The Application of Ultrasonic Meters in Multiphase Flow

Transcription

1 The Application of Ultrasonic Meters in Multiphase Flow A Report for NMSPU Department of Trade & Industry 151 Buckingham Palace Road London, SW1W 9SS Report No: 35/99 Date: 2 September 1999

2 Flow Centre The Application of Ultrasonic Meters in Multiphase Flow A Report for NMSPU Department of Trade & Industry London, SW1W 9SS SUMMARY This report describes the results of multiphase flow tests conducted in research project OR5 Application of Ultrasonic Meters in Non-standard Flows of the National Measurement System threeyear Programme for Flow Both reflective and transit time meters were assessed for flowrate measurement across a common range of liquid flowrates and gas volume fractions. Ultrasonic transit time measurements are adversely affected by the presence of secondary flow components. At low velocities the meter can continue to operate with high gas volume fractions but with large error. At higher velocities small gas volume fractions (as little as.4%) can prevent the transit time measurement of velocity. Reflective meters generally performed in a less accurate and less reproducible manner. However, at some conditions good results were obtained with the reflective meters. Tests with oil and high gas volume fractions were carried out to asses the potential of improving the accuracy of a transit time meter by incorporating ultrasonic interface level measurement. The measurements made were applied to correct the volumetric flowrate indications of the velocity-based meter. This approach improved measurements in certain conditions but was limited in applicability to low liquid velocities. A transit time meter was assessed for flowrate measurement at two liquid flowrates and a range of oil/water ratios. During these tests diagnostic parameters were logged and subsequently analysed for usable relationships. It was demonstrated that data relating to the oil/water ratio could be obtained. A transit time meter was assessed for flowrate and phase-fraction measurement by utilising the meter s diagnostic parameters and incorporating an ultrasonic interface level measurement. In multiphase conditions the applicability of both the velocity and interface level measurements was reduced relative to the oil/gas tests at high gas volume fraction. The information contained in this report provides a sound basis for recommendations on the selection and use of ultrasonic meters for multiphase applications. Prepared by: Mr C J Coull... Approved by: Mr G J Brown... Date: 2 September 1999 for Dr F C Kinghorn Report No: 35/99 Page 1 of 59

3 General Manager Report No: 35/99 Page 2 of 59

4 CONTENTS Page 1 INTRODUCTION BACKGROUND THE EXPERIMENTAL PROGRAM FACILITIES AND METHODS OIL FLOW WITH GAS VOLUME FRACTIONS OIL FLOW WITH HIGH GAS VOLUME FRACTIONS TWO COMPONENT OIL / WATER FLOW MULTIPHASE OIL / WATER / GAS FLOW DISCUSSION CONCLUSIONS Report No: 35/99 Page 3 of 59

5 1 INTRODUCTION This report partially describes the results of research project OR5 Application of Ultrasonic Meters in Non-standard Flows of the National Measurement System three-year Programme for Flow Two complementary reports Velocity Distribution Effects on Ultrasonic Flowmeters Part 1 Theoretical Analysis and Part 2 Determination by Computational and Experimental Methods detail further aspects of the project. Two further reports entitled Long-term Evaluation of Ultrasonic Flowmeters and Research into Clamp-on Ultrasonic Meters describe results of two related projects in the Programme. 2 BACKGROUND Ultrasonic flowmeters have the potential for non-intrusive, accurate and repeatable measurement. Their performance is, however, affected by non-standard conditions as was identified in the Flow Programme project Ultrasonic Meters for Oil Flow Measurement. It was identified that two of the major influences on meter accuracy in application are velocity distribution effects and multiphase flow effects. In this report the application of ultrasonic meters in multiphase flow conditions is addressed. Consideration is given to those ultrasonic flowmeter types typically encountered in practice. Ultrasonic flowmeters can be categorised as working on either reflective or transmissive principles of operation. Transit time difference and beam drift flowmeters are examples of meter types where the interaction measured is the effect of the flow on the signal transmitted through the fluid. Doppler and 'particle tracking' techniques utilise effects whereby the reflection of the signal by the fluid produces the measurable interaction. A principle that could be considered inclusive of both transmissive and reflective principles is the crosscorrelation flowmeter. However, as this technique does not derive its flow measurement from a fundamental dependence of the ultrasonic signal parameter on the flow velocity it is considered as a separate category. Another technique given consideration in this report, although not a flow measurement in terms of velocity or volumetric flowrate, is an interface measurement technique. This technique utilises the highly reflective nature of the gas/liquid boundary present in two-phase stratified and slug flows to locate this boundary by transit time measurement. 2.1 Transmissive Flowmeters There are essentially two types of transmissive ultrasonic flowmeter, the beam-drift flowmeter, and the transit time flowmeter. Beam-drift flowmeters If a beam of ultrasound is transmitted across a flowing fluid, its path will be deflected by an amount dependent on the fluid and sonic velocities. The beam drift can be determined from the amplitude of the signal received by a transducer on the far side. Design variations include differential receiving transducers and axial mechanical scanning of a single receiving transducer. The uncertainty in amplitude measurement and resultant low sensitivity to velocity is the limiting factor with this technique. Its industrial use appears to have been limited to one- Report No: 35/99 Page 4 of 59

6 off applications and no commercially available devices currently appear on the Report No: 35/99 Page 5 of 59

7 market. In multiphase application this technique would suffer from poor accuracy as the amplitude modulation due to phase variation in the beam would interfere with the velocity measurement. For these reasons, this technique was not considered further. Transit time flowmeters Ultrasonic transit time flowmeters are based on measurement of the propagation time of acoustic waves in a flowing medium. Generally, the assertion that the apparent velocity along a ray is given by the velocity of sound in the fluid at rest, c f, plus the component of fluid velocity along the ray is applied. To eliminate the velocity of sound from the subsequent derivation, transit times are determined both in the direction of flow and against it. Considering the general ray geometry shown in Figure 2.1, the upstream transit time and downstream transit times are given by t ab L = ( c v cos θ) f and t ba L = ( c + v cos θ) f (1) where v 1 = vdx L (2) a b d a a l L θ D b b Figure General ray geometry for transit time velocity measurement There are four basic methods by which transit time velocity measurement is performed; direct time differential, phase differential, phase control, and frequency differential. In modern ultrasonic flowmeters the direct time differential method is most common. Short time duration pulses are propagated either sequentially or simultaneously upstream and downstream and the time duration between excitation and reception is determined. The expressions for the upstream and downstream transit times are then solved for v as follows: 1 1 = t t ba ab ( c f + v cosθ) ( c f v cosθ) L (3) t t ab t t ab ba ba v = 2 cosθ (4) L v 2 L t L t = = 2 cosθ t t 2d t t ab ba ab ba (5) Report No: 35/99 Page 6 of 59

8 Multiplying v by the cross-sectional area of the flow, A, the volumetric flowrate is obtained. q A L V = 2d 2 t t t ab ba (6) In multiphase application the modulation of the signal by phase variations in the beam is the principle problem although variations in velocity distribution will also affect accuracy. As the transit time technique is employed widely in industrial applications, further understanding of its uses and limitations in this respect could be widely applied. 2.2 Reflective Flowmeters There are essentially two types of reflective flowmeter, which can be categorised as those based on frequency-domain signal analysis and those which utilise time-domain techniques Frequency domain reflective flowmeters These meters make use of the well-known Doppler effect. When ultrasound of frequency f t is transmitted into the flow medium, discontinuities such as entrained gas bubbles or solid particles scatter the ultrasound. Received ultrasonic energy, scattered by moving particles, is Doppler frequency shifted by an amount δf, given by δf = f t f r = vcosθ f t c r v cosθ + c t (7) where f r is the Doppler shifted frequency, v is the velocity of the particle with respect to the conduit axis, θ t and θ r are the transmission and reception angles, and c is the velocity of sound in the flow medium as illustrated in Figure 2.2. f t f r θ t θ r v c Figure The Doppler flowmeter principle The angles θ t and θ r in Figure 2.2 depend upon the configuration of the transducers employed in the meter. Typically, Doppler meters are designed as externally mounted or clamp-on devices. Therefore, the ultrasound is refracted at each acoustically discontinuous interface between the transducers and the flow medium as illustrated in Figure 2.3. Report No: 35/99 Page 7 of 59

9 Housing Internal coupling Transducer Pipe wall θ θ β β α Liquid Figure Beam refraction in clamp-on transducers This effect, while presenting problems in time-of-flight type usfm s can be used to advantage in Doppler meters to eliminate the dependence on the velocity of sound. Refraction of ultrasound is governed by Snell s law, which can be presented as cosα cosβ cosθ = = (8) c 1 c 2 c f If the transducers are configured such that α t = α r = α then Equation (1) can be rewritten as δf v cosθ = 2 ft c f (9) substituting cosα/c 1 for cosθ/c f and rearranging for v yields δfc1 v = 2 cosα f t (1) i.e. the measured velocity is proportional to δf and independent of the acoustic properties of the fluid and the pipe wall. The angle α and velocity of sound c 1 are assumed to have known values. However, temperature dependence of c 1, the velocity in the transducer potting or internal coupling material, may give rise to significant errors in the measurement of velocity. For example, if the transducer was coupled at angle via a nylon wedge, a shift in temperature from 25 to 4 C would result in an error in v of approximately 2.5%. Doppler Signal Processing The simplest way to measure the Doppler shift is with a zero crossing detector. The received signal is mixed with the oscillator frequency to produce a high frequency amplitude modulated signal which can be demodulated to extract the raw low frequency Doppler shifted signal. As this signal contains a broad band of frequencies of form similar to that shown in Figure 2.4, simple zero crossing analysis tends to under-estimate the peak frequency (and hence mean velocity) of the Doppler spectrum. To achieve a more precise estimate of the mean velocity more sophisticated methods such as Fourier analysis may be employed, although this does not eliminate the influence of factors discussed in following paragraphs. Report No: 35/99 Page 8 of 59

10 Amplitude Zero crossing frequency Frequency ~ mean velocity Frequency Range Gated Doppler Figure Inaccuracies in processing of Doppler signals using the zero-crossing technique The use of pulsed rather than continuous ultrasound permits the introduction of positional information in Doppler flow measurement. By exciting a short duration pulse, and time gating the reception interval, the position of the measuring zone can be controlled. By sweeping the time delay or sampling and processing the received signal in a series of discrete time intervals, it is possible to determine flow velocity profile information. This technique is not applied in industrial instruments but has proved useful in fluid mechanics and medical research Time domain reflective flowmeters Time domain reflective flowmeters are similar in principle to Doppler flowmeters. In fact, one could argue that they are a simply a variant of Doppler technology. Like Doppler meters they rely on discontinuities such as entrained gas bubbles or solid particles to scatter the ultrasound. Where they differ from Doppler meters is in the use of pulsed ultrasound and the time-domain derivation of the velocity measurement from the raw ultrasonic signals. Short duration pulses are transmitted at a high repetition rate into the fluid. The technique bears some resemblance to pulse-echo thickness measurement, except that in this case the reflectors are moving with respect to the transducer. Therefore, the pulse-echo transit time increases as the scattering centre moves though the ultrasonic beam as illustrated in Figure 2.5. θ v δx δd c f Figure A schematic diagram illustrating Doppler theory Report No: 35/99 Page 9 of 59

11 The transit time difference for successive pulse transmissions is given by 2δd t = (11) c f where the distance δ d is proportional to the axial displacement of the scattering centre, i.e. c f t δ d = δx cosθ = (12) 2 To obtain the velocity measurement we rearrange Equation (12) for δx and divide by δt, the time interval between succesive pulse transmissions, to give the result δx c f t v = = (13) δt 2cosθ δt Note the similarity of Equation (13) to Equation (1) in the previous section. Although restating the time intervals as frequencies (and performing the Snell s law substitution) gives an identical equation, it is useful to note that the pulse repetition rate ( 1 δt ) used in a time domain flowmeter can be varied with relative ease whereas the Doppler transmission frequency ( f t ) tends to be fixed. Sources of error in reflective flow meters Up to this point in the discussion, the measured Doppler shift or transit interval has been considered as representative of the mean velocity of the flow. In realistic industrial flow situations this is rarely the case. Extrinisc factors that affect the accuracy are discussed in the below Fluid/particle slip The flow signal originates from the scattering centres in the fluid whose velocity may not be the same as that of the bulk fluid. This can be quantified in terms of the slip velocity v sl, vsl = v f v p (14) where v f is the fluid velocity and v p is the particle velocity, error being given by v sl /v f. Attenuation/unevenly distributed scatterers In the desirable but unlikely situation of particles that are uniformly distributed in the fluid and each making an equal contribution to the received signal, the Doppler shift will comprise a range of frequencies. The spectrum will be dependent on the velocity profile and the area of the cross-section that the ultrasound integrates. Therefore, with knowledge of the function form of the flow profile and of the beam geometry the mean velocity may be determined by analysis of the Doppler spectrum. Report No: 35/99 Page 1 of 59

12 In the industrial situation it is unrealistic to assume that the scatterers are evenly distributed. Furthermore, even if the previous assumption holds, attenuation of ultrasound in the medium makes it unlikely that all scatterers contribute equally to the Doppler shift signal. If we consider particles heavier or lighter than the fluid these will tend to collect at either the bottom or the top of the pipe respectively and therefore the Doppler signal will be dominated by the velocity of the flow in these area. If we consider an attenuative medium, the signal penetration will be poor and the near-wall velocities will dominate the Doppler signal. These effects are illustrated in Figure 2.6(a), (b) and (c) respectively. In some multiphase conditions, attenuation may be so great that a measurable signal cannot be found. (a) (b) (c) Figure Particle distribution and attenuation effects 2.3 Cross-Correlation Flowmeters These flowmeters are based on the measurement of flow transit time between two sensors. Rather than relying on an injected tracer, a naturally occurring property of the flow is used to randomly modulate the signals of the upstream and downstream sensors. The result is two similar noise signals displaced by a time τ m = l/v where l is the axial spacing of the sensors and v is the flow velocity. The delay τ m is determined by computing the cross-correlation function R xy T lim 1 ( l, τ ) = T x t y t dt T ( + τ ) ( ) (15) where x(t+τ) is the upstream signal delayed by time τ, y(τ) is the down stream signal, and T is the integration time. In the idealised case R xy displays a distinct peak at τ = τ m. The use of ultrasonic sensors for cross-correlation techniques has received much research attention in the past, mainly due to the potential for non-intrusive multiphase flow measurement at higher accuracy than is currently achievable with Doppler techniques. However, despite the apparent simplicity of the method, many difficulties have been encountered. The following paragraphs cover some further aspects of ultrasonic crosscorrelation flow measurement. Modulation of Continuous Waves The measurement of amplitude and phase modulated continuous waves is the most commonly encountered sensing technique in ultrasonic cross-correlation flow measurement. As illustrated in Figure 2.7, two transmitter/receiver pairs produce beams that interrogate the fluid at a fixed separation and typically across the pipe diameter. Random variations in the acoustic impedance of the medium modulate the beams, the received signals being demodulated to produce the noise inputs to the cross-correlator. Report No: 35/99 Page 11 of 59

13 Transmitter Transmitter v Receiver Receiver x(t) Correlator y(t) Figure A schematic diagram of a CW cross-correlation flowmeter Difficulty in controlling the performance of continuous wave (CW) systems has been identified as being related to instabilities of the oscillator and resulting variations in the acoustic standing wave pattern. This problem is effectively eliminated by the use of a closed-loop phase-locking system. Clamp-on transducer arrangements Although the cross-correlation flowmeter was originally envisaged as a low cost clamp-on device, such clamp-on arrangements reduce achievable accuracy as prediction of the beam spacing and orientation is less reliable. Furthermore, the ultrasound travelling in the pipe wall short circuit limits the range of signal attenuation that can be tolerated. The acoustic short circuit has also been postulated as the source of phase errors, normally attributed standing wave variations. Pulsed ultrasonic cross-correlation flowmeters A cross-correlation flowmeter based on amplitude modulation of short time duration ultrasonic pulses effectively eliminates problems related to standing waves and acoustic short circuits. In this design, pulses are propagated across the flow, the reception system comprising a peak detector, sample-and-hold circuit and low pass filter to demodulate the noise signal. In order for a pulsed system to be effective the pulse repetition rate must be sufficiently high in order to recover the spectrum of the noise signal. A variation of the pulsed cross-correlation flowmeter uses the flow-modulated signal reflected from the back wall of the pipe. Industrial cross-correlators Much of the early developments in cross-correlation flow measurement addressed the need for a rapid low-cost processor to compute the cross-correlation function. The approach generally adopted is to convert the noise signal into a binary bit-polarity sampled waveform as illustrated in Figure 2.8. This allows the implementation to be greatly simplified although a longer integration time is required in order to achieve the same statistical accuracy. Crosscorrelators suitable for industrial measurements were introduced commercially in the 198 s. Report No: 35/99 Page 12 of 59

14 Random wavform x(t) t 1 Binary bit-polarity sampled waveform t Figure Signal pre-processing utilised in industrial correlators Like Doppler meters, cross-correlation ultrasonic flowmeters require markers in the flow in the form of a second component or property fluctuation, making them susceptible to sources of uncertainty similar to the Doppler technique. They are not ideally suited to single-phase flow of clean fluids, although measurements have been made on clean water and gases. Greater interest has been shown in application to difficult flow media such as paper pulp suspensions, hydraulically transported solids and liquid/gas mixtures. Despite improvements in high-speed signal processing and reduced costs of microprocessors, the ultrasonic cross-correlation flowmeter has not been a commercial success. Introduction of a cross-correlation meter by Kent in the 198 s was later abandoned. Due to the current lack of commercially available instrumentation, this technique was not considered further. 2.4 Interface Measurement In multiphase flows, propagating ultrasonic waves will undergo reflection at the interfacial boundaries between the gas and liquid components. The primary descriptor of this interaction is the characteristic acoustic impedance R = ρ c where ρ is the medium density and c is the propagation velocity of sound (VOS) in the medium. In general practice the VOS in fluids is considered independent of the ultrasonic frequency. The intensity transmission (α t ) and reflection (α r ) coefficients for normal incidence at a boundary may be obtained from the following expressions ( ) α t = R R R + R ( ) ( ) α r = R R R + R (16) (17) where the subscripts indicate the respective media forming the interface. Table 2.1 below gives approximate acoustic properties for a selection of gaseous, liquid and solid media. Report No: 35/99 Page 13 of 59

15 Table 2.1 Medium Steel Perspex Water Air Longitudinal wave velocity, c (m.s -1 ) Density, ρ (kg.m -3 ) Characteristic impedance, R (kg.m -.s -1 ) The significant acoustic impedance mismatch between gaseous and solid media gives rise to interactions that are problematic in relation to many ultrasonic flow measurement techniques. For example, previous experience at NEL has shown that flow velocity measurement based on transit time measurement or cross-correlation of ultrasonic signals were possible only to a limiting gas fraction. However, the strong reflective nature of the gas/liquid interface can permit the measurement of the interface location. The principle of interface position measurement using ultrasound is based upon measurement of the transit time of the ultrasonic signal as illustrated in Figure 2.9 and described by the simple equation d c1 t = 2 (18) Transducer c 1,ρ1 c 2,ρ2 d Figure The Ultrasonic Interface Measurement Technique 3 THE EXPERIMENTAL PROGRAMME 3.1 Oil Flow With Low Gas Volume Fractions Both reflective and transit time meters were assessed for flowrate measurement across a common range of liquid flowrates and gas volume fractions, the aim being to identify conditions most suited to each instrument. During these tests diagnostic parameters were logged and subsequently analysed for usable relationships. The test section and meters were orientated horizontally. The low gas fraction tests were conducted using oil of viscosity nominally 25 cst and nitrogen gas. Operating temperature and pressure were nominally 2 C and 1 bar gauge respectively. The matrix of nominal test conditions is shown in Table 3.1. Report No: 35/99 Page 14 of 59

16 Table Nominal Conditions for the Oil Flow with Low Gas Volume Fraction Tests Nom. Nominal GVF (%) Flowrate (l/s) X X X X 9 X X X X X 18 X X X X X 4 X X X X X 55 X X X X X 8 X X X X X 91 X X X X X 3.2 Oil Flow With High Gas Volume Fractions Tests with oil and high gas volume fractions in the region were carried out to asses the potential of improving the accuracy of a transit time meter by incorporating ultrasonic interface level measurement. Operating temperature and pressure were nominally 2 C and 1 Bar gauge respectively. The tests were conducted using oil of viscosity nominally 25 cst and nitrogen gas. The test section and meters were orientated horizontally. The matrix of nominal test conditions is shown in Table 3.2 Table Nominal Conditions for the Oil Flow with High Gas Volume Fraction Tests Flowrate (l/s) GVF (%) Two Component Oil/Water Flow A transit time meter was assessed for flowrate measurement at two liquid flowrates and a range of oil/water ratios. During these tests diagnostic parameters were logged and subsequently analysed for usable relationships. The test section and meter were orientated horizontally. The matrix of nominal test conditions is shown in Table 3.3. Table Nominal Conditions for the Two Component Oil/Water Flow Tests Nominal watercut 3.4 Multiphase Oil/Water/Gas Flow Nominal flowrates (l/s) 5 % % % 1 2 Report No: 35/99 Page 15 of 59

17 A transit time meter was assessed for flowrate and phase-fraction measurement by utilising the meters diagnostic parameters and incorporating an ultrasonic interface level measurement. The test matrix covered liquid flowrates of 5 to 16 l/s, oil/water ratios of 5, 4, and 75%, and gas volume fractions of 25 to 71.5%. The test section and meters were orientated horizontally. The matrix of nominal test conditions is shown in Table 3.4 Table Nominal conditions for the multiphase oil/water/gas flow tests Water Cut 5% Flowrate GVF (%) (l/s) Water Cut 4% Flowrate GVF (%) (l/s) Water Cut 75% Flowrate GVF (%) (l/s) FACILITIES AND METHODS 4.1 Flow Measurement Instrumentation Meters were obtained from three manufacturers for the programme. A brief description of the instrumentation supplied follows DANFOSS Sonoflow Manufacturer: Converter type: Converter Serial No: Sensor type: Sensor Serial No: Danfoss A/S Industrial Instrumentation SONO N355 SONO33 (DN1, CLASS15) 22N16 The SONO33 is a sensor tube with four transducers forming two parallel tilted chordal sound paths. Each path is formed at a mid-radius position in the cross-section and is inclined at an angle of 39 to the tube axis. The transducers are metal encapsulated piezoelectric elements that are mounted in the cast body by means of a screw fitting. A memory chip that is pre-programmed with the primary physical parameters and factory calibration factor is Report No: 35/99 Page 16 of 59

18 provided with the meter tube. The calibration report issued by Danfoss lists the primary physical parameters and factory calibration factor as reproduced below. Report No: 35/99 Page 17 of 59

19 Di =.124 [m] {Meter tube internal diameter} L =.1794 [m] {Distance between the windows of each transducer pair} h =.25 [m] {Radial displacement of the chordal paths}? = 39. [ ] {Angle between the pipe centreline and sound paths} Cal. Factor = The SONO3 signal converter performs the excitation and detection of ultrasonic signals and subsequent processing and flow computation. The transducers are excited by a sine wave of 8 cycles duration and a frequency of around 1 MHz. The received upstream and downstream signals are digitised and stored for processing. Automatic gain control (AGC) is utilised to optimise the analogue to digital conversion. The flow computation is based on measurement of the overall transit-time and the upstream-downstream transit-time difference, which are both treated separately by a correlation technique. The transit-time difference is determined by cross-correlating the upstream and downstream signals whereas the overall transit-time is determined by cross-correlating one of the received signals with a stored reference signal. This stored reference signal is dynamically adapted to the operation conditions. Set-up parameters of importance which were adjusted are listed below. All Other Tests Multiphase/2-Component Tests Vol. Flow Max: 1 l/s 4 l/s Low Flow Cut-off:.5% (of max).% (of max) Time Constant: 1 second 1 second Sound Velocity Min: 12 m/s 13 m/s Sound Velocity Max: 15 m/s 16 m/s Panametrics DF868 Model No: DF s Serial No: DF496-E Transducer No: SN , SN 79761, SN , SN The Panametrics DF868 is a dual channel flow transmitter for use with clamp-on or wetted transducers. Panametrics provided two pairs of clamp-on transducers. The meter was configured to operate in two modes, the transmissive transit time mode and a reflective timedomain Transflection mode. The transit time configuration employed a pair of 1 MHz shear-wave transducers which were mounted on the exterior of a length of 1 mm nominal bore Schedule 4 carbon steel pipe by means of a yoke-and-strap arrangement and coupled to the pipe wall using a coupling grease. The transducers were mounted such that a single-reflection path was formed in the horizontal plane. The transflection configuration employed a pair of 5 khz shear-wave transducers which were mounted on the exterior of a length of 1 mm nominal bore Schedule 4 carbon steel pipe by means of a yoke-and-strap arrangement and coupled to the pipe wall using a coupling grease. The transducers were diametrically opposed on the horizontal plane. Report No: 35/99 Page 18 of 59

20 The Panametrics DF868 uses a coded excitation and correlation detection scheme to determine the transit time of the ultrasonic signals. The excitation signal, rather than an impulse or sinusoid, is a phase-coded square-wave burst. AGC is utilised to optimise the signal level before digitisation and averaging of several successive signals. The digitised receive signals are then cross-correlated with a stored version of the excitation signal to determine transit time Buhler Montec Zytec Model No: Zytec 74 Serial No: RD1/2141 The Zytec 74 is a reflective flowmeter of the Doppler type. Two clamp-on transducers were attached to the pipe at 45 degrees above and below the horizontal on the same side of the pipe. On transducer acts as the transmitter operating at 2 MHz. The other transducer acts as the receiver and picks up the transmitted signal and the Doppler-shifted signal. The Doppler-shifted signal and the transmitted signal are mixed to produce a high frequency signal with a low-frequency amplitude modulation. The Doppler signal frequency is determined by eliminating the high frequency components and feeding the resulting signal into a zero-crossing detector The interface measurement equipment To achieve measurement of the interface position, a system comprised of pulser/pre-amplifier, digitising oscilloscope, and commercial 1 MHz ultrasonic transducer was utilised. The digitising oscilloscope was connected to a PC with data acquisition software in order to capture and process the raw ultrasonic signals. The system configuration is illustrated in Figure 4.1 below. The pulser/pre-amp excites the transducer with a high-voltage transient that generates an ultrasonic pulse that is transmitted into the adjoining medium. Returning ultrasonic signals (echoes) are converted to electrical signals by the transducer and amplified to a suitable level by the pre-amp. The amplified signals are captured by the oscilloscope, the pulse repetition rate and synchronisation being controlled by the pulser/pre-amp unit. The arming of the oscilloscope and data transfer is controlled by the PC, where subsequent processing of the raw signal is performed using software written specifically for the application. Digitising oscilloscope PC GPIB Transducer Waveform Pulser/ pre-amp Trig Figure A Schematic Diagram of the Ultrasonic System Configuration The ultrasonic transducer was coupled to the outer wall of a block-shaped Perspex flow cell with a 1 mm nominal bore as illustrated in Figure 4.2 (standard transducer). A mineral oil based couplant was utilised to improve transmission efficiency between the Perspex and the Report No: 35/99 Page 19 of 59

21 transducer face by eliminating the air gap that would otherwise be present as a result of object and transducer surface roughness. Gas-coupled transducer Cavity x 4 GAS x 1 c 1 LIQUID c 2 x 2 d Perspex section c 3 x 3 Standard transducer Figure A Schematic Diagram of the Flow Cell This buffered, non-invasive or non-wetted application of the transducer dictates that the transmitted ultrasound encounters the Perspex/liquid interface prior to the liquid/gas interface and the result is a waveform comprised of primary and secondary reflections. 4.2 Test Facilities The two-phase oil/gas tests reported here were conducted using a circuit of the main oil flow primary standard by introducing low flowrates of metered nitrogen into the test line on the downstream side of a reference positive displacement (PD) meter. The set-up is shown schematically in Figure 4.3. Turbine meters T P Nitrogen supply T P PD meter T P Test meters T P Viewing sections Bypass Pump Supply Tank Figure A schematic diagram of the oil/gas test facility Report No: 35/99 Page 2 of 59

22 The two-phase oil/water and three-phase oil/water/gas tests reported here were conducted in the multiphase calibration facility at NEL. A schematic diagram of the facility, as used for these tests, is shown in Figure 4.4. The total reserve of oil and water is held in a vessel that acts as a combined storage tank and multiphase separator with water, oil and mixture compartments. The oil and water are drawn from each of the single-phase compartments and are delivered to the flow loop via calibrated reference meters. Also present in the system are sampling loops for the determination of background quantities of oil-in-water and water-in-oil. The gas is injected via gas turbines from an external nitrogen supply. After the oil, water and gas are allowed to co-mingle the fluids then flow to the test section along a 5 m development length of straight pipe. Separator vessel Water Mixture Oil Pump Densitometer Pump Water-cut monitor Turbine meters Turbine meters T P Nitrogen supply Turbine meters Test meters Figure A schematic diagram of the multiphase facility The oil used in the facility was a crude-kerosene mix with approximate density and viscosity of 846 kg/m 3 and 6.6 cst respectively at the test temperature of 4 C. The water used was a brine solution of density approximately 123 kg/m 3 at the test temperature. 4.3 Interface Level Data Processing The interface level equipment records both the level and the amplitude of the received ultrasonic signal. During each test point, two hundred measurements are made at an interval of approximately half a second. Figure 4.5 shows a record obtained during slug flow, the upper trace showing the calculated interface level, the lower showing the recorded peak to peak voltage values. By rejecting the level measurements corresponding to peak voltage values below the threshold noise level in the ultrasonic waveform a filtered data set was produced. The threshold level was set at 15mVp-p. Figure 4.6 shows the results of applying a threshold level of 15 mvp-p to the data of Figure 4.5. Report No: 35/99 Page 21 of 59

23 Liquid Level (mm) Vp-p Time (s) Figure An interface level record obtained during slug flow 1 Liquid Level (mm) Valid Invalid Time (s) Figure Threshold filtered results The interface level is the average of the valid measurements: Sum of Valid IL' s IL = (19) No. of Valid Measurements The percentage of valid measurements (threshold pass rate, TPR) is used as a diagnostic tool to measure the performance of the interface level technique. No. of Valid Measurements TPR = *1% (2) Total No. of Measurements 4.4 Utilisation of Interface Level Results In two-phase oil/gas stratified flow the upper path in the Danfoss transit time meter would not be expected to operate due the gas in the upper section of the pipe. However the meter can and does operate on the results from its lower path. Report No: 35/99 Page 22 of 59

24 Applying the interface level measurement to the simple model shown in Figure 4.7 allows the ratio of liquid filled pipe area to total pipe area to be calculated using Equations (21) and (22). Gas Oil A L h L Figure A simple stratified flow model.25 π cos AL = A 1 ( 2 h ~ 1) + ( 2 h ~ 1) 1 ( 2 ~ h 1 L L L ) π 4 2 (21) where h ~ L = h L D (22) Area compensation factors the output of the meter operating on the lower path (V ind ) with the area ratio to give a predicted liquid flow (V AC ): V AL = V ind (23) A AC * The gas volume fraction (GVF) can be estimated by the following formula, however it does assume that the gas and liquid velocities are equal and is therefore of limited value. GVF = 1 AL *1% (24) A 5 OIL FLOW WITH LOW GAS VOLUME FRACTIONS The following sections present the results of tests on the four meters evaluated against this low gas volume fraction test matrix. The results are presented in terms of error relative to the total volumetric flowrate in the pipe, i.e. the sum of the liquid and gas flowrates. The results are described in the text in terms of the deviation from a specified reference condition. 5.1 Danfoss Sono 3/33 Test Results Previous tests on the Sonoflow indicated unstable behaviour when the upper path was affected by the presence of gas. Therefore these tests have been performed with the upper path disabled, to investigate the performance of the meter while only the lower path was functioning. Report No: 35/99 Page 23 of 59

25 5 Danfoss Sono 3/33 Total Vol. Error (%) GVF = % GVF = 1.9% GVF = 5.2% GVF = 27.% Figure Oil / Low GVF test at nominal liquid flow of 5 l/s Figure 5.1 shows the results of the oil / low gas volume fraction tests at a nominal flowrate of 5 l/s. At 1.9% GVF the deviation from the zero GVF error is between -2 and -4%. At 5.2% GVF the deviation against the zero GVF error has increased to between -5 and -8%. At 27.% GVF the deviation continues to increase to between 22.8 and -25%. During none of these tests did the meter indicate a fault condition. 4 Danfoss Sono 3/33 2 Total Vol. Error (%) GVF = % GVF =.9% GVF = 1.8% GVF = 4.% GVF = 9.4% -8-1 Figure Oil / Low GVF test at nominal liquid flow of 9.3 l/s Figure 5.2 shows the results at a nominal flowrate of 9.3 l/s. At.9% GVF the deviation from the zero GVF error was between -1.5 and -2%. At 1.8% GVF the deviation has increased to between -3.5 and -4%. At 4% GVF the deviation has increased to between -7 and -7.5%. At 9.4% GVF the deviation has increased to between -9 and -1%. All these results show a repeatability of 1% or better. Again during none of these tests did the meter indicate a fault condition. Report No: 35/99 Page 24 of 59

26 Danfoss Sono 3/33 1 Total Vol. Error (%) GVF = % GVF =.4% GVF =.9% GVF = 1.6% GVF = 4.5% -4-5 Figure Oil / Low GVF test at nominal liquid flow of 18 l/s Figure 5.3 shows the results at a nominal flowrate of 18 l/s. At.4% GVF the deviation from the zero GVF error was between -.4 and -.8%. At.9% GVF the deviation has increased to between -1. and -1.4%. At 1.6% GVF the deviation has increased to between -2 and -2.4%. At 4.5% GVF the deviation has increased to between -4.3 and -5.1%. Again the meter did not indicate a fault condition. Danfoss Sono 3/ Total Vol. Error (%) GVF = % GVF =.45% GVF = 1.% GVF = 1.9% GVF = 4.7% Figure Oil / Low GVF test at nominal liquid flow of 39.9l/s Figure 5.4 shows the results at a nominal flowrate of 39.9 l/s. At.45% GVF the deviation was between -.1 and -.7%. At 1% GVF the deviation has increased to between -.2 to - 2.1%. At 1.9% GVF the deviation has increased to between -6.3 and +1.1 %. At 4.7% GVF the deviation has increased to between and +21%. This graph illustrates the reduction in meter repeatability at higher gas volume fractions. However the meter still does not indicate a fault condition. Report No: 35/99 Page 25 of 59

27 3 Danfoss Sono 3/33 2 Total Vol. Error (%) GVF = % GVF =.1% GVF =.5% GVF = 1.8% GVF = 4.3% Figure Oil / Low GVF test at nominal liquid flow of 55.1 l/s Figure 5.5 shows the results at a nominal flowrate of 55.1 l/s. At.1% GVF no fault condition was observed and the deviation was between -.5% and +.1%. At.5% GVF the deviation increased to between -1.2 and +2.8%, and still no fault condition was indicated. At 1.8 and 4.3% GVF the meter indicates that the lower path (the only operating path), signal is too weak to make a valid measurement. The output goes to its forced condition, which is set at zero. Danfoss Sono 3/ Total Vol. Error (%) GVF = % GVF =.1% GVF =.4% GVF =.9% GVF = 1.8% Figure Oil / Low GVF test at nominal liquid flow of 78.9 l/s Figure 5.6 shows the results at a nominal flowrate of 78.9 l/s. At.1% GVF, during one test point the meter indicated a high signal gain warning, which indicated a low signal level. However the meter showed a deviation less than +.3%. At.4% GVF the meter indicated the following errors, velocity of sound out with measuring range, high signal gain, and fatal measuring error, output in forced condition. The fatal measuring error occurred intermittently, hence the error was slightly less than -1%. At.9 and 1.8% GVF the meter signal was continually forced low due to the weak detected signal. Report No: 35/99 Page 26 of 59

28 Danfoss Sono 3/ Total Vol. Error (%) GVF = % GVF =.2% GVF =.35% GVF =.9% GVF = 1.8% Figure Oil / Low GVF test at nominal liquid flow of 9.5 l/s Figure 5.7 shows the results at a nominal flowrate of 9.5 l/s. At.2% GVF no fault condition was observed, the deviation was less than +.3%. At GVF s of.35% and above the meter signal was continually forced low due to a weak received signal. 5.2 Danfoss Sonoflo 3/33 Diagnostics The Danfoss meter has a service menu the can be utilised to display a number of diagnostic parameters including the signal-to-noise ratio of the ultrasonic signal. The following graphs illustrate the relationship between the lower path signal-to-noise ratio and the meter error observed during three of the tests Danfoss Sono 3/ Total Vol. Error (%) SNR Danfoss Total Vol Error (%) SNR (Lower Track) GVF (%) 1 5 Figure Oil / Low GVF test at nominal liquid flow of 9.3 l/s Figure 5.8 shows the meter error and the signal-to-noise ratio (SNR) for the lower path, against the gas volume fraction for a nominal liquid flowrate of 9.3 l/s. The SNR remained constant in the range 35 to 45, while the meter error increased from +2 to 8%. At this low superficial liquid velocity of 1.2 m/s the gas is concentrated in the upper portion of the pipe and does not affect the lower path signal. Thus at low velocities the lower path SNR does not provide an indication of meter error. The SNR of the upper path would have provided an indication of meter error at lower GVF if it had been connected. Report No: 35/99 Page 27 of 59

29 25 2 Danfoss Sono 3/ Total Vol. Error (%) SNR Danfoss Total Vol Error (%) SNR (lower track) GVF (%) -2-3 Figure Oil / Low GVF test at nominal liquid flow of 39.9 l/s Figure 5.9 shows meter error and the signal noise ratio against gas volume fraction for a nominal liquid flowrate of 39.9 l/s. Over a GVF range of to 4.7%, the signal-to-noise ratio drops from around 4 to approximately 15. This corresponds to an increase in result scatter from.25% at % GVF to 33% at 4.7%GVF. At this superficial liquid velocity of 5.1 m/s the gas was fairly evenly across the pipes cross section ensuring that the gas bubbles affected the lower path. The graph shows that the decreasing SNR correlates with the reduction in repeatability providing an indication of the onset of measurement failure. 1 5 Danfoss Sono 3/ Total Vol. Error (%) SNR Danfoss Total Vol Error (%) SNR (lower track) GVF (%) Figure Oil / Low GVF test at nominal liquid flow of 78.9 l/s Figure 5.1 shows meter error and the signal noise ratio against gas volume fraction for a nominal liquid flowrate of 78.9 l/s. At GVF s between and.1% the meter error was between.4 and.8%. Over this range the signal noise ratio dropped from approximately 38 to around 28. Above.4 % GVF the meter output was forced low due to a weak signal on the lower path. This is graphically illustrated on the graph, as the SNR drops below 1. This graph shows at superficial velocities of 1 m/s the SNR acts as a switch indicating whether the meter is working or not. 5.3 Panametrics DF868 Test Results (Transit Time Mode) Report No: 35/99 Page 28 of 59

30 Panametrics DF868 Transit Time 2 1 Total Vol. Error (%) GVF = % GVF = 2.% GVF = 2.3% GVF = 39.4% -4-5 Figure Oil / Low GVF test at nominal liquid flow of 5 l/s Figure 5.11 shows the results of the oil flow with low gas volume fraction tests at a nominal flowrate of 5 l/s. At 2% GVF the deviation from the zero GVF error is between and -2.3%. At 2.3% GVF the deviation has increased to between and -33%. At 39.4% GVF the meter indicates a signal amplitude error, and as a result the deviation increased to between and -48.7%. 6 Panametrics DF868 Transit Time Total Vol. Error (%) GVF = % GVF = 1% GVF = 2% GVF = 4.8% GVF = 9.3% Figure Oil / Low GVF test at nominal liquid flow of 9 l/s Figure 5.12 shows the results obtained at a nominal flowrate of 9 l/s. At 1% GVF the deviation was between -.7 and -2%. At 2% GVF the deviation has increased to between -1 and -2.6%. At 4.8% GVF the deviation has increased to between -3.5 and 4.5%. At 9.3% GVF the deviation has increased to between -6.7 and 8.9%. Report No: 35/99 Page 29 of 59

31 Panametrics DF868 Transit Time Total Vol. Error (%) GVF = % GVF =.5% GVF = 1% GVF = 1.9% GVF = 4.5% Figure Oil / Low GVF test at nominal liquid flow of 18 l/s Figure 5.13 shows the results obtained at a nominal flowrate of 18 l/s. At.5% GVF the deviation was +1%. At 1% GVF the deviation has reduced to -.65%. At 1.9% GVF the deviation has increased to between -.65 and +.9% and at 4.5% GVF the deviation has again increased to between 2.5 and +1.6%. The scatter of results at % GVF was.5%. Panametrics DF868 Transit Time Total Vol. Error (%) GVF = % GVF =.5% GVF =.9% Figure Oil / Low GVF test at nominal liquid flow of 4 l/s Figure 5.14 shows the results obtained at a nominal flowrate of 4 l/s. At.5% GVF the deviation is less than -1%. At.9% GVF the deviation has increased to between +1.4 and -2.9%. The meter was also tested at 1.9 and 4.9% GVF. However reviewing the meter diagnostics showed that at these GVF s the meter had failed and was operating in a error hold mode. The meter was continually outputing a signal corresponding to the last acceptable transit time measurement before signal failure. Hence to avoid confusion these results have not been shown. Report No: 35/99 Page 3 of 59

32 Panametrics DF868 Transit Time Total Vol. Error (%) GVF = % GVF =.1% GVF =.5% Figure Oil / Low GVF test at nominal liquid flow of 55 l/s Figure 5.15 shows the results obtained at a nominal flowrate of 55 l/s. At.1% GVF the deviation is less than +1.6%. At.5% GVF the deviation has increased to between +1.5 and 1.7%. The results at 1.9 and 4.6 % GVF are not shown as in these conditions the meter was operating in error hold mode. 3 Panametrics DF868 Transit Time 2.5 Total Vol. Error (%) GVF = % GVF =.1% Figure Oil / Low GVF test at nominal liquid flow of 8 l/s Figure 5.16 shows the results obtained at a nominal flowrate of 8 l/s. At.1% GVF the deviation was between +.4 and -1.8%. At.4% GVF and above the meter diagnostics indicated a weak signal strength and displayed erratic time difference measurements. These results are not shown as the meter was operating in an error hold mode. Report No: 35/99 Page 31 of 59

33 Panametrics DF868 Transit Time Total Vol. Error (%) GVF = % GVF =.1% Figure Oil / Low GVF test at nominal liquid flow of 91 l/s Figure 5.17 shows the results obtained at a nominal flowrate of 91 l/s. At.1% GVF the deviation was between -.25 and -2.5%. The results at.4,.9 and 1.8% GVF s are not shown as the meter was operating in error hold mode. 5.4 Panametrics DF868 Test Results (Transflection Mode) As the transflection technique requires the presence of a reflector in the fluid, in this case gas bubbles, conditions at or about 2% GVF have been taken as the reference against which deviations are stated. The specific reference condition for each flowrate is detailed in each test condition. 2 Panametrics DF868 TransFlex 1 Total Vol. Error (%) GVF = % GVF = 2% GVF = 2.3% GVF = 39.4% -3-4 Figure Oil / Low GVF test at nominal liquid flow of 5 l/s Figure 5.18 shows the results of the oil / low gas volume fraction tests at a nominal flowrate of 5 l/s. The high error level for point 1 at 2% GVF can be ignored as it seems likely that the point was recorded before the meter achieved a stable condition. The meter was operating in a statistics mode, which allows the meters averaging function to be varied to produce a smooth but responsive output. As this point was recorded immediately after the 18 l/s conditions it seems likely that the meter response had not stabilised following the change of conditions. At % GVF the deviation from the 2% GVF reference condition was between Report No: 35/99 Page 32 of 59