TORSION WAVE BASED FLUID DENSITY MEASURING DEVICE AND METHOD

Abstract

A fluid density measuring device uses a pipe with a pipe wall that has an inner wall surface with a non-circular cross-section at least in an axial segment of the pipe. Preferably, the inner wall surface comprises one or more protrusions extending inward into the pipe and along the axial direction of the pipe. An ultrasound transducer located on the pipe wall is used to generate local motion of the pipe wall with a circumferential direction of motion. Preferably, the ultrasound transducer is located between successive protrusions. An ultrasound receiver located on the pipe wall receives an ultrasound torsion wave generated by said local motion after the torsion wave has traveled through the axial section wherein the inner wall surface has a non-circular cross-section. The fluid density is determined from the propagation speed of the torsion wave.

Claims

1. A fluid density measuring device, comprising a pipe, comprising a pipe wall that circumferentially surrounds an inner space for fluid in the pipe, the pipe wall in at least an axial segment of the pipe having an inner wall surface that has a non-circular cross-section with a virtual plane perpendicular to the axial direction of the pipe; a transducer located on the pipe wall, directed to generate local motion of the pipe wall in a circumferential direction; a receiver located on the pipe wall, configured to receive a torsion wave generated by said local motion after propagation through said axial segment.

2. A fluid density measuring device according to claim 1, wherein the transducer and the receiver are an ultrasound transducer and an ultrasound receiver respectively, and wherein the torsion wave is an ultrasound torsion wave.

3. A fluid density measuring device according to claim 1, wherein the inner wall surface comprises a protrusion that has a convex cross-section part in the cross-section with the virtual surface and extends in the axial direction along said axial segment of the pipe.

4. A fluid density measuring device according to claim 3, wherein the inner wall surface comprises a plurality of protrusions at successive circumferential positions, separated by inter-protrusion parts of said space, the inter-protrusion parts of said space being open to a central part of said space.

5. A fluid density measuring device according to claim 4, wherein the protrusions are a plurality of evenly spaced fins extending axially and radially inward from outermost parts of the inner wall surface.

6. A fluid density measuring device according to claim 1, wherein the pipe wall has an outer wall surface that has a substantially circular cross-section with said virtual plane along said axial segment of the pipe.

7. A fluid density measuring device according to claim 4, wherein the transducer is located on the inner wall surface, located between a circumferentially successive pair of the protrusions.

8. A fluid density measuring device according to claim 7, comprising transducers, preferably ultrasound transducers, located on the inner wall surface between each circumferentially successive pair of the protrusions.

9. A fluid density measuring device according to claim 1, wherein said axial segment includes an axial location of the transducer and an axial location of the receiver.

10. A fluid density measuring device according to claim 9, comprising a further receiver, preferably a further ultrasound receiver, located on the pipe wall at an axial location between the axial locations of the transducer and the receiver, and a control circuit configured to measure a delay between reception of a torsion wave from the transducer by the receiver and the further receiver.

11. A fluid density measuring device according to claim 1, comprising a sound reflector, or an ultrasound reflector, in or on the pipe wall, configured to reflect at least a torsion wave travelling through the pipe wall, wherein the axial segment includes an axial location of the reflector and the axial location or axial locations of the receiver and the transducer.

12. A fluid density measuring device according to claim 11, comprising a transceiver that forms both said transducer and the receiver.

13. A fluid density measuring device according to claim 11, comprising a further sound reflector, or an ultrasound reflector, located in or on the pipe wall, the axial location of the receiver being located at the further reflector or between the further reflector and the reflector.

14. A fluid density measuring device according to claim 13, comprising a control circuit configured to measure a delay between arrival of respective different reflections of a torsion wave from the reflector.

15. A fluid density measuring device according to claim 1, wherein the pipe wall in a further part of the pipe wherein the inner wall surface has a substantially circular cross-section, the fluid density measuring device comprising a control circuit configured to measure traveling speed of a torsion wave in both the part of the pipe wherein the inner wall surface has a non-circular cross-section and the part of the pipe wherein the inner wall surface has a substantially circular cross-section.

16. A fluid density measuring device according to claim 1, comprising a control circuit coupled to the receiver and the transducer, configured to cause the transducer to excite a torsion wave in the pipe wall, to receive a response to arrival of the torsion wave from the receiver and to determine wave propagation delay of the torsion wave through the pipe using the response by the receiver.

17. A fluid density measuring device according to claim 16, wherein the control circuit is configured to determine wave propagation delays of torsion waves travelling in opposite axial directions through the pipe wall and to determine an indication of average torsion wave speed from the propagation delays of the torsion waves travelling in the opposite axial directions.

18. Use of the fluid density measuring device of claim 1, in an oil well to monitor a fluid density of a mixture of oil and water flowing through the pipe.

19. A method of measuring fluid density of fluid within a pipe, comprising generating local circumferential motion of a pipe wall of the pipe by applying vibrations with a transducer located on the pipe wall; receiving a torsion wave generated by said local motion after the torsion wave has traveled through an axial segment of the pipe wherein an inner wall surface of the pipe has a non-circular cross-section with a virtual plane perpendicular to the axial direction of the pipe, the torsion wave being received by means of a receiver located on the pipe wall, the inner wall surface preferably comprising one or more protrusions extending inward into the pipe and along the axial direction of the pipe; determining the fluid density from a measured propagation speed of the torsion wave.

20. A method according to claim 19, wherein the transducer and the receiver are an ultrasound transducer and an ultrasound receiver respectively, and wherein the torsion wave is an ultrasound torsion wave.

Description

BRIEF DESCRIPTION OF THE DRAWING

[0024] These and other objects and advantageous aspects will become apparent from a description of exemplary embodiments with reference to the following figures.

[0025] FIG. 1 schematically shows a fluid density measuring system

[0026] FIG. 2a-d show examples of cross sections of a pipe in a fluid density measuring system

[0027] FIG. 3 shows an embodiment of part of a fluid density measuring system

[0028] FIG. 3a shows a flow chart of density measurement

[0029] FIG. 4 shows an embodiment of part of a fluid density measuring system

[0030] FIG. 4a shows a flow chart of density measurement

[0031] FIG. 5 shows an embodiment of part of a fluid density measuring system

[0032] FIG. 6a,b show use of shared transceivers

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS

[0033] FIG. 1 schematically shows a fluid density measuring system, comprising a pipe 10, a transducer 12 and a receiver 14 attached to pipe 10, and a control circuit 16 coupled to transducer 12 and receiver 14. Control circuit 16 may comprise a programmable computer, programmed with a program with instructions to cause transducer 12 to excite local torsion in pipe 10 and to receive wave measurements from receiver 14.

[0034] Pipe 10 may be located in an oil well, where its main function is to transport well fluid, e.g. a mixture of oil and water, and the fluid density measuring system measures the density of well fluid flowing through in pipe 10. Pipe 10 comprises a pipe wall that circumferentially surrounds a space for containing the fluid. The space may be open at the axial ends of pipe 10 to enable fluid flow. Pipe 10 is cylindrical and has a non-circular inner cross-section. That is, the pipe wall has an inner surface and the cross-section of the inner surface with a virtual plane perpendicular to the axial direction of pipe 10 is non-circular. Obviously no manufactured pipe will have a perfectly circular inner cross-section. But the inner pipe is intentionally non-circular, i,e, it deviates more from a circular cross-section than due to normal manufacturing tolerance, The cross-section may be the same for all axial positions over a continuous range of axial positions.

[0035] FIG. 2a-c show examples of possible cross sections. Preferably, the selected cross section stays the same over length of the pipe in the axial direction, but alternatively it may vary. In FIG. 2a, the inner wall forms a plurality of radially and axially extending fins that protrude radially inward into the inner space of the pipe. In the cross section, first parts of the inner surface lie along parts of a circular ground shape and these first parts 21 are connected by second parts 20 that protrude radially inward from the circular ground shape. In an embodiment, all second parts extend inward by the same amount, so that the inward most tips of the pipe wall in the second parts lie on an inward circle with a smaller radius than the circular ground shape. In a further embodiment all second parts may have the same shape. The number of second parts in the figure is an arbitrary example. A pipe with sixteen fins may be used for example. In an embodiment, the outer pipe radius is 70 millimeter, the inward circle radius is 38 millimeter, the protruding parts protrude 17 millimeter in the radial direction and are 6 millimeter wide at their base.

[0036] In FIG. 2b, the inner wall forms a plurality of radially and axially extending grooves that protrude radially outward. In the cross section, first parts of the inner surface lie along parts of a circular ground shape and these first parts are connected by second parts that extend radially outward from the circular ground shape. Thus, in this embodiment, the first parts form the top of protrusions that extend radially inwards from the largest radial extent of the inner space of pipe 10. In an embodiment, all second parts extend outward by the same amount. In a further embodiment all second parts may have the same shape. In the illustrated embodiment there are four such grooves, with rectangular bottoms.

[0037] As may be noted, the embodiments of both FIGS. 2a, b effectively show grooves that extend radially outward from an inner circle on which the innermost parts of the inner surface of the pipe wall are located. In other words, the embodiments result in radially inward protrusions from the largest radial extent of the inner space of pipe 10, wherein the protrusions have a convex shape, i.e. where there lines can be drawn between points on the inner surface of the pipe wall that run entirely through the pipe wall rather than through the inner space. The presence of such convex shapes means that fluid on both sides of the convex shape must be displaced in a circumferential direction. Circumferential displacement of the pipe wall requires that at least the fluid in these grooves must be displaced circumferentially. Hence such grooves increase the effect of the fluid on torsion wave propagation. Due to viscosity an amount of fluid of fluid within the inner circle must be displaced circumferentially as well. This increases the effect on torsion wave propagation.

[0038] In FIG. 2c, the inner wall has a rectangular cross-section (e.g. square) this also results in an association of the circumferential displacement of the pipe wall with the displacement of the fluid. Similarly, other polygonal cross sections may be used, e.g. triangular or hexagonal. It may be noted that these do not result in protrusions with a convex shape. It has been found that the embodiments with protrusions of convex shape on the inner wall surface produce a larger effect of the fluid on torsion wave measurement.

[0039] Another issue with the pipe cross-section involves stress in the pipe wall. In high pressure environments such as oil wells attention is needed to avoid failure of the pipe due to excessive peak stresses. This imposes a minimum thickness on the pipe wall. Use of polygonal cross-sections increases the minimum required pipe wall thickness needed to avoid failure, which in turn reduces torsion wave dependence on fluid properties. The required thickness can be reduced by using increased rounding. But in the polygonal cross-sections rounding also reduces the amount of fluid displacement in the circumferential direction. Either way, sensitivity is reduced. The use of protrusions with convex shape makes it possible to realize higher sensitivity with a smaller pipe wall thickness at the locations where the pipe wall thickness is at minimum and hence to more sensitivity.

[0040] Transducer 12 and receiver 14 are designed to excite a time dependent local torsion in pipe 10, for example with a frequency of 40 kHz, i.e. in the ultrasound frequency range of 20 kHz upward. Ultrasound transducers and receivers may be used as transducer 12 and receiver 14. However, for large pipe diameters lower frequencies may be used. Although the frequency is not critical, the optimum frequency is roughly proportional to the inner pipe diameter (of the free part not containing protrusions), e.g. 300 kHz for an inner pipe diameter of 1 cm, but any frequency in a range of 75-525 kHz may work for such a diameter. Ultrasound may be used for pipes with inner diameter of less than about 25 cm. For a pipe with an inner diameter of one meter any frequency in a range of 750 Hz-5 Khz may work.

[0041] FIGS. 2a, d illustrate an embodiment of the location of transducers 12. By way of example FIG. 2a shows an embodiment where a single transducer 12 is shown. FIG. 2c shows a preferred embodiment with a plurality of transducers. But a single or a few transducers may suffice. A transducer 12 may be used located at the bottom of a groove formed between the fins, attached to the inner surface of the pipe wall. As is known per se, transducers act by causing relative displacement of mass, which, because the transducers as attached to the wall surface, involves local movement of the pipe wall relative to a remainder of the pipe wall. A piezo transducer or an electromagnetic acoustic transducer (EMAT) may be used for example. For example piezo transducers of 11 millimeter circumferential length, 30 millimeter axial length and 5 millimeter radial thickness may be used. In the fluid density measuring device transducer 12 is directed so that it displaces mass in a circumferential direction. A temporally periodic displacement may be used that oscillates as a function of time, when transducer 12 is caused to do so by control circuit 16.

[0042] Preferably transducers 12 are located in a plurality of grooves, located at the bottom of the grooves at corresponding axial positions. In this embodiment transducers 12 are preferably caused to oscillate in phase, so that all transducers 12 always locally move the pipe wall in the same circumferential direction at the same time. In this way, wave excitation is concentrated in a lowest order mode that provides for the most straightforward relation between torsion wave propagation speed and fluid density. However, higher order torsion wave modes may be used instead. Transducers 12 located at different axial positions and/or excitation with non-zero phase offset may be used. This may be used to excite higher order modes or the lowest order mode. For example, transducers in different grooves at different axial positions may be excited with phase offsets corresponding to the propagation speed of the lowest order mode. Preferably transducers 12 are located in all grooves. This may be used to optimize the concentration of wave excitation in the lowest order mode.

[0043] A measuring system with transducers 12 on the inside of the pipe wall may involve electrical wiring (not shown) passing to the control circuit through the pipe wall. This may reduce robustness of the pipe against stress when under high pressure. As an alternative transducers 12 may be located on the outward surface of the pipe wall. This avoids increase of stress due to wiring through the pipe wall. It has been found that in the design with convex protrusions with a transducer or transducers in the grooves formed between the protrusions a higher measurement sensitivity can be achieved without excessive peak stresses.

[0044] Preferably, the outward surface of the pipe wall has a circular cross-section, as shown in FIGS. 2a-c. In this way the relative effect of the density of the fluid located on the outside on the torsional wave propagation in the pipe is minimized. The pipe may in turn be surrounded by a casing and optionally further fluid between the pipe and the casing. Use of a circular outer cross-section reduces the effect of the casing and/or further fluid on torsion wave propagation.

[0045] FIG. 3 shows an embodiment wherein a receiver 14 is located in pipe 10 at a different axial position than transducer 12, separated by a pipe section with axial length L. Pipe 10 has a non-circular inner surface cross section (e.g. as shown in FIGS. 2a-c) in the section of pipe 10 along the length L.

[0046] In an embodiment, transducer 12 and receiver 14 are both transceivers that may be used both for exciting and receiving. In this embodiment, the control circuit may be configured to measure wave propagation delay in opposite axial directions along the pipe section of length L. The control circuit may be configured to compute an average wave speed, or more generally a sum of these delays to measure the fluid density. In this way effects of fluid flow in pipe 10 on the measurement may be reduced.

[0047] FIG. 3a shows a flow chart of operation of the control circuit in this embodiment. The control circuit may comprise a stored computer program to control execution of the steps of the flow-chart. In a first step 31, the control circuit causes application of a pulsed displacement exiting signal to transducer 12 or a plurality of transducers at the same axial positions. In an embodiment, a control signal may be used that cause a plurality of oscillations of the displacement at a predetermined oscillation frequency (e.g. 40 kHz) during a pulse time interval. In a second step 32, the control circuit records a time count at the time of transmission, or resets a time count. In a third step 33, the control circuit monitors wave reception amplitude at receiver 14, and detect when a peak reception amplitude occurs, optionally from a filtered signal component at the oscillation frequency. In a fourth step 34, the control circuit determines the difference between the time count at the time at which the peak wave reception amplitude occurred and the time count at the time of transmission. In a fifth step 35, the control circuit uses the difference e.g. by computing an estimated density, or by controlling further equipment (not shown) dependent on the difference.

[0048] Preferably, pipe 10 has the same non-circular inner surface cross section along the entire length of the section of length L. However, it may suffice that pipe 10 has such a cross section only along part of the section of length L. Having the same non-circular cross section along the entire section of length L has the advantage that the effect of the fluid on propagation is maximized and confusing reflections are reduced. Parts of the pipe outside this section need not have such a cross-section: they may have a circular inner surface cross-section. Instead of a single receiver 14, a plurality of receivers may be used at the same axial position. The same goes for transducer 12.

[0049] FIG. 4 shows an embodiment wherein a first and second receiver 40a,b are located in pipe 10. Second receiver 40b is separated from transducer 12 by a pipe section with axial length L. First receiver 40a is located axially between transducer 12 and second receiver 40b, separated by first and second pipe sections of length L1 and L2 respectively (L=L1+L2). Pipe 10 has a non-circular inner surface cross section (e.g. as shown in FIGS. 2a-c) at least in the second pipe section of L2 between the axial positions of first and second receiver 40a, b and preferably over the entire length L. The control circuit may be configured to compute a difference between the times of arrival of torsional waves at first and second receiver 40a, b due to travel over the length L2, after excitation by transducer 12 and propagation over the length L1. This may be used to eliminate the effect of excitation timing on the delay between excitation and reception.

[0050] FIG. 4a shows a flow chart of operation of the control circuit in this embodiment. Similar steps as in FIG. 3a have been labeled with the same number. In a first step 31, the control circuit causes application of a pulsed displacement exiting signal. In a second step 42, the control circuit records signals received by receivers 40a, b during a predetermined time interval or time intervals that extend over expected times of reception of the signal from transducer 12 at receivers 40a, b. In a third step 43, the control circuit computes cross-correlation values between the recorded signals from receivers 40a, b for a plurality of different time shifts of these signals relative to each other. In a fourth step 44, the control circuit determines a time shift value for which the computed cross-correlation has its largest value, optionally of an interpolation of the cross-correlation value for discrete time shift values for which the cross-correlation value has been calculated. In a fifth step 35, the control circuit uses the difference as an indication of the fluid density.

[0051] In an embodiment, transducer 12 and at least second receiver 40b are both transceivers that may be used both for exciting and receiving. In this embodiment, the control circuit may be configured to measure wave propagation delay in opposite axial directions along the pipe sections between first and second receiver 40a,b and between first receiver 40a and transducer 12 acting as receiver. The control circuit may be configured to compute an average wave speed of these measurements or more generally a sum of the delays (weighed if L1 and L2 are not equal). In this way effects of fluid flow in pipe 10 on the measurement may be reduced.

[0052] FIG. 5 shows an embodiment for using wave reflection. In this embodiment receiver 14 and transducer 12 may be located at the same or nearby axial positions, or transducer 12 may act both as receiver and transducer. An ultrasound reflector 50 is located at a different axial position than transducer 12, separated by a pipe section with axial length L. Pipe 10 has a non-circular inner surface cross section (e.g. as shown in FIGS. 2a-c) in the section of pipe 10 along the length L. Reflector 50 may be a transition between non-circular inner surface cross section and circular inner surface cross section of pipe 10, or any other structure that changes the wave impedance or geometry as a function of axial position, such as a weld joint between pipe sections. Use of reflection measurement has the advantage that directional effects on the measurement due to flow are reduced.

[0053] In an embodiment, transducer/receiver 12 may be located at the axial location of a further reflector, or between the further reflector and reflector 50. The further reflector may be realized in any of the ways described for reflector 50. In this embodiment, transducer/receiver 12 will receive multiple reflections of torsion waves traveling back and forth between the reflector and the further reflections. When pulse transmission is used, these lead to reception of successive reflection pulses at receiver 14, which may be numbered in order of reception. The control circuit may be configured to measure the effect of the fluid on the delay between transmission and reception of the nth reflection (n>1). In this way the change of delay due to the fluid can be increased. In an embodiment the control circuit is configured to measure the effect of the fluid on the delay between reception of the nth reflection and the mth reflection (m>n>1). In this way time delay/phase effects due to the electronics, the transmitter/receiver/transceiver can be eliminated from the measurement. When m=n+k with k>1, a plurality of k wave travel delays back and forth may be used to increase sensitivity.

[0054] An operation of the control circuit like that of FIG. 4a may be used, with cross-correlation of the nth and mth received reflection signal instead of cross-correlation of the signals from first and second receiver 40a,b. Instead the deconvolution of a product of the signal at the mth reflection by the signal at the nth reflection may be used.

[0055] Instead of cross-correlation, deconvolution may be used. Deconvolution may be realized for example by computing the Fourier transforms of the recorded signals of the nth and mth reflection from receiver 14, computing and ratios of the Fourier transform of the recorded signal of the mth reflection divided by the Fourier transform of the recorded signal from the nth reflection at corresponding frequencies. The ratios correspond to a deconvoluted signal in the Fourier transform domain. Optionally, an inverse Fourier transform of the frequency dependent signal formed by the ratios may be computed to obtain the deconvoluted signal in the time domain.

[0056] Deconvolution eliminates, or al least reduces the effect of the transfer functions of the electronics, the transmitter and receiver/transceiver and the phase response of the voltage excitation. Because the reflection coefficients of the reflector usually have zero phase, the effect of the reflection on the phase is also removed. The wave speed may be determined from the phase Phi of the deconvoluted signal at the oscillation frequency f, according to the relation c=L2*f/Phi, where c is the wave speed.

[0057] Deconvolution may also be used in the other embodiments. Assuming that the first and second receivers 40a,b of the embodiment of FIG. 4 have the same response functions, use of deconvolution of their signals instead of cross-correlation may be used to eliminates the effect of transmission to the first receiver 40a and the receiver response functions from the measurement. The same holds when the function of first and second receivers 40a,b is performed by a single receiver in combination with a reflector. Deconvolution may make a separate summing step for summing delays unnecessary.

[0058] The control circuit may be configured to translate measured delays into density measurements using calibration data. In one embodiment control circuit may be configured to use a memory and a look-up table stored in the memory to look up density values corresponding to delays. Optionally interpolation may be used between densities corresponding to delays for which calibrated values are available. The look up values for a specific pipe design may be determined by means calibration measurements using fluids of known density.

[0059] In an embodiment wherein the fluid is a mixture of two known fluids, such as a mixture of water and oil, the measured delays may be translated into a quantitative measure of the density ratio of the fluids. A water cut ratio R may be determined, which corresponds to R=(dd1)/(d2d1), wherein d1 is the density of water and d2 is the density of oil, when an additional data is available, e.g. a measurement of the fluid flow speed or knowledge of the density of one of the components. By means of calibration data R may be determined directly from the delay.

[0060] When the measurements are used to generate a signal, e.g. a signal to control fluid flow shut off, a signal to control fluid flow redirection or an alarm signal, translation may not be not needed. A threshold value may be set and the control circuit may be configured to generate the signal when it measures a delay that exceeds the threshold value. Similarly, calibration may not be needed when the measurements are used in a feedback loop that is configured to regulate fluid density to a set value. In this case a delay value may be used as a set value instead and the feedback loop may be configured to regulate fluid density in a direction wherein the difference between the measured delay and the set value is reduced.

[0061] In addition to torsion wave speed in a pipe with a non-circular inner wall surface, wave speed in a pipe with a circular inner wall surface may be measured in the presence of the same fluid. The latter may be used to correct for effects of temperature and viscosity variations if they occur, e.g. by subtracting wave speed measured with a circular inner wall surface, from wave speed measured with a non-circular inner wall surface. Such wave speeds may be measured in different sections of a pipe, that have a circular and non-circular inner wall surface cross-section. Alternatively the wave speeds may be converted into a density (from measurements using the pipe section with the noncircular cross-section) and a viscosity value (from measurements using the pipe section with circular cross-section). When the viscosity is variable, its measurement can then be used to determine a correction of the density, using a calibration of the dependence of the correction of the torsion wave speed on viscosity or a model of this dependence (e.g. an analytical model, which is known per se).

[0062] FIG. 6a,b show shared transceivers for performing such measurements. In the embodiment of FIG. 6a the pipe comprises first and second reflector 60a,b and a transceiver 62 or transceivers in different grooves between first and second reflector 60a,b, at distances L1 and L2 from the first and second reflector 60a,b, wherein L1 and L2 may differ from each other. Reflector 60a,b may be realized as described in the preceding. The part of the pipe between first reflector 60a and transceiver 62 has a circular inner wall surface cross-section. The part of the pipe between transceiver 62 and second reflector 60b has a non-circular inner wall surface cross-section such as in FIGS. 2a-c. In this embodiment, transceiver 62 may be used to measure torsional wave propagation speeds of waves that have been reflected from first and second reflector 60a,b respectively.

[0063] In the embodiment of FIG. 6b the pipe comprises first and second transducers 64a,b and a receiver 66 or receivers axially between first and second transducers 64a,b. The part of the pipe between first transducer 64a and receiver 66 has a circular inner wall surface cross-section. The part of the pipe between receiver 66 and second transducer 64b has a non-circular inner wall surface cross-section such as in FIGS. 2a-c. In this embodiment, receiver 66 may be used to receive torsional waves from first and second transducers 64a,b respectively. Instead receivers may be provided at the location of first and second transducers 64a,b and transducers may be provided at the location of receiver 66.

[0064] A fluid density measuring device uses a pipe with a pipe wall that has an inner wall surface with a non-circular cross-section at least in an axial segment of the pipe. Preferably, the inner wall surface comprises one or more protrusions extending inward into the pipe and along the axial direction of the pipe. A transducer such as an ultrasound transducer located on the pipe wall is used to generate local motion of the pipe wall with a circumferential direction of motion. Preferably, the transducer is located between successive protrusions. A receiver, such as an ultrasound receiver, located on the pipe wall receives an ultrasound torsion wave generated by said local motion after the torsion wave has traveled through the axial section wherein the inner wall surface has a non-circular cross-section. The fluid density is determined from the propagation speed of the torsion wave.