Fig. 6: Schematic diagram of the high pressure wet gas test facility at NEL Gas is drawn from the top of the separator and driven round the loop by a 200 kw gas blower up to a maximum (pressure independent) actual dry gas flowrate of 1400 m 3 /hr. Liquid is injected through an injection spool over 60D upstream of the test section. The gas and liquid temperatures are both controlled with heat exchangers to maintain equal temperatures in the test section. The gas and liquid reference flowrates are measured using traceable, calibrated, reference turbines. The expanded uncertainty is 0.35% for the gas mass flowrate and 0.15% for the liquid mass flowrate (both at the 95% confidence level, k=2). All temperature and pressure measurements are taken using traceable calibrated instrumentation. A modified subsea video camera can be used to monitor the two-phase flow in the test section. The camera view allows the transition from stratified flow to annularmist flow to be observed. The test conditions were different for each type of meter, due to the differences between the meter types and sizes. The actual test conditions used are given within the particular sections describing the results for each meter type. Report No: 2002/100 Page 11 of 82

4 TEST PROGRAMME The test conditions were different for each type of meter, due to the differences between the meter types and sizes. The actual test conditions used and the results for each meter type are described within the following sections. 4.1 Turbine Meter 4.1.1 Application Turbine meters are normally used for custody transfer measurement of processed natural gas flows. This necessitates the removal of any component likely to be present in the liquid state or likely to drop out of the gas stream prior to metering. The damaging effect of a liquid presence on turbine meters is well known and usually avoided. However, there may be situations where gas is metered that has the potential for liquid formation if it is not processed sufficiently, or if unanticipated pressure drop occurs. The presence of even a small quantity of liquid is likely to have a measurable effect on what is a sensitive and finely balanced piece of equipment. To examine the effect of a liquid-containing gas stream on a typical design of turbine meter would therefore allow a measured response were it found in the field that a small (and measurable) quantity of liquid was passing through an assumed dry gas metering station. Very little literature is available on the effect of liquid on turbine meters, particularly at high pressures. The only significant reference found on this specific topic was published by Jones and Ting [11] in 1996. In fact, the authors state in their paper that Current turbine meter research has been confined to the areas of installation, pulsation, and pressure effects on meter performance, ie no wet gas data available. The paper of Jones and Ting is therefore the main reference for this work, which is an attempt to expand the available database and to provide confirmatory or contradictory evidence compared with that in [11]. Jones and Ting described a series of wet gas tests on both a 6-inch turbine meter and an orifice plate system at CEESI in Colorado. In their work the authors describe how small quantities of water were injected into an air stream at approximately 51.6 bar, over a Reynolds number range of 2 to 10 x 10 6, at 22D upstream of the turbine meter. The shift in the meter factor was measured at equivalent pipe Reynolds numbers to a dry gas baseline test that was performed initially. The authors found that the liquid presence (liquid mass fractions of 0.05% and 0.1%) shifted the behaviour of the meter at Reynolds number above 5 million, by a mean value of approximately -0.3%. As will be discussed later, this published behaviour differs markedly with that obtained at NEL. 4.1.2 NEL test programme A 6-inch Instromet gas turbine meter (model: G-650, type: SM-RI-XE and serial no: 61648) was initially calibrated in dry gas at 20 barg, 40 barg, and 60 barg. This gave a baseline calibration curve across a range of Reynolds from 1 x 10 6 to 9 10 6 against which the wet gas tests could be compared. The meter was subsequently tested in wet gas at 60 barg. Four sets of tests were carried out across the gas flow range 200 to 1000 m 3 /hr. In the first set of tests, the liquid flowrate was held constant at 0.045 l/s, ie the minimum liquid rate that could be injected in the NEL wet gas facility. This flowrate corresponds to a liquid mass fraction (LMF) of 0.93% at the lowest gas flowrate and 0.18% at the highest gas flowrate. The next three tests were carried out with liquid mass fractions of 0.5%, 1.0% and 2.0%. A comparison of the test conditions with those at CEESI is given in Fig. 7, which is a reproduction of a Shell flow pattern map Report No: 2002/100 Page 12 of 82

based on the gas and liquid densiometric Froude numbers (Fr), expressed as equations 2 and 3 below. Fig. 7: Shell flow pattern map with NEL and CEESI test points Fr g A D q g gd l g g (2) Fr l A D q l gd l l g (3) where q is the volumetric flowrate in m 3 /s, A D is the pipe area, D is the pipe internal diameter, g is acceleration due to gravity and is the density. g and l are gas and liquid identifiers respectively. 4.1.3 Results and discussion Dry gas baseline The dry gas baseline calibration is shown below in Fig. 8 below. The calibration was carried out across a Reynolds number range of 1 x 10 6 to 9 10 6. The spread of the data around the best-fit curve is within 0.1%. This calibration was subsequently used for comparison with the wet gas tests. Report No: 2002/100 Page 13 of 82

Fig. 8: Dry Gas Baseline Calibration Wet gas tests As stated in Section 4.1, four sets of wet gas tests were carried out at 60 barg, across a gas flow range of 200 m 3 /hr to 1000 m 3 /hr with a constant liquid rate of 0.045 l/s, at liquid mass fractions (LMFs) of 0.5%, 1.0% and 2.0%. It had been hoped that the LMFs used in these tests would allow a direct comparison with the published results [11], where a typical shift in K-factor of -0.3% relative to the dry gas baseline was obtained. However, the LMFs used at NEL are generally an order of magnitude larger than those used by Jones and Ting, which were stated as 0.05% and 0.1%. However, it was anticipated that both a similar type and size of shift would be obtained at NEL. To keep these liquid contents in perspective, the generally accepted definition of wet gas allows LMFs of 50% or more, considerably larger than were used here. These were not used for two reasons, namely that there was no particular relevance to actual field operating conditions, as a turbine meter is unlikely to be used intentionally under such conditions and also that it was felt that damage to the meter would have been certain to occur. The results from the wet gas tests are shown in Figs 9 to 12. In the following discussion no attempt is made to interpret the results obtained from the point of view of the flow behaviour within the turbine meter itself, as this would be somewhat speculative in nature. Acceleration of the gas within the meter (due to the presence of the annular space directly upstream of the rotor) will likely cause a local shift in the flow pattern. The area available for flow in a typical meter is less than half of the cross sectional area in the upstream pipe. As a consequence the gas Froude number values presented in Fig. 7 would increase by more than a factor of 2, shifting the flow behaviour to more of an annular-mist type. It is not clear however that sufficient time would be available within the upstream section of the meter to fully accelerate any liquid droplets present to the gas velocity at the rotor. The gas acceleration would also cause a partial thinning of the film at the pipe wall combined with an Report No: 2002/100 Page 14 of 82

increase in liquid entrainment. It is uncertain as to what effect, if any, the entry vanes of the meter would have on the liquid passing into the meter. Fig. 9 Wet gas tests with a constant liquid flow of 0.045 l/s The data shown in Figs 9 to 12 can be divided into approximately three regions according to the type of wet gas flow, particularly for the results shown in Figs 11 and 12. Region 1 is actually only a single test point, at the minimum Reynolds number value of approximately 1.8 x 10 6, at which the flow is completely stratified. The maximum measured shift is 0.38% at an LFM of 2%, while at lower LMFs the shift is closer to -0.15%. The main influence on the turbine meter in this flow regime is the drag effect of the blades having to pass through a small layer of liquid running along the base of the pipe. This will have a decelerating effect on the turbine rotor, and will decrease the measured frequency compared with the dry gas flow. Region 2 covers the Reynolds number range 2.7 10 6 to approximately 5.4 10 6. Over this range the stratified flow becomes transitional (towards annular mist flow) in nature with jets (ie local concentrations of liquid flow) of liquid being partially suspended in the gas flow, plus a shift in the concentration of the annular-mist nature of the flow. These flow patterns were observed directly within the pipeline using the subsea camera fitted to a specially modified pipe spool. The trend in the data indicates that the shift away from the baseline data increases quite significantly with increasing LMF. The maximum shift measured was of the order -0.75% at a Reynolds number of 2.7 10 6. Report No: 2002/100 Page 15 of 82

Fig. 10: Wet gas tests with a constant LMF of 0.5% Fig. 11: Wet gas tests with a constant LMF of 1.0% Report No: 2002/100 Page 16 of 82

As the Reynolds number increases above 5.4 10 6, the flow pattern gradually becomes dominated by the annular-mist regime, and all visible traces of liquid jets in the flow disappear. It is suspected that the liquid jets have a more significant effect on the deceleration of the turbine rotor than stratified flow due to the fact that they occur in a random manner over a wider cross-section of the pipe, and interact with a larger blade surface area than stratified flow. Another effect of the entrained liquid is the increased random scatter observed in the meter factor. This is most apparent in Figs 11 and 12 where the maximum size of the band is around 0.15% (Re = 3.6 10 6 and 4.5 10 6 ), compared to less than 0.1% at most other flowrates. The increase in scatter at these Reynolds numbers can most likely be attributed to the fluctuating nature of the observed liquid jets as the gas attempts to suspend the bulk of the liquid in the main flow. It is possible that these jets could either accelerate or decelerate the turbine rotor during data acquisition, producing a wider range of measured frequencies. Fig. 12: Wet gas tests with a constant LMF of 2.0% Region 3 covers the Reynolds number range 5.4 10 6 to 9 10 6. At these flowrates only mist flow is observed with the camera, although the flow will still be of the annular-mist type. As the gas velocity increases, so the mean droplet diameter decreases, thereby ensuring that the droplets are more likely to remain completely suspended in the gas flow while moving at a significant fraction of the gas velocity. The observed shift in the meter factor decreases to within the measured band of the dry gas baseline curve for the two largest LMFs used. However, it is not clear why this should happen when at lower liquid contents there is a noticeable difference between the dry and wet gas data at the high gas flowrates. Perhaps at the higher liquid/gas mass ratios some component of the liquid droplet momentum is transferred to the turbine rotor, in effect increasing the measured frequency and therefore the meter factor. The suggestion from the NEL data at the LMFs of 1% and 2% is that bulk mist flow does not appear to have a major effect on the behaviour of the meter. An increased distribution of the Report No: 2002/100 Page 17 of 82

liquid throughout the pipe appears to effectively reduce the impact of the liquid on the meter performance. As the liquid droplets are travelling much closer to the superficial gas velocity then the meter appears to behave as if it is just measuring a less stable gas flow. Comparison with previous work The results of Jones and Ting [11] are reproduced in Fig. 13. It is apparent from comparison with Figs 9 to 12 that the NEL results are clearly different to those obtained at CEESI. In the work of Jones and Ting the shift in the meter factor occurs above a pipe Reynolds number of 5 x 10 6, whereas in the NEL work the main shifts in the data are all below a Reynolds number of 5 x 10 6. Considering that the two LMFs used in the CEESI tests were quite small relative to values used at NEL, it is perhaps surprising that a shift of as much as -0.3% was obtained at CEESI. A suitable explanation for the differences in the two data sets is not obvious. Fig. 13: Reproduction of Jones and Ting s results Figures 9 and 12 also indicate that at the smallest liquid injection rates used, the liquid presence only reduces the measured turbine frequency by between 0.1% and 0.2% across the entire Reynolds number range tested, from stratified through to annular-mist flow conditions. The reason for this difference, which is contrary to the behaviour observed at higher liquid contents is not known. Report No: 2002/100 Page 18 of 82

4.1.3 Conclusions A 6-inch turbine meter was tested in wet gas across a range of gas flow rates from 200 m 3 /hr to 1000 m 3 /hr, corresponding to a Reynolds number range of approximately 2 10 6 to 9 10 6. Four sets of tests were carried out, one with a constant liquid flowrate of 0.045 l/s, the other with constant LMFs of 0.5%, 1.0% and 2.0%. The presence of even these small quantities of liquid produced a significant effect on the turbine meter at Reynolds numbers below 5 10 6. The maximum shift in observed meter factor was -0.75% at 2% LMF. At higher Reynolds number the liquid appeared to have little effect on the meter. These results do not agree with the only previously published data on turbines in wet gas. In this previous work, the most significant effect was at Reynolds numbers above 5 10 6. Additional testing, preferably on more than one model of turbine meter would be required to allow a better examination of turbine meters in wet gas flow. 5 VORTEX METER Section 2.2 states that the vortex meter is used in situations were it might encounter wet gas flows, e.g. on the outlet from a separator. A vortex meter was therefore included in the test programme for this project to increase the small amount of information that exists on the performance of vortex meters in wet gas flows. 5.1 Test Programme A 4-inch Fisher-Rosemount vortex meter (model: 8800A, serial no: 27208) was tested in a horizontal orientation in wet gas at three gauge test pressures: 15 bar, 30 bar and 60 bar. The meter was tested across a range of gas and liquid flowrates to investigate the effect of liquid content and superficial gas velocity on the meter performance. The modified Lockhart- Martinelli parameter, X, and the gas densiometric Froude number, Fr g, (both described in section 6) were also investigated. The three test pressures chosen allowed the influence of the gas density on the meter performance to be checked. The overall test envelope is shown in Table 5.1, expressed by the superficial gas velocity (SGV), gas densiometric Froude number, Fr g, and maximum achieved X value. The liquid volume fractions used at each superficial gas velocity were 0.1%, 0.25%, 0.5%, 1%, 2.5% and 5%. The vortex meter was installed in a 100 mm test section with 30D of 100 mm pipe upstream of the meter and 9D downstream. The static pressure and temperature were measured at 5.2D and 7.6D downstream of the meter respectively. The change from 150 mm to 100 mm line sizes was achieved using an eccentric reducer which was used to prevent excessive liquid holdup in the upstream 150 mm line. The liquid was injected approximately 70D upstream of the eccentric reducer. The vortex meter was configured to give a frequency output that had a 1:1 correspondence with volumetric flowrate (i.e. a meter factor of 1) in units of m 3 /hr. This allowed direct comparison with the reference meter volumetric flowrate, corrected for the conditions at the test meter. The maximum superficial gas velocity of 30 m/s used in these tests represented approximately 40% of the maximum stated flow range of the meter. Report No: 2002/100 Page 19 of 82

Table 1 Test envelope for Vortex meter tests Pressure SGV Fr g Max Pressure SGV Fr g Max Pressure SGV Fr g Max X X X (barg) (m/s) (-) (-) (barg) (m/s) (-) (-) (barg) (m/s) (-) (-) 3 0.47 0.163 3 0.67 0.244 3 0.95 0.174 5 0.77 0.346 5 1.09 0.243 5 1.53 0.180 10 1.54 0.341 10 2.2 0.244 10 3.13 0.178 15 15 2.32 0.168 30 15 3.3 0.120 60 15 4.67 0.180 20 3.09 0.065 20 4.4 0.047 20 6.3 0.085 25 3.87 0.031 25 5.6 0.146 25 7.9 0.033 30 4.69 0.033 30 6.7 0.024 30 9.5 0.033 The superficial gas velocities used in these tests were chosen to produce a range of flow patterns in the test line. These flow patterns ranged from wavy-stratified, through a transitional zone and into an annular-mist region. However, the static pressure in the line also influences the flow pattern, tending to shift it toward an annular-mist regime as pressure increases. In wavy-stratified flow conditions the liquid runs along the bottom of the pipe. The surface of the liquid is however quite turbulent with the presence of waves as well as an indistinct, mixed two-phase interface. The slip between gas and liquid in this regime is quite high, with the gas generally travelling much faster than the liquid. As the gas velocity increases, so the flow pattern changes into a transitional form, with some liquid moving along the pipe bottom, some starting to spread out around the pipe wall and some being picked-up as droplets moving with the gas. At high gas velocities a large fraction of the liquid is suspended in the gas flow, with a continuous interchange between a liquid film at the wall and the gas core. 5.2 Results and Discussion The combined dry and wet gas test results for the vortex meter, across the entire pressure range used, are shown in Figs 14 to 16. The data is given in terms of the meter error and superficial gas velocity, as a function of the liquid volume fraction (LVF). The Gas Flow Error indicated in Figs 14 to 16 was determined using equation 4. Gas Flow Error (%) Q meter Q Q ref,corr ref,corr where Qref, corr ref meter.q ref.100 (4) (5) and Q meter is the average volumetric flowrate measured by vortex meter, Q ref,corr is the reference volumetric flowrate at the test meter and ref and meter are the reference and the test gas densities in kg/m 3. All flowrates are in m 3 /h. Inspection of Figs 14 to 16 highlight a number of points: (a) As pressure increases, so the size of the error decreases, eg at 15 bar and 0.5% LVF there is a mean error of 10%, yet at 60 bar the error reduces to a mean value of 2.5%. Report No: 2002/100 Page 20 of 82

This reduction is considered to be mainly due to a decrease in the liquid holdup level in the pipe at a fixed SGV. (b) The meter error increases with increasing LVF at all pressures, however the higher the pressure the smaller is the error increase, particularly up to an LVF of 1%. (c) For some of the high LVFs the meter error behaves in an erratic fashion, with large fluctuations in the error from one SGV to another. The effect is most pronounced at 15 bar, and diminishes as the line pressure increases. An example of this is the 1% LVF data. At 15 bar there is a spread in the data of 20%, yet at 30 bar an average error of 10% was measured with a spread of approximately 5%. At 60 bar the error reduces to around 5% and is almost totally linear across the test SGV range. Fig. 14: Meter error as a function of gas velocity and liquid volume fraction at 15 barg Report No: 2002/100 Page 21 of 82

Fig. 15: Meter error as a function of gas velocity and liquid volume fraction at 30 barg Fig. 16: Meter error as a function of gas velocity and liquid volume fraction at 60 barg NB: The exact reason for the fluctuation in the high LVF errors at 15 barg is unknown, but it may possibly be connected to some interaction between the vortex shedding mechanism and the liquid fraction/flow regime. Most of the inconsistent behaviour occurs at the lower test velocities, where it is likely that a transitional flow pattern exists, even at 15 barg. Unfortunately it was not possible to test the meter at high SGVs and LVFs together, so it is not possible to determine if the errors would return to Report No: 2002/100 Page 22 of 82

a stable level. However, some further repeat test runs conducted at the unstable vortex-meter test points at 15 barg in fact produced almost the same error values as the original test runs, indicating a consistent behaviour in the meter performance. These repeat points are included in Fig. 14 as part of the overall data set. (d) The overall performance of the meter in wet gas (with the apparent exception of some of the higher LVF results) appears to be independent of the superficial gas velocity, ie the error remains relatively constant across the entire SGV/pressure range tested at a fixed LVF. This result strongly implies that even though the gas-liquid flow pattern is changing from wavy-stratified to annular-mist as the SGV is increased, so the fraction of the total liquid content that is capable of affecting the meter performance (i.e. the stratified or annular component of the liquid flow) actually remains constant. The mist component of the liquid flow at high SGV in fact becomes invisible to the meter, as the droplets suspended in the gas core are moving much closer to the gas velocity than the annular component of the flow. 5.3 Comparison with Previous Work 5.3.1 Nederveen, Washington and Batstra As pointed out by Nederveen, Washington and Batstra [1] the vortex meter measures the fluid velocity directly. Any blockage in the line that reduces the available area for gas flow will increase the actual gas velocity at the meter and produce an error in the meter reading. The presence of liquid in the pipeline provides this blockage, so that a vortex meter exposed to wet gas flows will over-read the actual gas flowrate. The extent of the over-reading will depend upon the liquid holdup, which is itself influenced by the physical properties of the gas and liquid phases, and the slip velocity, as generally the gas travels faster than the liquid. In their 1989 paper Nederveen et al describe some (rare) vortex meter wet gas test results, obtained as part of a wider test programme investigating wet gas flow measurement under actual field and laboratory conditions for NAM in the Netherlands. The vortex meter used had a nominal diameter of 75mm and tests were conducted at 80 bar and at two gas flowrates: 85,000 and 135,000 Nm 3 /day. These correspond to superficial gas velocities of approximately 2.9 and 4.6 m/s respectively and densiometric gas Froude Nos of 0.81 and 1.29 respectively. These values put the tests in the region of the transitional boundary between stratified-wavy and annular-mist flow regimes. Water was injected into the natural gas flow line downstream of a reference Venturi meter and upstream of the vortex meter. Liquid contents (LGR=Liquid/Gas Mass Ratio) ranged from 0 to 300 m 3 liquid/10 6 Nm 3 gas (20 l/min maximum injected flowrate). This equates to a maximum liquid volume fraction of 2.56% at the minimum gas flowrate and 1.63% at the maximum gas flowrate. Report No: 2002/100 Page 23 of 82

The results of their tests are reproduced in Fig. 17 below: Fig. 17: Wetness effect on vortex meter An increasing liquid content (at both gas flowrates) clearly increases the meter over-reading relative to the dry gas values, with a maximum shift in the meter reading of around 14% at the maximum SGV and LGR tested. The authors concluded that the meter over-reading was dependent mainly on the liquid flowrate, and that the magnitude of the change in the meter error depends on the slip rate between the gas and liquid phases. The authors also suggested that vortex meters were not suitable for measuring gas flows containing entrained liquids due to an inability to produce predictable readings. Report No: 2002/100 Page 24 of 82

The results obtained at NEL are presented in a similar format to Fig. 17. In this case, however, the meter over-reading has been determined relative to the dry gas baseline data obtained at each test pressure. The results from Figs 14 to 16 are presented in the new format as Figs 18 to 20. Fig. 18: Meter over-reading as a function of the LGR and SGV at 15 barg At low LGRs (< 500), the over-readings at different SGVs all lie within a band of 2 to 3%. However, as the LGR increases the over-readings become unstable (particularly at 15 barg). The overall spread in the over-reading decreases as pressure increases but the smallest data spread is generally below a LVF of 0.5% to 1%. The slope of the over-reading curve at low LVF tends to decrease as pressure increases, and this is highlighted in Fig. 21. Report No: 2002/100 Page 25 of 82

Fig. 19: Meter over-reading as a function of the LGR and SGV at 30 barg Fig. 20: Meter over-reading as a function of the LGR and SGV at 60 barg Report No: 2002/100 Page 26 of 82

Fig. 21: Meter over-reading as a function of the LVF and test pressure The results at low LGR tend to agree qualitatively with the published data, however the absolute over-reading values are much lower for the NEL data at a given LGR. For example, at an LGR of 250 m 3 liquid/10 6 Nm 3 gas, the data of Nederveen et al suggest a meter over-reading range of between 10% and 15%. At the same LGR for the NEL data a meter over-reading of around 5% is produced, a significant difference. The larger slope of the over-reading curves of Nederveen et al may be due to a higher relative holdup of liquid in the test line. Nederveen et al did use water as the liquid phase in their tests, while at NEL a kerosene substitute was used. This results in a difference in liquid density of nearly 20%, which even at the higher test pressure of Nederveen et al appears likely to produce a significant difference in holdup. Another point to note with respect to Fig. 17 is the separation in the over-readings at the two test gas flowrates used. At first glance it appears as if there is a gas flowrate effect. This is even stated by the authors. However, closer inspection of Fig. 17 shows that some dry gas points (ie LGR = 0) have been included. At the low gas test flowrate the over-reading is in fact slightly less than 1, while at the higher flowrate it is closer to 1.03 (a 3% shift). If the wet gas data were calculated relative to these dry gas values, then both data sets would collapse directly onto each other and it could be concluded that the gas flowrates used had no effect on the meter performance. This would then agree with the NEL results presented in Section 5.2. 5.3.2 Washington Washington published another paper in 1989 [2] that contained either additional or data not reported in reference [1]. The vortex meter used was the same as in [1] and both high pressure field data and atmospheric pressure laboratory data were presented. The relevant results from the Washington paper are reproduced below as Figs 22 and 23. The format is similar to that of Fig. 17. Report No: 2002/100 Page 27 of 82

Fig. 22: Vortex meter field test results (from Ref 2) Fig. 23: Vortex meter laboratory results for horizontal installation (from Ref 3) The field test results presented in Fig. 22 were obtained at a pressure of 80 bar and at superficial gas velocities between 3.3 and 5.5 m/s. The results presented in Fig. 23 were obtained at atmospheric pressure over a velocity range of 5.6 to 28 m/s. Both data sets show the expected increase in meter reading error with increasing liquid content, but the level of the over-reading is different due to differences in holdup resulting from different liquid fractions, pressures and gas flowrates. Washington also states that one of the data sets in Report No: 2002/100 Page 28 of 82

Fig. 22 showed a shift either side of an LGR of 100 m 3 liquid/10 6 Nm 3 gas. Inspection of the data suggests a shift in the performance of the vortex meter itself. One apparent trend noticeable in the data is the increasing slopes in the lines fitted to the values of the meter over-reading, as gas flowrate increases, suggesting a gas flowrate dependency. However, the superficial gas velocity range is relatively small. Alternatively, comparison with the atmospheric test data in Fig. 23, which was conducted over a much larger gas flow range but at the much lower pressure, suggests that there is no gas velocity dependency, as overall the data is randomly scattered with no clear trend. Washington also carried out two further sets of tests on the meter, in an attempt to mitigate the effect of the liquid on the meter. The first set of tests was with the meter in a horizontal orientation with a (completely liquid filled) bypass line for the liquid, as, at atmospheric pressure the liquid it generally traveled as a stratified layer at the bottom of the pipe. This prevented the majority of the liquid from interacting with the vortex meter and so kept the over-reading to a minimum, except at the highest gas velocity where the partial annular flow pattern allowed a large fraction of the liquid to circumvent the bypass line. The second set of tests were performed with the meter in a vertical, downward orientation in an attempt to force the flow pattern into a total mist condition. However, the flow regime quickly stabilized into annular flow as close as 1.5 m upstream of the meter, even when the liquid was injected at the centre of the pipe. Consequently the author concluded that a vertical installation is not a practical way of reducing the effect of entrained liquid on a gas vortex meter. 5.3.3 Hussein and Owen A third paper dealing with the effect of wet gas on a range of meters was published by Hussein and Owen in 1991 [3]. Of some relevance to this work, it contains wet steam data for a small (2-inch) vortex meter. The meter was tested in a steam-water system at two static pressures (4 and 6 bar) and steam flowrates ranging from 294 to 929 kg/h. Liquid flowrates ranged from 56 to 177 kg /h (dryness fraction 0.84). The steam flowrates quoted lead to a range of superficial gas velocities in a 50 mm nominal bore pipe of between 12.7 and 53.9 m/s. These are relatively high velocities in comparison to those of the NEL tests, and it would be expected that the flow patterns would be between the transitional and annular-mist regimes. Figure 24 is a reproduction of the results of these tests, with a correction factor F given as a function of dryness fraction and gas mass flowrate. The correction factor (F) calculated by Hussein and Owen is not the same as the meter over-reading determined by Nederveen et al [1]. The correction factor (F) corrects the indicated meter reading (in wet gas) to produce an estimate of the total (steam and water) mass flowrate passing through the meter, rather than just the dry steam flowrate. In keeping with this approach the equivalent form of the correction factor applied to use the NEL data is presented as equation 6. Report No: 2002/100 Page 29 of 82

Fig. 24: Wet-Steam correction factor for vortex shedding meter (from Ref 3) m F g,ref meter m.q l,ref meter (6) where m g,ref and m l,ref are the reference gas and liquid mass flowrates, meter is the gas density at the vortex meter and Q meter is the apparent gas volumetric flowrate measured by the vortex meter. Figures 25 to 27 show the NEL wet gas data represented in a form matching that of Hussein and Owen. The correction factor is seen to increase as the gas quality decreases, ie an increasing liquid fraction, in-line with the data in Fig. 24. The maximum correction factor found by Hussein and Owen was 1.1 at a dryness fraction of approximately 0.84. Inspection of Figs. 25 to 27 shows that the mean correction factors found at NEL agree roughly with this value. The Hussein and Owen data also lie scattered around a correction factor which is a function of the form 1 x. This factor has been included on Figs. 25 to 27 for comparison and shows that the NEL data lies scattered either close to or around this function, over the same gas quality range. However, below a value of x = 0.7 the NEL data start to deviate strongly from 1 x, and this is most marked at the lowest test pressure. As the factor 1 x is only shown for comparison purposes, the reason for the deviation cannot be determined, because, as Hussein and Owen point out there is no physical reason for a relationship between the vortex meter behaviour and this function. Report No: 2002/100 Page 30 of 82