magnetic resonance multiphase flowmeter: measuring ... · 2/26/2015 · applications. one of the...

Copyright 2015, Letton Hall Group. This paper was developed for the UPM Forum, 25 – 26 February 2015, Houston, Texas, U.S.A., and is subject to correction by the author(s). The contents of the paper may not

necessarily reflect the views of the UPM Forum sponsors or administrator. Reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Letton Hall Group is

prohibited. Non-commercial reproduction or distribution may be permitted, provided conspicuous acknowledgment of the UPM Forum and the author(s) is made. For more information, see www.upmforum.com.

1. Introduction This paper describes a magnetic resonance based

multiphase flow meter. During the last decades, magnetic resonance technology has enabled many industrial applications. One of the most well-known developments can be found in medical imaging. High resolution, three dimensional body scans can provide crucial information for medical diagnosis with many innovations being published every year. Another important field of application is the oil and gas industry. For many years, magnetic resonance based logging tools have been successfully used for detailed formation evaluation [Ref. 1]. A scan of the vicinity of the bore hole provides important information about the reservoir rock structure and saturation condition with pore fluids, which is very important for oil production optimization.

Magnetic Resonance (MR) has specific advantages that can be used for multiphase flow measurement as well. A motivation of these advantages is presented in this paper. After a brief explanation of the principle of MR, a description is given how this principle has been applied in a multiphase flow meter. The conceptual design of the MR-based flow meter has been published previously [Ref. 2-5]. Significant efforts have been made to translate the conceptual prototype to an industrial design that meets oil-field requirements and specifications. A detailed description of this design is given in this paper too.

Finally, this paper summarizes the experimental and calibration results obtained during extensive three-phase testing in various industrial multiphase flow loops.

The paper finishes with a summary and conclusions.

2. Principles of magnetic resonance By using the principle of Magnetic Resonance (MR),

experiments are performed on the nuclei of atoms, by exposing them to a strong magnetic field. The strength of the magnetic field used in this technology depends on the application involved. Chemical applications utilize magnetic field strengths that are up to one million times stronger than the Earth’s magnetic field measured at the equator. The magnetic field applied in the flow meter described in this

paper is approximately a thousand times stronger than the Earth’s magnetic field.

MR is a consequence of the intrinsic magnetic moment of protons and neutrons. For most atoms, such as

12C and

16O, the individual magnetic moments of protons and

neutrons offset each other, and the effective magnetic moment of such nuclei vanishes. Those atoms are invisible to MR. Other nuclei possess either an odd number of protons (

1H), or neutrons (

13C), or both protons and

neutrons (2H). If those atoms have a sufficiently high natural

abundance, they can be utilized in MR experiments. Hydrogen protons (

1H) (Fig. 1) provide the strongest MR

response and are targeted in most oil-field applications of magnetic resonance [Refs. 6 and 7]. Another advantage of 1H is that it has a very high natural abundance (99.985%).

Figure 1: 1H atom. The atom consists of a proton and an

electron. The spinning proton has a magnetic moment, .

2.1. Precession / Resonance Frequency. Similar to

bar magnets, when a magnetic nucleus is placed between the poles of an external magnet, it too will try to align itself with respect to the externally applied magnetic field. The static magnetic field exerts a torque on the axis of the spinning magnetic moments, which, in turn, causes the axis to move perpendicular to the direction of the applied torque, and, hence, around the direction of the background magnetic field (Fig. 2). This

UPM 15060

Magnetic Resonance Multiphase Flowmeter: Measuring Principle and Multiple Test Results Jankees Hogendoorn, André Boer and Mark van der Zande, Krohne; Rick de Leeuw and Piet Moeleker, Shell Global Solutions International B.V.; Mathias Appel and Hilko de Jong, Shell International Exploration and Production inc.

Upstream Production Measurement Forum 2015 2

motion is referred to as Larmor precession and is analogous to the motion of a spinning top in the Earth’s gravitational field.

Figure 2:

1H proton. Left-hand figure: no external

static magnetic field. Right-hand figure: with external static magnetic field; the

1H proton starts to precess

along the magnetic field direction with the Larmor frequency, f0 .

The Larmor frequency, f0 , of this precession around

the direction of the background magnetic field, 0B

, is

determined by the gyromagnetic ratio, :

00

2Bf

………………………….(Eq. 1)

The gyromagnetic ratio is a material constant. Commercial logging tools, as well as the magnetic resonance flow meter, utilize permanent magnets of such strength that aligned protons precess with frequencies between several hundreds of kHz to a few MHz.

2.2. Magnetization. In the macroscopic world, magnets

can be aligned in an infinite number of orientations (Fig. 3, left-hand side). At the atomic level, however, only distinct alignments of the nuclei relative to the direction of an external magnetic field (also called spin states) are possible (Fig. 3, right-hand side). That is because direction and

magnitude of the angular momentum of a nucleus, I

, are quantized, i.e., they can only vary by integer multiples of

(Planck’s constant, h , divided by 2 ).

For 1H, two distinct alignments are possible. The nuclei

can be aligned at a certain angle either with or against the external magnetic field. The population ratio of these two spin states is given approximately by the ratio of magnetic to thermal energy: at room temperature, there is approximately the same number of proton nuclei aligned with the main magnetic field as counter aligned. The aligned position is slightly favored, as the nucleus is at a lower

energy state in this position. For every one million nuclei,

there is about one extra nucleus aligned with the 0B

field as

opposed to the field. This phenomenon creates a net or macroscopic magnetization, in this paper subsequently

referred to as M

, pointing in the direction of the main magnetic field (z-direction). This is illustrated in the right-hand picture of Fig. 3.

The magnetization build-up is time-dependent. For a single fluid, this build-up of magnetization can be characterized by a time constant called longitudinal relaxation time, T1, as shown in Fig. 4. For mixtures of fluids, or more complex fluids, the magnetization build-up is typically characterized by a spectrum of longitudinal relaxation times. Short T1-times result in fast magnetization build up.

Figure 4: Time-dependent magnetization build-up. For a single-phase fluid, the magnetization builds-up exponentially

and can be characterized by the so-called longitudinal

relaxation time, T1.

Figure 3: Left-hand figure: no external magnetic field; random orientation of protons, no net magnetization.

Right-hand figure: external magnetic field; protons start to precess with the Larmor frequency.


2.3 Radio-Frequency Pulses. The alignment of magnetic moments can be disturbed by radio-frequency (RF) pulses transmitted from an antenna to the fluids measured with the device (Fig. 5, left-hand figure). As a consequence,

the orientation of the macroscopic magnetization, M

, can be modified by applying radio-frequency pulses with the appropriate intensity, duration, and frequency. Changes in the orientation or intensity of the macroscopic magnetization can be detected as a small voltage in an appropriate RF coil. It is this voltage which is measured as the MR signal in magnetic resonance experiments.

2.3.1 Hahn-echo. A well-known pulse experiment is the

so-called Hahn-echo (spin-echo) experiment [Ref. 6]. An RF pulse is created with the intensity and duration such that

the magnetization vector, M

, is tilted along the x-axis by 90° (from the z-orientation to the xy-plane). This pulse is called a P90 pulse. After this pulse, the magnetization vector is rotating with the Larmor frequency in the xy-plane.

However, due to spatial inhomogeneity in the magnetic field strength (leading to slightly differing Larmor frequencies of protons located in different positions), some protons lag behind and others move ahead during their Larmor precession. This is illustrated in Fig. 5b. As a consequence, the phase coherence slowly disappears, which, in turn, diminishes the magnetization vector and leads to a fast decaying signal. By applying a P180 pulse at

t=, all protons are flipped along the y-axis by 180°. As a result, the slower moving protons are placed ahead and the faster moving protons are placed behind. Since the individual resonance frequency is unchanged, all protons

start to re-phase (Fig. 5c), leading to an echo at t=2 (Fig. 5d). This echo is called the Hahn echo.

2.3.2 CPMG pulse sequence. Hahn echoes can be created multiple times by repeatedly applying P180 pulses. This pulse sequence is called the CPMG pulse sequence and is illustrated in Fig. 6. The name CPMG stands for the inventors of this sequence: Carr, Purcell, Meiboom and Gill [Ref. 7]. By varying the time delay between subsequent P180 pulses, TE, different physical and chemical processes can be studied. Inter-echo spacings between tens or hundreds of microseconds to several milliseconds are typically applied during the CPMG sequence. Measuring several thousand echoes in one run is not uncommon.

Figure 6: Illustration of the CPMG pulse sequence,

creating multiple echoes.

During a CPMG sequence, the amplitudes of succeeding

echoes are slightly attenuated relative to the previous echo because the reversal of the direction of the Larmor precession by a P180 pulse does not perfectly compensate for differences in the Larmor frequencies of different spins. Therefore, the signal envelope of all echoes acquired during

Figure 5: Hahn echo experiment. Left-hand figure: A P90 RF pulse flips the magnetization vector Mz in the horizontal xy-plane. Right-hand figure: a) P90 pulse b) de-phasing signal mainly due to differences in the local magnetic field, c) re-phasing signal

after P180 pulse at t=, d) echo at t=2.


the CPMG sequence eventually decays to noise level. This signal decay describes the return of the magnetic moments to their equilibrium state prior to the disturbance by RF pulses and is characterized by the so-called transverse relaxation time, T2. This is shown in Fig. 7.

Figure 7: The decay of the spin echoes (blue line) is characterized by the transverse relaxation time, T2.

In weak, homogenous magnetic fields, and for simple,

symmetric molecules, the T1- and T2-relaxation times of fluids are similar, since both are governed by the molecular structure of the fluid (such as inter- and intra-molecular distances between the hydrogen atoms) as well as the ratio of viscosity to temperature [Ref. 8].

Besides CPMG, a variety of RF pulse sequences have been developed for measuring T1 or T2. By varying the timing of these pulse sequences, the measurement can be optimized to the expected MR response of the flowing fluids.

3 Magnetic resonance for multiphase flow measurement.

Conceptually, a magnetic resonance experiment usually consists of three sequential steps:

1. Creating a net magnetization by aligning the magnetic moments of hydrogen atoms in an applied, constant magnetic field B0.

2. Perturbing the alignment of hydrogen atoms by employing electro-magnetic radio frequency (RF) pulses.

3. Detecting the radio-frequency signal emitted by the hydrogen atoms during their return to equilibrium alignment in the external magnetic field.

This measurement principle also forms the basis of magnetic resonance based flow metering.

Figure 8: Simplified drawing of the magnetic resonance multiphase flow meter. The flow meter comprises two major

components: the pre-magnetization section and the measurement section.

The flow meter comprises two major components: a pre-

magnetization section and a perturbation and measurement section involving a radio-frequency coil (figure 8).

3.1 Water Liquid Ratio Determination. In the pre-

magnetization section, the hydrogen atoms are magnetized and a net macroscopic magnetization is created. Fig. 4 shows the magnetization of the protons as function of time. In the flow meter, magnetization as function time can be transformed to magnetization as function of distance to the start of the pre-magnetization section by substituting the simple relation t=x/v where x is the position within the pre-magnetization section and v is the flow velocity. The magnetization as function of x for a specific velocity and longitudinal relaxation time, T1, is shown in Fig. 9.

Figure 9: Magnetization build-up as a function of position inside the pre-magnetization section for a specific velocity

and T1. (v= 2[m/s], T1 oil =0.15s).

An important property of oil and water, which is exploited in the MR multiphase flow meter, is the difference in longitudinal relaxation time, T1. For most oil viscosities encountered in the industry, the longitudinal relaxation

Pre-magnetization

section Measurement

section

RF coil

Flow


time, T1, is significantly shorter than that of water. As a consequence, oil magnetizes faster than water (see Fig. 10, left-hand figure). This difference in the rate of magnetization build-up is used to create contrast between oil and water. By varying the effective length of the pre-magnetization section, the signals originating from oil and water, respectively, will be built-up to different levels, leading to quantitative information on the water-oil ratio. This process is illustrated in Fig. 10. In the left-hand figure, the full pre-magnetization length (labeled as configuration 111) is used. For the particular flow velocity displayed, this length is sufficient to achieve complete polarization of the oil signal, whereas the signal originating from water has been built up to approximately 30%. Consequently, the oil-water signal ratio (S111oil/S111water) in the measuring section adjacent to the pre-magnetization section is about 3. When the pre-magnetization length is reduced to just the length of the measuring section magnet (labeled as configuration 000), both the oil- and water signals are reduced but the signal from water is more strongly attenuated than the oil signal. This leads to an oil-water signal ratio (S000oil/S000water) of about 10 (right-hand side).

Figure 10 demonstrates that a strong contrast can be created between the signals originating from oil and from water based on the difference in T1 times by varying the pre-magnetization length. This contrast enables precise quantification of the oil-water ratio even for very high water liquid ratios. This capability is illustrated in Fig. 11. In this figure, the signal ratio S000/S111 of both oil (red) and water (blue) are plotted as a function of flow velocity.

Figure 11: Signal ratio S000/S111 as function of flow velocity

for various oil-water ratios. The signal intensities have been calculated assuming longitudinal relaxation times of 0.15 s

for oil and 2.3 s for water. A strong signal contrast between oil and water can be achieved even for high flow rates.

For multiphase flow of oil and water, the S000/S111 curve

will always fall between the pure oil and pure water curves. The shape of curves for different water liquid ratios can be calculated and has been added to Fig. 11 for water liquid ratios of 10%, 50% and 90%, respectively. Figure 11 also shows that the sensitivity of the MR flow meter for determining the water liquid ratio (WLR) increases with increasing water liquid ratio, demonstrated by an increased

Figure 10: Build-up of magnetization and, hence, signal levels, for oil and water achieved with maximum pre-magnetization length (left-hand figure) and minimum pre-magnetization length (right-hand figure). The magnetization build-up has been calculated for a flow velocity, v, of 2 m/s and assuming longitudinal relaxation times, T1, of 0.15 s and 2 s for oil and water,

respectively.


separation of the water liquid ratio lines with increasing water liquid ratio. This method remains robust even when phase slip between the oil and water phases is present.

The water liquid ratio determination method is largely independent of the gas volume fraction (GVF) up to high GVF’s because the gas signal is weak compared to the liquid signal. Once the gas fraction has been determined, as discussed in the next paragraph, the gas signal can be subtracted from the respective signals S000 and S111. This process produces liquid-only MR signals, which are used for water liquid ratio determination

3.2 Gas Hold-Up Determination. The MR flow meter is

capable of directly measuring the gas hold-up as well as the gas phase velocity. The underlying principles will be explained in this and subsequent paragraphs.

Assume a simplified two phase flow situation comprising a gas-liquid mixture. For this system holds that the sum of the liquid and gas hold-ups equals unity:

1 GL ……………………………(Eq. 2)

where L and G are the hold-ups for liquid and gas, respectively. Furthermore, we know that the measured signal, Smeas., is a superposition of the signals generated by the liquid and gas fractions, respectively:

.%100%100 .. measGGLL SSS ……………..(Eq. 3)

where S100%L and S100%G are the signal amplitudes corresponding to 100% liquid and gas filling, respectively. These particular-signals amplitudes are determined as part of the calibration procedure for each meter.

Equation 2 in combination with Eq. 3 can be written as:

GL

measL

GSS

SS

%100%100

..%100

………………………….(Eq. 4)

Smeas., S100%L and S100%G are known. Consequently, G can be calculated. Considering the strong contrast between the liquid and gas signal amplitudes due to their contrast in hydrogen density, this methodology is a very robust and reliable way to accurately determine the gas hold-up. This approach can be extended for the situation in which the liquid phase is a composition of oil and water.

3.3 Velocity Determination. The flow velocity is determined by using the so-called convective decay method. In the measurement section, the alignment of hydrogen

atoms relative to the direction of the external magnetic field is perturbed by irradiating RF energy using a RF coil and a pre-defined pulse sequence, such as a CPMG sequence introduced earlier. The spin echoes produced by the sample are detected by the same RF coil also used for irradiating the RF energy. The amplitude of the initial Hahn echo has the highest value since all excited protons are still inside the RF coil. In contrast, when the second Hahn echo is formed about 1 ms later, a certain fraction of excited protons has already flown out of the RF coil. Consequently, the amplitude of the second echo will be somewhat lower than that of the first echo, even if there was no transverse relaxation. All following echoes are successively attenuated, since more and more protons that were excited by the initial RF pulse have left the RF coil. This ‘convective decay’ of the echo signals is shown in Fig. 12.

Figure 12: Velocity determination using the so-called convective decay method. Ignoring magnetic relaxation

effects, the measured echo amplitude decreases linearly with time since the excited protons are leaving the RF coil due the

flow. At a certain time all excited protons have left the RF coil. The RF coil length divided by the time at which the red

dashed line intercepts the horizontal axis yields the flow velocity.

The signal decay is proportional to the flow velocity. The

higher the flow the faster the echo decay. The flow velocity is equal to the length of the RF coil, Lc (typically 10 cm), divided by the time at which the red dashed curve intercepts the x-axis, tS=0, so v = Lc/tS=0 [Ref. 9].

Combining the convective decay method with varying the pre-magnetization length enables the direct measurement of velocity slip between the flowing liquid phases. As shown in Fig. 10, signal contrast can be created between water and oil by modifying the effective pre-magnetization length. For a long pre-magnetization length, the initial signal, S0, in Fig. 12 is a composition of signal contributions originating from both oil and water.


The velocity that is obtained from the convective signal decay is a composition of the oil and water velocities. In contrast to this, the initial signal amplitude corresponds predominantly to oil if a short pre-magnetization length is selected, cf. Fig. 12. Consequently, the velocity that is being measured for this pre-magnetization configuration predominantly reflects the oil velocity. This, in combination with the measured water liquid ratio, makes it possible to determine both the oil and water velocities independently.

3.4 Composition And Velocity Profile Velocity

Determination. Similar to Magnetic Resonance Imagers developed for medical applications, the MR multiphase flow meter is capable of producing an image of the spatial distribution of hydrogen protons inside the pipe.

The resonance frequency of detecting a MR signal is directly proportional to the strength of the applied magnetic field, B0.

By adding a linear magnetic field gradient, Gz, in the vertical direction, see Fig. 13, the magnetic field strength varies linearly inside the pipe and now becomes a function of height, as indicated by the blue shaded area. Consequently, the resonance frequency of the protons, f, becomes a function of the height inside the pipe, z.

By means of an RF frequency band selection, only protons in the corresponding resonance frequency band are read out, effectively leading to a slice-selective measurement. The approach to determine the WLR, gas hold-up and velocity, as just discussed, could be applied, but

now in a spatially selected slice. Using this spatial information, the fluid composition as well as the velocity can be determined as a function of height. An example of this data acquisition is shown in Fig. 14.

Figure 13: Application of a linear magnetic field gradient in

the vertical direction, z, makes the proton resonance frequency to become a function of height inside the pipe. By means of an RF frequency band selection, only protons in the

selected slice can be measured. It should be noted that the flow is in the x-direction.

4 Mechanical design of the MR multiphase flow meter

The current flow meter is operated in a horizontal configuration. The design is an industrialized version of the improved and tested prototype version and has a full bore pipe. No specific upstream flow conditioning measures are required. The connecting flanges are made from stainless

Figure 14 - The lower left-hand figure shows the composition profile for the multiphase stratified wavy flow as shown in the upper figure. In the right-hand lower figure the corresponding time averaged velocity profile is depicted.

(Gas: 5.8 (a)m3/h, Oil: 1.6 m

3/h, Water: 6.4m

3/h, WLR: 80%, GVF: 43%).


steel (SS) or carbon steel (CS). The overall flange-to-flange length is about 3m. Maximum process temperature is 100°C and the minimum and maximum ambient operating temperatures range from -40°C to +65°C. All electronics are mounted directly on the flow meter in two flame-proof boxes (see Fig. 15).

An important feature of the MR flow meter is that there are no protrusions into the pipe. Inside the flow meter, the transition is made from steel flanges to a glass fiber reinforced epoxy (GRE) pipe (see Fig. 16). A non-conductive pipe section is required to allow high frequency electromagnetic waves to be transmitted into the pipe. All measurements are made from outside of the tube. There is no feed-through or anything else puncturing the tube.

The selected GRE pipe and flanges are frequently used in the oil and gas industry and possess all required certifications and approvalsThe forces from the connected piping system are guided through the external housing in order to prevent mechanical loading of the GRE pipe.

The magnetic field strength in the main magnet is equal to the magnetic field strength in the pre-magnetization section. However, the magnetic field homogeneity of the main magnet is orders of magnitude higher than the homogeneity of the pre-magnetization section. The high field homogeneity is necessary for proper proton resonance measurements.

In contrast to the magnetic field strength inside the pipe, the magnetic field strength outside the meter housing is

Figure 15: External mechanical design of the industrialized version of the MR multiphase flow

meter. All electronics are mounted on the flow meter in explosion-safe boxes. The overall length is about 3 m.

Figure: 16 The internal mechanical design of the industrialized MR multiphase flow meter. The

flow meter has a full bore design.


very weak (< 0.5 mT). This level is below the maximum field strength legally specified by standards. The value 0.5 mT is about 10 times stronger than the Earth’s magnetic field strength. In between the main magnet and the GRE pipe, the radio frequency coil is located. The RF coil is transmitting and receiving the signals and is matched and tuned to the Larmor precession of flowing phases by a matching and tuning circuit mounted in the electronics housing. 5 Brief description of the various multiphase test facilities

The magnetic resonance flow meter has been tested at four different flow loops (see Fig. 17). With the exception of the single phase (water) testing at the XCaliber loop in Dordrecht, NL, all flow loops have been visited a number of times with the various generations of the MR flow meter:

Multiphase flow loop of Southwest Research Institute (SwRI), San Antonio, TX, USA.

Single phase (water) flow loop (XCaliber loop), Dordrecht, NL.

Multiphase flow loop of Shell (Donau loop), Rijswijk, NL.

Multiphase flow loop of DNV-GL, Groningen, NL.

The test conditions of each flow loop are briefly discussed.

5.1 SwRI Test Facility, San Antonio, March 2013. Fresh water, oil (Regal R&O 32 VSI, viscosity about 43 cSt @ 32°C) and methane have been used as test fluids. The operating static pressure and temperature were 82.7 bar(g) and 32°C respectively. The Gas Volume Fraction (GVF) versus Water Liquid Ratio (WLR), as well as gas and liquid flow rates are included in Figs. 18 and 19. The flow regimes tested included single phase, stratified, stratified wavy, plug and slug flow. The test results obtained at this loop have been presented before [Refs. 3, 4 and 10].

5.2 XCaliber Flow Loop, Dordrecht, February 2014.

XCaliber is a single phase, fresh water test loop located in Dordrecht, NL. The operating pressure and temperature were 3 bar(g) and 35 °C, respectively. The flow rates have been varied between 15 to 200 m

3/h.

5.3 Shell Donau Multiphase Test Loop, Rijswijk NL,

Feb/Apr/Sept/Oct 2014. The Donau flow loop is using compressed air for the gas phase. The MR flow meter has been tested using oils of three different viscosities (Renolin 10 cSt, 50 cSt and 120 cSt). Saline water (100 g/l NaCl) is used for the water phase. The line pressure for the experiments with the MR multiphase flowmeter was kept at 3 bar(g), and the line temperature at 40 °C. The Gas Volume Fraction (GVF) versus Water Liquid Ratio (WLR), and gas versus liquid flow rates that have been tested are included

Figure 17: Picture of the various flow test loops. Upper left-hand figure: SwRI, San Antonio, USA. Lower left-hand figure:

XCaliber, Dordrecht, NL. Upper right-hand figure: Donau, Shell Rijswijk, NL. Lower right-hand figure: DNV-GL, Groningen, NL.

SwRI (TX, USA)

Xcaliber (NL) DNV–GL (NL)

SwRI (TX, USA) Donau Shell (NL)


in Figs. 18 and 19.

5.4 DNV-GL Test Loop, Groningen, NL, April/June

2014. The DNV-GL flow loop [Ref. 11] comprises a multiphase pump, combined with a separator and storage vessels for the different liquid types. A heat exchanger is used to control the temperature of the loop. The complete loop is at test pressure. The reference flow rates for oil and water were measured with Coriolis meters in different ranges. The gas reference flow is measured using ultrasonic flow meters. A Hysys flow model is linked to the data acquisition system to account for phase transition between the reference meters and the MUT due to temperature and pressure differences. The fluids used in this flow loop are Groningen natural gas (81 vol.% CH4, 14 vol.% N2), Exxsol D120 API 40 (about 4.1 cSt) and saline water (41 gr/l NaCl).

6 Overview of the multiphase flow meter tested range.

The MR multiphase flowmeter has been tested over a wide range of conditions at the four multiphase test facilities. Fig. 18 summarizes the GVF versus WLR test points. This diagram shows that the entire GVF and WLR range has been covered systematically. The increased concentration of test points at higher WLR is related to our particular interest of utilizing the MR flowmeter for these conditions.

Figure 19 shows all test points in the flow map; liquid flow rates versus actual gas volume flow rates. The flow rates for both gas and liquid have been varied across a wide range, covering about two decades of each parameter. Tests at four different pressures have been carried out in the range of 3 to 83 bar(g). The temperature at the various test loops varied from 25°C to 40°C. The salinity varied from 0 g/l (fresh water) up to 100 g/l NaCl concentration. The viscosity of the oil in the multiphase test facilities varied from 4.1cSt (1 cSt on single phase water) up to 43 cSt. An overview of this range is shown in Fig. 20.

7 Overview of the test results

Prior to the measurements at DNV-GL, tests at the Donau flow loop of Shell have been carried out. During these tests at the Donau, the calibration parameters of the MR flowmeter were determined that will be used by the data evaluation algorithm. These parameters have been kept constant for the processing and interpretation of data acquired during the tests at the other flow facilities.

Figure 21 shows the test results that have been obtained at the DNV-GL loop in the flow rate map. In the left-hand figure the results are shown for the tests acquired at 12 bar(g) line pressure. The right-hand figures shows the 30 bar(g) test data. The green squares indicate the reference flow rates. The red circles correspond to the measured values (in accordance with [Ref. 12]).

Figure 18: Overview of all Gas Volume Fraction (GVF) versus Water Liquid Ratio (WLR) test points for the

tests at various multiphase flow loops.

Figure 19: Overview of the liquid flow rate versus gas flow rate test points for all the tests. Actual

flow rates have been used as reference.


A number of points draw attention. At first glance, the 30 bar data looks significantly better than the 12 bar data. For the gas measurements this is indeed true. However, closer analysis shows that this is not true for the liquid data.

For the gas phase, it is clear that a lower pressure leads to a lower gas density. This results in a weaker signal generated by the gas, and thus to a lower signal-to-noise ratio (SNR). Detailed analysis of the test data demonstrates that the errors obtained in the gas measurement results can be attributed to the relatively high level of noise of the measured gas signal. To address this finding, the sensitivity

of the latest version of the MR multiphase flowmeter has been improved by a factor of three. The result of this improvement is a signal-to-noise level which is comparable to the signal quality of 30 bar as presented in the right-hand figure. This hardware improvement should lead to improved results for lower gas pressures.

The liquid data around the value of 20 m3/h in Fig. 21

(left-hand) show a larger deviation between measured and reference values. Further analysis of the measured data, and comparison with video recordings of flow behaviour acquired simultaneously through sight glasses during the

Figure 20: Overview of the test pressure, temperature, salinity and viscosity.

Figure 21: Flow map with an overview of the test results. The green squares indicate the reference flow rates. The red circles indicate the measured flow rates. The left-hand figure shows the results obtained with 12

bar(g) line pressure. The right-hand figure corresponds to 30 bar(g) line pressure.

Result prior to SNR

improvement


flow tests revealed that this larger deviation is related to a mismatch between data acquisition frequency and slug frequency. For these particular flow conditions, the video recording illustrated that the slug frequency is noticeably smaller than the typical measurement interval.

A longer measuring time should lead to better averaging of flow rate and fluid fractions, and consequently to better results. This relation is confirmed by the measurements at 30 bar line pressure with identical actual liquid and gas flow rates. Video recordings demonstrate a more regular slug pattern at 30 bar line pressure, to which the selected data acquisition pattern was better suited.

Figure 22 shows the test results for the 30 bar test conditions in the composition map. The left-hand figure shows the error in the liquid measurements; the right-hand figure the gas results. If the meter’s performance would have been inferior for a specific GVF or WLR range, this shortcoming would have become visible in these diagrams. However, both for the liquid and the gas measurements, there seems to be no specific range in which the error deviates significantly.

For all the test points, the error associated with the gas flow measurement is less than 10% of the MV, even though two of the measured points were outside the operating range for this version of the MR flowmeter. For the error associated with the liquid flow measurement, all test points are within 5% of the MV, noticeably both inside and outside the specified operating range. An exception is the liquid measurement at a GVF of 99%. The uncertainty in the quantification of the 1% liquid fraction flowing in the gas stream is 14.5% MV. When related to total volume flow this reading translates to an error of 0.145%, which is still good.

The results of the liquid measurements acquired at Shell’s Donau flow loop are shown in Fig. 23. Due to the fact that this installation uses compressed air for the gas phase, no gas NMR signal is generated. Because of the absence of hydrogen in compressed air. As a consequence, the gas

velocity cannot be measured. For this reason the ‘measured’ gas flow was set equal to the reference gas flow during data interpretation.

The measured liquid flow rates correspond well with the reference values. Higher GVF’s appear to be associated with larger measurement uncertainties. At the same time, video recordings at these flow conditions show that the flow pattern is very unstable. As discussed in the previous section, this shortcoming will be addressed by selecting longer data acquisition periods, which are expected to lead to an improved precision of the MR flowmeter for these conditions.

Figure 23: Overview of the test results obtained at Shell’s Donau flow loop plotted on a flow map. The green squares

indicate the reference flow rates. The red circles indicate the measured flow rates. Test pressure was 3 Bar(g). As

explained in the text above, the ‘measured’ gas flow rate is set equal to the reference gas flow rate.

Analysis of the data at different water salinities validates that a variation in salinity, and hence, conductivity of the

Figure 22: Overview of the test results in the composition map. The left figure shows the liquid flow measurement error in %MV. The right figure shows the gas flow measurement error in %MV.


water phase, does not have a noticeable effect on the accuracy of the volume flow measurements. This implies that the accurate WLR measurements obtained with fresh water [Ref. 3] have also been confirmed for higher water salinities. The insensitivity of the MR flowmeter to emulsification of the oil and water phases has been confirmed as well.

Compared to the results as presented last year, the accuracy of liquid and gas measurements has been significantly improved with the latest generation of the MR flowmeter. This is illustrated by the cumulative error plot

[12]

as shown in Fig. 24. For liquid flow, the uncertainty has been reduced by a

factor of 3 to 6 compared to earlier results reported. For gas flow, the improvement is significantly higher as increasing the meter’s sensitivity for detecting small signals, as well as further improving the automated data evaluation algorithm, resulted in a reduction of measurement error by approximately a factor of 8. We repeat that the same evaluation algorithm, including constant calibration parameters, has been used for the entire application range (flow, GVF, WLR, salinity, temperature and pressure). 8 Summary and conclusions

This paper provides an explanation of the physical concept of a magnetic resonance based multiphase flowmeter. Utilizing the concept of magnetic resonance imaging, it has – for the first time – be possible to directly quantify the flow of a free gas phase during multi-phase flow at industrial conditions.

The improved and industrialized design of the MR multiphase flowmeter has been tested at four different flow loops over a wide range of conditions of gas and liquid flow rates, GVF, WLR, pressure, salinity and viscosity. This paper provides an overview of the results of these tests.

Progress in both hardware development and the data interpretation algorithm has resulted in a significant improvement in measurement accuracy both for liquid flow (factor of 3 to 6 improvement) and for gas flow (factor of 8) characterization compared to the previous prototype version. The high accuracy of the WLR determination has been proven again by additional measurements. The experiments have demonstrated that a variation in water salinity neither affects the accuracy of WLR, nor the accuracy of the liquid and gas volume flow rate.

The tests have shown that the MR flowmeter’s accuracy of gas and liquid measurement is not dependent on volume flow rate, below a GVF of 0.95. It was realized that the data acquisition period of the flow meter needs to be long enough to eliminate the effect of the characteristic time scale of slugs.

Line pressure starts to affect the accuracy of gas flow measurements once the pressure is below approximately 10 bar. Analysis of acquired data showed that this uncertainty is related to a relatively high noise level. Further hardware improvement has resulted in a reduction of the data noise level by more than a factor of 3. This leads to a proportionally better performance of the gas measurements at lower pressures. 9 Acknowledgements

The authors of this paper like to express their appreciation to the entire development team for the tremendous amount of work that has been done throughout the last year. Further acknowledgement is made to the various operators of the flow facilities at which the meter has been tested.

Figure 24: Cumulative error plot for liquid and gas measurement. Left-hand figure: results obtained at SwRI in 2013 at 83 bar(g). Right-hand figure: result obtained with the improved industrialized MR multiphase

flowmeter obtained in June 2014 at 30 bar(g).

Initial design current design


10 References 1. M. Appel, Nuclear Magnetic Resonance and Formation

Porosity, Petrophysics, Vol. 45, No.3, (May-June 2004); p296-307.

2. M. Appel, et al., Robust Multi-Phase Flow Measurement Using Magnetic Resonance Technology, Society of Petroleum Engineers, MEOS, Manama, Bahrain, 6-9 March, 2011.

3. Jankees Hogendoorn, et al., Magnetic Resonance Technology, A New Concept for Multiphase Flow Measurement, 31

st International North Sea Flow

Measurement Workshop, Tønsberg, Norway, 22.-25. Oct. 2013.

4. Jankees Hogendoorn, et al., Magnetic resonance technology: An innovative approach to measure multiphase flow, 9

th North American Conference on

Multiphase Technology, Banff, Canada, 11.-13. June, 2014.

5. Jankees Hogendoorn, et al., Magnetic Resonance Multiphase Flow Meter: Gas Flow Measurement Principle and Wide Range Testing Results, 32

nd

International North Sea Flow Measurement Workshop, St. Andrews, Scotland, 21.-24. Oct. 2014.

6. E. L. Hahn, Spin Echoes, Phys. Rev. 80 (1950) 580. 7. P.T. Callaghan, Principles of Nuclear Magnetic

Resonance Microscopy, Clarendon Press, Oxford, 1991. 8. N. Bloembergen, E. M. Purcell, R. V. Pound,

Relaxation Effects in Nuclear Magnetic Resonance Absorption, Phys. Rev. 73 (1948) 679-712.

9. T.M. Osán, et al. : Fast measurements of average flow velocity by Low-Field

1H NMR, Journal of Magnetic

Resonance (1969) 209(2):7, 2011. 10. Jankees Hogendoorn and Matthias Appel, Magnetic

Resonance for the Future; A New Methodology to Measure Multiphase Flow, 4th International EMBT Conference, Hannover, 20.-21. March, 2013.

11. A.F. van den Heuvel, D.van Putten and B. Bergsma, Are Multiphase Flow Meters ready for calibration?, Milestones in Metrology, Venice, Italy, 2012.

12. Handbook of Multiphase Flow Metering, Norwegian Society for Oil and Gas Measurement, revision 2, March, 2005.

magnetic resonance multiphase flowmeter: measuring ... · 2/26/2015 · applications. one of the...

Documents