WO1999005479A1

WO1999005479A1 - Velocity measurement

Info

Publication number: WO1999005479A1
Application number: PCT/GB1998/002248
Authority: WO
Inventors: Michael Rigby
Original assignee: Pcme Ltd.
Priority date: 1997-07-28
Filing date: 1998-07-28
Publication date: 1999-02-04
Also published as: GB9715897D0; GB2327761A; AU8549298A

Abstract

Method and apparatus for measuring the velocity of a phase of a fluid flow, wherein, to ascertain the time shift required for close correlation of a signal from a first signal transmitter or detector and a signal from a second signal transmitter or detector, samples of the signal associated with the first transmitter or detector and samples of the signal associated with the second transmitter or detector are obtained and stored, at least some of the samples are retrieved from storage and subjected to a first correlation over a relatively wide range of time shifts and at least some of the samples are retrieved from storage and subjected to a second correlation over a relatively narrow range of time shifts selected according to the results of the first correlation stage.

Description

Velocity Measurement

This invention relates to a method and apparatus for measuring the velocity of a phase of a fluid flow and in particular, but not exclusively, particles in a fluid flow. The invention particularly relates to an arrangement in which particles in a fluid flow are detected by detectors and cross-correlation of signals from the detectors is used to determine the particulate velocity. The invention is, however, also applicable to the measurement of the velocity of a single phase fluid flow (single phase, in the sense that all components of the flowing fluid travel together as in the case of, say, an air flow where nitrogen and oxygen travel together but not as in the case of a typical particulate flow in which the particulate would represent a first phase and would travel more slowly than the fluid carrying the particles, which fluid would represent a second phase) .

The technique of cross-correlation has been applied before to the problem of measuring the velocity of a fluid flow. The technique has been implemented in apparatus utilising various sensing methods, such as electrodynamic or triboelectric sensing. Commercial instruments are available from ABB Kent and Endress & Hauser. A problem with existing cross-correlation devices is that they are either slow, if they are based on a microprocessor, or expensive, if they are based on signal -processing hardware.

Those problems result from the fact that cross- correlation requires a large number of arithmetic calculations. The response speed of a cross-correlation device will generally be approximately inversely proportional to the square of the number of samples correlated; thus, a trade off must be made between the need for an adequate response time and the need for reliable data. Endress & Hauser manufacture a device which is microprocessor-based and correlates 256 samples from its particle detectors. However, in order to increase its speed of operation, it correlates only eight points in every scan of the input cycle. In order to build up the whole result 32 scan cycles are required, and if the particulate velocity changes whilst that processing is taking place, the device will not produce an accurate result . GB 2039110 describes a device which is hardware based and which attempts to correlate data in two stages to enable a velocity measurement to be made whilst providing a continuous output of the velocity. First a coarse correlation of the data is carried out and thereafter a fine correlation is undertaken; the problem with this approach is that the coarse correlation has to be carried out before the fine correlation and, if there is any change in the two sets of data being correlated after the coarse correlation and before the fine correlation, a wrong result can easily be obtained. If the velocity being measured changes suddenly, or if there are two peaks of similar magnitude in the correlation curve, incorrect results are very likely.

ABB Kent manufacture a device which is hardware- based and is capable of correlating up to 1024 points, in one scan, at a higher speed of operation. However such hardware devices are much more expensive than devices based on microprocessors.

When correlating values of x and y the value of the correlation of n samples at time shift position m is given by: ^T (I x, - x iy, _* m - y_mj = ^ lx,y, . „, + xy_m - xy, - m - ymxλ

= xy, <- m + jT xy_m - xyi • m - ∑ ymx, n n

Xtatal 'ι - m ym total /^ Xi

Xtotaiym total

= ∑x.y, ,+ n n n

Xtotaiym total Xtotaiym total Xtotaiym total

= ∑x.y, ,+ n n n n

Xtotaiym total

= ∑x.y ι=0 n

Where:

Xtotal - XO + X\ + . . .+Xn - I ym total = ym + ym + 1+ +y - n

It will be understood that in the terms x^ and y^, "i" designates the number of the x or y sample in the group of samples being correlated.

The result is evaluated for m=0 to m=n-l; i.e., the above calculations are carried out n times. Thus, if for example 4096 samples are taken from detector 1 and 8192 are taken from detector 2, the first calculation correlates samples 0 - 4095 of set 1 against samples 0 -

4095 of set 2 to give result number 0, the second calculation correlates samples 0 - 4095 of set 1 against samples 1 - 4096 of set 2 to give result number 1 and so on. The correlation of 4096 samples therefore requires

4096 * 4096 = 16777216 multiply-add calculations (macs) . A full correlation of 4096 samples is therefore not a practical possibility; for example, an Hitachi Super-H processor would perform the mac calculation in approximately 150 ns, and so the entire correlation of the full set of samples would take approximately 2.5 seconds; that is too slow for the effective monitoring of particle velocity in most applications. Furthermore, in the case of analogue correlation, because each sample can take any value between -128 and +127, the magnitude of the correlation value depends not only on the degree of matching of the samples but also on their amplitude. That can result in a maximum correlation result at a point other than where the signals are best matched. The correlation result is therefore normalised. To normalise the correlation result, it is divided by:

As ∑(x, - ϊ) = ∑{x,² + (x)² - 2χ-x,) = ∑_x;- + ∑(x)² - 2∑^~xx,

Σ 2 flX total Xtotal ■<— . ,

X, + 2 Xtotal = 2_ X,- ^{~ ■} n n n the final normalised correlation result is therefore

Normalisation of the correlation result requires still further processing time, and a full correlation of 4096 samples from each detector would result in the particle velocity measurement apparatus having an unacceptably slow response speed if it were microprocessor-based. Apparatus based on hardware would have a faster response speed, but it would also be significantly more expensive. The additional steps required with analogue correlation, such as normalisa- tion, make hardware-based analogue correlation impractical, so hardware correlators use polarity correlation.

It is an object of the invention to provide a method and apparatus for determining the velocity of a phase of a fluid flow that avoids or mitigates the above problems and gives a reliable indication of the particle velocity. The present invention provides a method of measuring the velocity of a phase of a fluid flow, the method including the following steps: providing a first signal transmitter or detector at a first location in the flow, providing a second signal transmitter or detector at a second location in the flow downstream of the first location, at least one of the first and second signal transmitters or detectors being a signal detector, ascertaining the time shift required of the signal associated with the second transmitter or detector relative to the signal associated with the first transmitter or detector to obtain close correlation of the signals, and determining the velocity of the phase of the flow from the time shift, wherein, to ascertain the time shift required for the close correlation, samples of the signal associated with the first transmitter or detector and samples of the signal associated with the second transmitter or detector are obtained and stored, at least some of the samples are retrieved from storage and subjected to a first correlation over a relatively wide range of time shifts and at least some of the samples are retrieved from storage and subjected to a second correlation over a relatively narrow range of time shifts selected according to the results of the first correlation stage. By carrying out the correlation in more than one stage, it is possible to reduce considerably the total number of calculations that have to be carried out to find the time shift required for close correlation. Thus the speed at which the calculations have to be carried out in order to arrive at a measurement of velocity is much reduced. For example, as described in more detail with reference to the drawings, the correlation referred to above and requiring 4096 * 4096 = 16,777,216 calculations when carried out in a single step can be carried out in two steps in accordance with the invention using only 512 * 512 + 100 * 4096 = 671,744 calculations .

Furthermore by storing the samples it is possible to carry out both the first and second correlation stages on the same samples, which is very advantageous and eliminates the possibility of the data changing after the first stage of correlation and before the second stage.

It should be understood that "close correlation" refers to a condition in which the correlation curve resulting from plotting the correlation result against the time shift of the signals from the detectors is substantially at its maximum value. The substantially maximum value of the correlation curve may be defined in various ways. In the simplest case the maximum value of the correlation curve is defined as the highest peak in the curve but that approach is not well suited to the case where the correlation curve has twin peaks of similar height and the position corresponding to the true velocity lies between the two peaks. Accordingly it is preferred to adopt an additional "averaging" step in relation to the correlation curve. Such an "averaging" step should transform a correlation curve with twin peaks into one with a single peak. Because the results representing the correlation curve can be stored, such an averaging step can readily be carried out and need not unduly lengthen the total time required to complete correlation.

As indicated, it is most preferable that the stored samples that provide the samples for the first correlation also provide the samples for the second correlation. It is desirable from a performance point of view for the second correlation to use the same samples as the first correlation, but not essential. The first and second correlations may be carried out on samples which are obtained and stored at different times. Preferably each of the first and second signal transmitters or detectors are particle detectors, each providing an output signal representative of particle flow at the respective detector. Another possibility, however, is for there to be one signal transmitter and one signal detector; for example, the first detector may be an ultrasonic transmitter and the second detector may be an ultrasonic receiver. The velocity of the flow influences the time taken for the signal from the transmitter to reach the receiver and thus the time shift between the transmitted and received signals provides an indication of flow velocity.

Preferably, the first correlation is carried out using average output values, each derived by averaging a plurality of individual samples, and preferably by averaging four or more samples . The step of averaging individual samples during the first correlation reduces the loss of information resulting from the fewer number of samples that are employed in the first correlation compared to the number that would be employed in a conventional single stage correlation and helps to ensure that the first correlation stage reaches the correct conclusion as to the approximate time shift required to give close correlation.

In principle, where particle detectors are employed, the detectors may take any suitable form but in a preferred embodiment they detect particles by detecting an electrical effect caused by the particles. The electrical effect may include a triboelectric effect, although it is possible that a significant part of the effect is from other electrodynamic effects such as charges induced from charged particles passing in the vicinity of the detector. The particle detectors may project into the particle flow in which case the detector may be in the form of a rod which may be of circular cross-section and may comprise an electrically conducting core which may be exposed to the particle flow or may be covered with an insulating layer which insulates the core from the particle flow. Suitable detecting arrangements are described in GB 2266772 and GB 2277154, the disclosures of which are incorporated herein by reference. Alternatively the particle detectors may each extend around the boundary of the particle flow and may be in the form of a ring defining a length of the passageway along which the fluid flows; such detectors are known per se . Alternatively, the detectors may be stud sensors; that is a set of studs forming a discontinuous ring which extends around the boundary of the particle flow. Such an arrangement allows for installation of a ring-like detector without the need to cut the conduit containing the flow, as is necessary in the case of a conventional ring sensor.

The invention may be applied to measuring the velocity of particles in a liquid flow but preferably the fluid is a gas and the particles are liquid or solid particles suspended in the gas. The particles may be flowing along a stack and emitted through the stack.

Preferably, a.c. components of the output signals from the detectors are used to provide the samples that are employed in the first and second correlation stages. Although d.c. levels may be used we have found that it is advantageous when using a triboelectric detector to eliminate the d.c. component of a signal and use the a.c. component since that provides a signal more representative of particle flow.

Preferably more than 1000 samples, and more preferably more than 4000 samples, are obtained from each transmitter or detector on each occasion that the time shift required for the close correlation is ascertained. Sampling as many as 4000 samples enables an accurate measurement of the time shift required for close correlation to be obtained and can be achieved in a reasonable time without requiring unduly expensive hardware, when the method of the invention is used.

Preferably consecutive samples are taken alternately from the first and the second signal transmitters or detectors. The samples from the two transmitters or detectors thus span substantially the same period of time .

Preferably, the second correlation is carried out over a range of time shifts at least four times narrower, preferably at least eight times narrower, than the range of time shifts over which the first correlation is carried out . Because the second correlation is carried out over a narrower range of time shifts it improves the resolution of the time measurement and so provides a more accurate measurement of the time shift required for close correlation.

In an embodiment of the invention described below with reference to the drawings, the samples are subject to two stages of correlation, but it is possible for the samples to be subject to more than two stages of correlation, and in the one or more further stages the samples are then subjected to correlation over progressively narrower ranges of time shifts. Such an approach can result in a reduced number of calculations but requires more complicated software for its implementation.

The method of correlation employed may be analogue correlation; in that case the apparatus may include an analogue-digital converter, in which the analogue signals from the probes are converted into digital represen- tations of numbers between, for example, -128 and +127. Preferably, the correlation data resulting from a correlation stage are normalised before the time shift required for close correlation is ascertained.

Alternatively, the method of correlation employed may be polarity correlation; in that case the apparatus may include a comparator, in which the signals from the detectors are sampled as binary values (high or low) . The correlation calculation that is repeated in polarity correlation when correlating values of x and y for n samples at time shift position m is given by:

Correlation Result = χ„ ® y_n ■ m

In words, the Correlation Result can be said to be the bitwise Exclusive-Or of x_n and y_n+m. Using simple microprocessors, the Exclusive-Or and add calculation required in polarity correlation is much quicker than the multiply and add calculation (mac) required in analogue correlation. Because of that speed increase, and the reduction in hardware complexity, previous instruments have tended to use polarity correlation. The advantage of the analogue method is that although the computational requirements are more severe, if a powerful micro- processor is used, the hardware required for implementation is simpler, and so the hardware cost is lower. An Hitachi Super-H processor, for example, performs the mac calculation in about 150 ns . If the polarity correlation is performed in hardware for maximum speed, then, with a 10 MHz clock, the Exclusive-Or operation takes about 100 ns . Thus the polarity correlation in hardware may be very little faster than the analogue correlation performed by a powerful microprocessor (ignoring time taken to normalise the data with the analogue method) , but the analogue correlator may be significantly cheaper and is therefore preferred.

The present invention also provides an apparatus for measuring the velocity of a phase of a fluid flow, - li ¬

the apparatus comprising a first signal transmitter or detector at a first location in the flow, a second signal transmitter or detector at a second location in the flow downstream of the first location, at least one of the first and second signal transmitters or detectors being a signal detector, and electrical apparatus, which may include a microprocessor, coupled to the first and second signal transmitters or detectors, the electrical apparatus including storage means for storing signals associated with the transmitters or detectors, evaluating means for measuring the velocity of the phase of the flow from the sampled outputs of the transmitters or detectors, subjecting at least some of the samples of the signals associated with the transmitters or detectors to a first correlation over a relatively wide range of time shifts and then subjecting at least some of the samples to a second correlation over a relatively narrow range of time shifts selected according to the results of the first correlation stage, thereby ascertaining the time shift required of the signal associated with the second transmitter or detector relative to the signal associated with the first transmitter or detector to obtain close correlation of the signals.

The velocity measurement may also be combined with a particle concentration measurement enabling a measurement of mass flow to be obtained. Thus the present invention further provides a method of measuring the mass flow rate of particles in a fluid flow, the method including a method as defined above of measuring the velocity of the particles; similarly, the present invention provides an apparatus for measuring the mass flow rate of particles in a fluid flow, the apparatus including an apparatus as defined above for measuring the velocity of the particles . In an especially preferred embodiment one of the particle detectors used for measuring the velocity of the particles can also be used for measuring the concentration of the particles and certain signal processing circuitry connected to said one particle detector can be used for both purposes . An embodiment of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

Fig. 1 is a sectional view of two particle detectors of a particle velocity measuring apparatus mounted in a wall of a stack through which dust particles in a flow of air are emitted;

Fig. 2 is a block diagram representation of an electrical system of the particle velocity measuring apparatus ; Fig. 3 is a memory map representing storage of sample data in a random access memory (RAM) of the electrical apparatus of Fig. 2;

Fig. 4 is a flow diagram showing the steps involved in a correlation process carried out by the particle velocity measuring apparatus;

Fig. 5 is a typical correlation curve generated from data from the particle velocity measuring apparatus; and

Fig. 6 is a section of the correlation curve of Fig. 5, the curve of Fig. 6 being derived from more detailed data near the time shift required for close correlation.

Fig. 1 shows a first particle detector 1 and a second particle detector 2 mounted in the wall of a stack 3. The particle detectors are insulated probe devices of the kind described in GB 2277154. As shown in Fig. 2, the outputs of the detectors 1, 2 are connected to electrical apparatus comprising electrical signal processing circuitry 6 and then to evaluating means comprising an analogue-digital converter 7, an input/output interface 8, external RAM 9, a Direct Memory Access Controller (DMA) 10, and a microprocessor unit 11. The individual elements of the evaluation means are well-known per se and will not be described in detail herein. As will now be described also with reference to the other drawings and in particular the flow diagram of Fig. 4, the apparatus just referred to processes signals from the detectors 1, 2 and, by ascertaining the time shift required for close correlation of signals from the two detectors, measures the velocity of particles in the stack. Dust particles suspended in air flowing in the stack interact with the first and second detectors 1, 2 and, as a result of electrodynamic effects, cause charging of the detectors; this is shown as Step A in Fig. 4. The signals from each of the probes 1, 2 are processed by the electrical circuitry 6 in order to remove the d.c. component from each signal (Step B) . The charging of the detectors and the processing circuitry for measuring the d.c. component of the signal are described in GB 2266772 and GB 2277154. The remaining a.c. component of each signal passes through the 8 -bit analogue-digital converter 7 which converts the magnitude of the a.c. signal into a digital signal with a value of between -128 and +127 (Step C) obtained by sampling the a.c. component at a fixed clock rate. The sampled signal passes through the input/output interface 8 and is stored in external RAM 9 by the DMA 10 (Step D) . In the particular example of the invention being described, the external RAM 9 takes the form of two 32 Kbyte*8 bit RAM chips which together provide 64 Kbytes of 16-bit RAM.

Data is acquired from the two particle detectors using an interleaved method as illustrated in Fig. 3: consecutive samples are acquired alternately from the two particle detectors. In the particular example described, 8192 samples are acquired from each detector, but the second half of the array of samples from the first detector is disregarded, because only 4096 samples are required from the first detector for the correlation. Correlation of the samples is controlled by the microprocessor unit 11, which in the particular example described is an Hitachi Super-H 32 -bit Rise processor which includes as part of the unit the Direct Memory Access Controller 10.

For the array of samples from detector 1 and for the array of samples from detector 2, the respective average value of the array is calculated and that value is subtracted from the value of each sample within the array (Step E) . That operation removes any residual d.c. component from the samples.

In the first correlation stage, samples from each of the particle detectors are averaged in blocks of eight readings (Step F) . As will be understood, each block of eight readings comprises eight readings from the same detector that are adjacent to one another. Those blocks of samples are then correlated in the internal RAM of the microprocessor 11 (Step G) ; in the particular example described, the first stage of correlation requires 512*512 macs and takes 39 ms . The results of the first correlation stage are then normalised (Step H) , the process of normalisation being as described in this specification. The normalised results can be represented by a correlation curve, as shown in Fig. 5 and the highest peak of the normalised correlation curve determined (Step I) . The time shift corresponding to that highest peak is the time shift required to obtain a close correlation of the outputs from the first and second detectors .

Having established during the first, somewhat coarse, stage of correlation the approximate time shift (in this example, about 100 units in the x-axis) required to obtain a close correlation of the sampled signals from the two detectors, the same stored data from the particle detectors 1, 2 are subject to a second stage of correlation (Step J) . The second stage of correlation correlates the original samples of data individually (without averaging) , but correlates them only in a narrow window, in this particular example a window of 100 time shifts, around the time shifts at the highest peak of the correlation curve resulting from the first stage of correlation. The second stage of correlation therefore requires only 100*4096 macs, taking, in principle, 61.4 ms (although implementation-related factors may increase the calculation time) . The results of the second correlation stage are then normalised (Step K) . The normalised results can be represented by a correlation curve and the highest peak of the normalised correlation curve determined (Step L) .

Fig. 6 shows the correlation curve obtained from combining the results of the second stage of correlation (between 50 and 150 on the x-axis) with some of those of the first stage (outside that area, where the resolution is reduced by a factor of eight, namely 0 to 50 and 150 to 4096 on the x-axis ; note that Fig. 6 shows only the first 400 points) . The location of the peak of the correlation curve is thus determined to within one sample duration. The location of the peak gives the time shift required to obtain a good correlation of the outputs from the first and second detectors. The particulate velocity is obtained (Step M) by dividing the spacing of the first and second particle detectors (HeadSpacing) by the time shift (TimeDelay) required to obtain a good correlation:

Veloc.ty = ^^dSP^acιng. TimeDelay

In the particular example shown in Fig.5, the peak of the correlation curve occurs at position 95 on the x-axis. The data from which Fig. 5 was generated was originally sampled with a 30 μs sample duration and so the time shift required for good correlation is 95*30 μs; i.e., 2.85 ms . The head spacing in the particular example was 50 mm, and so the particulate velocity is 50/2.85 = 17.5 m/s. The particle velocity measuring apparatus of the particular example of the invention just described generates a particulate velocity measurement approximately twice per second. That response frequency is adequate for the monitoring of particulate velocities in many applications.

The invention enables analogue sampling and rapid correlation of a large number of data points (in the example described, 4096) , using a standard microprocessor rather than expensive custom-made hardware. In terms of random access memory, the apparatus of the particular example only requires the internal memory of the microprocessor and two 32 Kbyte external RAM chips. Fast 32 Kbyte RAM chips are available at low cost and are ideal for this application. The particle velocity measuring apparatus requires no more than 64 Kbytes of RAM, avoiding the need to buy more expensive memory devices. The Hitachi Super-H processor contains 8 Kbytes of internal RAM: the correlation data stores require 4096 bytes of memory, and the internal memory is also used to hold the processor working stack. Data acquisition requires 16384 bytes of external RAM to store two sets of 8192 samples; processor mac operations require 4096 + 8192 16-bit locations (i.e. 12288 16-bit locations, or 24576 bytes of memory) ; the correlation result array consists of 4096 16-bit words, and two arrays are stored (i.e. 16384 bytes are required); and the normalisation factor array contains 512 normalisation factors, with each factor normalising 8 results. The total memory requirement for the above operations is therefore 58368 bytes. The remaining 7168 bytes in the memory are available for the storage of other system variables. If desired, the apparatus for measuring particle velocity can be combined with apparatus for measuring particle concentration to provide a mass flow meter. One of the detectors 1, 2 used in the velocity measurement can also be used to provide a measurement of particle concentration as described in GB 2277154 (or if the insulated probes are replaced by ones where the electrically conducting cores of the probes are exposed, as described in GB 2266772) . A particular advantage of such an arrangement is that the same electrical circuitry 6 that is connected to the detector for the velocity measurement can be used for the particle concentration measurement .

Claims

Claims :

1. A method of measuring the velocity of a phase of a fluid flow, the method including the following steps: providing a first signal transmitter or detector at a first location in the flow, providing a second signal transmitter or detector at a second location in the flow downstream of the first location, at least one of the first and second signal transmitters or detectors being a signal detector, ascertaining the time shift required of the signal associated with the second transmitter or detector relative to the signal associated with the first transmitter or detector to obtain close correlation of the signals, and determining the velocity of the phase of the flow from the time shift, wherein, to ascertain the time shift required for the close correlation, samples of the signal associated with the first transmitter or detector and samples of the signal associated with the second transmitter or detector are obtained and stored, at least some of the samples are retrieved from storage and subjected to a first correlation over a relatively wide range of time shifts and at least some of the samples are retrieved from storage and subjected to a second correlation over a relatively narrow range of time shifts selected according to the results of the first correlation stage.

2. A method as claimed in claim 1 in which the stored samples that provide the samples for the first correlation also provide the samples for the second correlation.

3. A method as claimed in claim 1 in which the first and second correlations are carried out on samples which are obtained and stored at different times.

4. A method as claimed in any of claims 1 to 3 in which the first correlation is carried out using average output values, each derived by averaging a plurality of individual samples .

5. A method as claimed in claim 4 in which the average output values are derived by averaging four or more samples .

6. A method as claimed in any preceding claim in which each of the first and second signal transmitters or detectors is a particle detector, each providing an output signal representative of particle flow at the respective detector.

7. A method as claimed in any preceding claim in which the detectors detect particles by detecting an electrical effect caused by the particles.

8. A method as claimed in claim 7 in which the electrical effect is or includes a triboelectric effect.

9. A method as claimed in any one of claims 6 to 8 in which a.c. components of the output signals from the detectors are used to provide the samples that are employed in the first and second correlations.

10. A method as claimed in any preceding claim in which the fluid is a gas carrying particles and the particles are liquid or solid particles suspended in the gas.

11. A method as claimed in any preceding claim in which more than 1000 samples are obtained from each transmitter or detector on each occasion that the time shift required for the close correlation is ascertained.

12. A method as claimed in any preceding claim in which more than 4000 samples are obtained from each transmitter detector on each occasion that the time shift required for the close correlation is ascertained.

13. A method as claimed in any preceding claim in which consecutive samples are taken alternately from the first and second signal transmitters or detectors.

14. A method as claimed in any preceding claim in which the second correlation is carried out over a range of time shifts at least four times narrower than the range of time shifts over which the first correlation is carried out .

15. A method as claimed in any preceding claim in which the samples are subject to more than two stages of correlation, and in the one or more further stages the samples are subjected to correlation over progressively narrower ranges of time shifts.

16. A method as claimed in any preceding claim in which the method of correlation employed is analogue correlation.

17. A method as claimed in claim 16 in which the correlation data resulting from a correlation stage are normalised before the time shift required for close correlation is ascertained.

18. A method as claimed in any one of claims 1 to 15 in which the method of correlation employed is polarity correlation.

19. A method as claimed in any preceding claim in which the first correlation is carried out in a plurality of sequential steps providing intermediate results which are stored and the second correlation is carried out in a plurality of sequential steps providing further intermediate results which are stored.

20. An apparatus for measuring the velocity of a phase of a fluid flow, the apparatus comprising a first signal transmitter or detector at a first location in the flow, a second signal transmitter or detector at a second location in the flow downstream of the first location, at least one of the first and second signal transmitters or detectors being a signal detector, and electrical apparatus coupled to the first and second signal transmitters or detectors, the electrical apparatus including storage means for storing signals associated with the transmitters or detectors and evaluating means for measuring the velocity of the phase of the flow from the sampled outputs of the transmitters or detectors by subjecting at least some of the samples of the signals associated with the transmitters or detectors to a first correlation over a relatively wide range of time shifts and then subjecting at least some of the samples to a second correlation over a relatively narrow range of time shifts selected according to the results of the first correlation stage, thereby ascertaining the time shift required of the signal associated with the second transmitter or detector relative to the signal associated with the first transmitter or detector to obtain close correlation of the signals.

21. Apparatus according to claim 20 in which each of the first and second signal transmitters or detectors is a particle detector, each providing an output signal representative of particle flow at the respective detector.

22. Apparatus as claimed in claim 21 in which the evaluating means is arranged to sample the a.c. components of the outputs of the detectors.

23. Apparatus as claimed in any one of claims 20 or 22 in which the electrical apparatus includes a microprocessor .

24. Apparatus as claimed in any one of claims 21 to 22 in which the evaluating means includes an analogue- digital converter.

25. Apparatus as claimed in any one of claims 21 to 24 in which the evaluating means include a comparator.

26. A method of measuring the mass flow rate of particles in a fluid flow, the method including a method of measuring the velocity of the particles as claimed in claim 6, or any one of claims 7 to 19 when dependent upon claim 6.

27. A method as claimed in claim 26, in which the output signal from one of the detectors is used both for measuring the velocity of the particles and for measuring the concentration of the particles.

28. An apparatus for measuring the mass flow rate of particles in a fluid flow, the apparatus including an apparatus for measuring the velocity of the particles as claimed in claim 21 or any one of claims 22 to 24 when dependent upon claim 21.

29. An apparatus as claimed in claim 28, in which the apparatus includes means for determining the concentration of the particles in the fluid flow, the determining means receiving an input signal from one of the particle detectors whose output is sampled by the evaluating means.