The invention concerns a method for measurement of the flow velocity, the volumetric flow, and/or of the mass flow of a gas that flows in a pipe, in which method sound detectors are fitted in the flow pipe at a certain distance from one another in the longitudinal direction of the measurement pipe, and in which method a long-wave sound that proceeds in the flow pipe downstream and upstream exclusively in the base mode as a plane-wave front is fed into the flow pipe by means of sound sources from outside the measurement distance defined by said detectors, and the flow quantities are determined from the travel times of said sound downstream and upstream over the measurement distance, from the cross-sectional area of the pipe, and from the density of the gas to be measured.

Further, the invention concerns a device for measurement of the gas flow velocity and/or of quantities derived from same, such as volumetric flow and/or mass flow, which device comprises a measurement pipe, in which the flow to be measured runs, and which device comprises loudspeakers as the transmitters of the wide-band, low-frequency sound signals that proceed exclusively as a plane- wave front in the base mode and power amplifiers for said loudspeakers, and microphones as the sound detectors and amplifiers for the microphone signals, and which device is provided with calculator devices and programs so as to calculate said flow quantity or quantities from the distance between the sound detectors in the longitudinal direction of the flow pipe, from the cross-sectional area of the measurement pipe, from the density of the flowing gas, and from the travel times of the sound signals downstream and upstream over the distance between the sound detectors.

It is well known that the speed of progress of an acoustic piston mode in a pipe does not depend on the flow profile or on other profiles that produce local variations in the sound speed, but on the average flow velocity integrated across the cross-section of the pipe and on the sound speed prevailing in the gas composition that fills the pipe at rest [B. Robertson, "Effect of Arbitrary Tem¬ perature and Flow Profiles on the Speed of Sound in a Pipe", J. Acoust. Soc. Am., Vol 62, No. 4, p. 813...818, October 1977, and B. Robertson, "Flow and Tempera¬ ture Profile Independence of Flow Measurements Using Long Acoustic Waves", Transactions of the ASME, Vol. 106, p. 18...20, March 1984]. Thus, precise measurement of flow velocity independent from the profile can be carried out so that the travel times of the piston mode downstream and upstream are measured over a certain measurement distance, preferably by using a wide-band sound, as is described in the applicant's FI Patent No. 76,885. In order that the higher wave modes which disturb the measurements should already be attenuated at the near vicinity of the sound sources, the sound frequency must be sufficiently below a certain limit frequency, which depends on the shape and dimensions of the flow pipe and on the travel speed of the sound. For a pipe of circular section, this limit frequency is

Limit frequency f _c = (c - v _maχ )/(1.7*D) (1)

wherein c is the travel speed of sound in the gas at rest, v _maχ is the highest dimensioned flow velocity, and D is the diameter of the pipe. As is well known, the flow quantities can be calculated from the following formulae:

How velocity [m/s] v = O^L*^ ^'1 - £ ^~ ) (2)

Volumetric flow [cu.m/s] Q = v*A (3)

Mass flow [kg/s] M = Q*p (4)

wherein

v = average flow velocity

L = distance between sound detectors in longitudinal direction of pipe t _χ = travel time of sound downstream over the distance L t ₂ = travel time of sound upstream over the distance L Q = volumetric flow

A = cross-sectional area of the pipe M = mass flow p = density of gas

The above FI patent describes the measurement of the travel times by means of the correlation technique, and in said FI patent the choice of the wide-band transmission or of the correlator type has not been restricted in any way. In sub- claims, the use of a polarity correlator is suggested, which can be accomplished as fully parallel. In practice, it has been noticed that, in order that the required signal-to-noise ratio could be reached by means of the polarity correlation technique in the measurement conditions, which are usually noisy, it is required that the sound transmissions downstream and upstream are transmitted alternatingly and that the microphone signals are filtered so that they pick up exclusively the sound that is being transmitted at each particular time and eliminate any interfering sound. A sound that is transmitted in the form of frequency scanning and a filter connected to follow said sound, in principle, also permit that sound is transmitted in both directions at the same time, but with a different momentary frequency, in which case the downstream signals and the upstream signals of the two sound detectors can be separated by means of a total of four scannable filters, as is described in the FI Pat. Appl. No. 916102.

It can be considered that it is a drawback of the prior art discussed above that frequency scanning is, in principle, non-stationary, in which case the necessary filters are also time-dependent. This results in dispersion, i.e. the time delay of the filters is dependent on the frequency, which further readily results in an error in the determination of the travel time if the scanning is not chosen correctly, if the scanning is chosen too rapid, or if the beginning and the end of the scanning

are not cut off from the measurement. These drawbacks can be avoided if, in stead of frequency scanning, wide-band transmissions are employed in which all frequencies sound constantly with an invariable amplitude. At the same time, this means that sound is transmitted in both directions simultaneously. For example, the present DSP-technique permits even a versatile processing of signals of audio frequency in real time and provides a number of possibilities for determination of the travel-time information in a noisy environment.

In view of elimination of the drawbacks mentioned above and in view of achiev- ing the objectives that will come out later, the method of the invention is mainly characterized in that said sound is transmitted into the measurement pipe as stationary, periodic sequences that proceed over the measurement distance simultaneously in both directions, downstream the sequence S _d (t,T) and upstream the sequence S _u (t,T), wherein T is the length of the period and t is the relative time measured from the beginning of each period, and that said sequences S _d (t,T),S _u (t,T) are orthogonal in relation to one another, i.e. they have no common frequency components different from zero.

On the other hand, the device in accordance with the invention is mainly characterized in that the sound transmitted through said loudspeakers is com¬ posed of wide-band sequences of equal periods, which are transmitted during the period T simultaneously downstream and upstream, the downstream sequence S _d (t,T) and the upstream sequence S _u (t,T), which sequences S _d (t,T),S _u (t,T) are arranged as orthogonal in relation to one another, i.e. they include no common frequency components different from zero.

In the following, the theoretical background of the invention and some exemp¬ lifying embodiments of the invention will be described in detail with reference to the figures in the accompanying drawing, wherein

Figure 1 is a general block-diagram illustration of an acoustic system of measurement in accordance with the invention,

Figure 2 illustrates sound sequences transmitted downstream and upstream,

in Figure 3, the upper curve represents a correlation peak consisting of even frequencies and imagined as measured downstream, and the lower curve represents a correlation peak consisting of odd frequencies and imagined as measured upstream,

Figure 4 illustrates, in a way corresponding to Fig. 3, an application in which, in stead of a second sequence to be correlated, a Hubert transform of same has been chosen,

Figure 5 shows graphs of a phase difference obtained with even and odd frequencies in an application in which the modulation arises from reflec- tion echoes alone, and

Figure 6 shows the cumulative distributions corresponding to Fig. 5.

Fig. 1 is a general block-diagram illustration of a system by whose means the invention can be carried into effect advantageously. The flow to be measured, e.g. a natural-gas flow, flows at a velocity v in the flow pipe 1, in which the sound sources 2a and 2b, preferably loudspeakers, are fitted, by whose means the, sound sequences are produced in the flow pipe 1. Between the sound sources 2a and

2b, sound detectors 3a and 3b are fitted, the scaling factor of the flow meter being determined by the length L and by the cross-sectional area A of the pipe portion between said sound detectors. In Fig. 1, 4a and 4b are power amplifiers,

5a and 5b are signal amplifiers, 6 is a real-time processor, preferably a digital signal processor provided with necessary analog inputs and outputs, by means of which processor 6 it is possible to carry out the procedures based on FFT- algorithms or on FIR-filter algorithms, 7 is a system processor, through which the communication between the real-time processor 6, the display device 8 and a separate PC work station 9 is accomplished.

In the determination of the travel time of a stationary, wide-band sound for purposes of flow measurement, it is essential that the sounds that are transmitted simultaneously in both directions from the loudspeakers 2a and 2b do not interfere with each other and that the measurement is also as insensitive to the noise in the pipe 1 as possible. Sound sequences, which have been transmitted by the loudspeakers 2a and 2b, which are periodic with respect to a certain measur¬ ement period T, but which are in the other respects arbitrary, will be examined. It is well known that such signals consist of discrete, equally spaced frequency components A _m *cos(m*2*7r*t/T+ø _m ), which are multiples of the fundamental frequency f = 1/T, and which are orthogonal in relation to one another, i.e. the correlation functions calculated across the period disappear when m and n are of unequal magnitudes (Formula 5).

— *J__ _J _cos(_.*2π— +φ *_i _R cos(π*2π— +φ„)*<ft = Λ _m *._i _Λ *δ _M ^*cos(_-*2π*— ) (

The principal idea of the invention is to transmit, from the loudspeakers 2a and 2b into the measurement pipe 1, downstream and upstream, such wide-band sound sequences S _d (t,T),S _u (t,T) of equal periods as do not include any common frequency component and as are, thus, orthogonal, i.e. their correlation function disappears identically:

C β) ^~ /5 f,7)*S _κ (.-θ,2)*Λ = 0 (6)

In such a case, the signals of the microphones 3a and 3b, herein, let D _L (t,T) represent the signal of the microphone 3a at the left end of the measurement distance L, and let D _R (t,T) represent the signal of the microphone 3b at the

right end of the measurement distance L, with the exception of the background noise, can be reduced unequivocally to downstream and upstream components orthogonal in relation to one another:

The principal idea of the present invention permits the solution of a number of problems that deteriorate the signal-to-noise ratio and leads to several different alternative embodiments. In an examination over a long period of time, the proportion of the external pipe noise that occurs in these discrete frequency components is the smaller, the higher the number of the periods is over which the situation is examined. In an averaging made over N periods, the periodic signal increases in proportion to N, whereas the incommensurable noise increases in proportion to the square root of N.

The phase angles of the discrete frequency components can be chosen freely, in principle. In order to avoid the effect of non-linearity of loudspeakers 2a and 2b or of equivalent sound sources and, for example, owing to safety regulations concerning natural-gas pipes, it is advisable to avoid momentary peaks of sound power. This aim is carried into effect readily if said phase angles are chosen at random. The frequency range to be used is limited at the upper end by the so- called cut-off frequency [formula (1)], at which, in addition to the piston mode, the first non-attenuating higher mode occurs, which has a different travel speed and which thereby disturbs the precise determination of the travel speed of the piston mode. At the lower end, it is often advisable to cut off the mains fre¬ quency of the AC electricity network and frequencies that are lower than its first upper harmonic.

The division of the remaining frequencies between the sound sequences trans-

mitted downstream and those transmitted upstream can be carried out in different ways. The simplest way is to choose the even frequencies for one sequence and the odd frequencies for the other. This choice has the result that each sequence consists of two half-periods, of which the half-periods of the even frequencies are identical with each other, as is illustrated by the upper graph in Fig. 2, and the half-periods of the odd frequencies are also identical with each other when the signs of one of them are changed, as is illustrated by the lower graph in Fig. 2. Thus, the sounds that proceed downstream and upstream can be separated from each other simply by adding together an even number of half- periods, in one case as such, and in the other case by first changing the sigh of every second half-period. A second alternative is to allot the available frequencies irregularly but evenly among the downstream and the upstream sequences.

The travel times of the measurement sound over the measurement distance L downstream and upstream can be determined in a number of different ways even from the same measurement data. The basic alternatives are to form a filtered correlation function of the signals of the microphones 3a and 3b as a function of the time difference or to examine the phase difference of successive frequency components of the signals concerned as a function of frequency. Rapid Fourier transforms of the digital signal processors with FFT-algorithms permit processing of a signal in a time and frequency space, if necessary, by means of fully the same apparatuses while interfering with the real-time program only.

The determination of the travel time of the measurement sound from correlation functions takes place as follows. The situation is most ideal for determination of the middle point of a correlation peak when there is only one peak in the whole correlation function, i.e. the amplitudes of other possible peaks are so little that the distortion caused by them in the area of the main peak remains below the permitted level. The origin of interfering peaks may be in background noise that proceeds in the wrong direction, from which noise frequency components chosen for the main peak have been left over by the correlation filter, or connections

and joints of the detector, which produce interfering echo peaks. If the interfer¬ ing peaks cannot be attenuated to a sufficiently low level, attempts must be made to bring their location and range of effect outside the range of variation of the main peak. The location of the echoes can be affected by means of the locations of the connections of the detectors 3a and 3b, and the range of effect can be affected thereby that the frequency-dependence of the amplitudes of the frequency components is made even, while avoiding abrupt points of discontinu¬ ity. For the signals of the microphones 3a and 3b, three different correlation functions can be formed:

n'T ^c " ^(θ) ~ ^~ 0ZT f ^{Fu *Aver} « ^[ ^'^ ^ ^W «W - ^θ ' ^7)] **

wherein F _d and F _u denote filter operators, which pick up the downstream portion only or the upstream portion only from the detector signals, Aver _m means averaging over m periods, and n means averaging of correlation functions by integration over n periods. The first correlation function is the simplest one and substantially includes two peaks, one of them with a positive time-delay value and the other one with a negative time-delay value, corresponding to travel of sound downstream and upstream. This one is also most sensitive to disturb¬ ance. Each of the following, filtered correlation functions includes substantially one peak only, one representing the travel of sound downstream only, and the

other one upstream only. Fig. 3 shows, in the upper half, a correlation peak consisting of even frequencies and imagined as measured downstream, and, in the lower half, a correlation peak consisting of odd frequencies and imagined as measured upstream. The small side peaks are produced by negative reflections from the loudspeaker branches.

In stead of the possibility that the signals of both of the microphones, restricted to the frequencies chosen for downstream or upstream, are correlated directly, the signal of each microphone 3a and 3b can be correlated separately with a reference sequence R _d (t,T) and R^T) of invariable phase and of the desired distribution of amplitude, which sequences have been derived from correspon¬ ding transmission sequences S _d (t,T) and S _u (t,T) or from microphone sequences measured in a zero flow situation, the following correlation functions being produced:

in which case the sound travel time over the measurement distance L down¬ stream is determined as the time difference between the middle points of the peaks of the correlation f nctions C _j (θ,T) and C _Rd (θ,T), and upstream as the time difference between the middle points of the peaks of the correlation functions C _Lu (θ,T) and C _Ru (θ,T). For this, it is possible to use either two pairs of FTR-filters (finite impulse response), the downstream reference sequence

being chosen as the coefficient vector for one of them, and the upstream reference sequence for the other one, respectively, or it is possible to carry out the operation in a frequency space as the following multiplications of Fourier- transformed vectors carried out component by component:

(10 )

wherein as a superscript means a complex conjugate, and by reverse-transform¬ ing the product vectors into correlation functions of the time space. Also in this case, it is possible to carry out the averaging operations corresponding to those that were carried out in connection with the correlation functions.

In stead of an ordinary correlation function, which produces fully symmetric peaks in an ideal case, in all of the above cases it is alternatively possible, in stead of a second sequence to be correlated, to choose its Hubert transform, in which case the correlation peaks become fully antisymmetric. Their middle point, which is obtained as an intersection with the zero level, is perhaps determined more precisely than the middle points of symmetric peaks. This case is illustrated in Fig. 4, which is, with the exception of symmetry, in the other respects similar to the case shown in Fig. 3.

Determination of the travel time as dependence on the phase angle takes place as follows. If the correlation function includes substantially one peak only, the phase angle of its Fourier transform is increased or decreased in a linear way as a function of the frequency, the angle coefficient being proportional to the shift of the peak away from the origin, for the shifting of time by means of the shift θ is carried out in a time space by multiplication by the phase factor exp(2_-

r^ω-^θ). An estimate for the travel time of the measurement sound over the measurement distance L is now obtained by forming the average of the phase- angle differences calculated over successive frequency steps of equal parity over the distance l^..!^

2π *(i ₂ -t ₁ ) _/

(11)

In the formulae (11), the Fourier transforms of the signals of the microphones 3a and 3b have always been used as the frequency vectors. As was stated above in connection with the correlation functions, also in the formula (11), it is always possible to replace the vector derived from the signal of one of the detectors 3a,3b by the Fourier transform of the reference sequences R _d (t,T) and R _u (t,T), in which case the following four time shifts θ _ά , θ _Rd , θ _Lu , and θ _Ru are obtained. The travel times downstream are obtained as the difference between the first two, and the travel time upstream as the difference between the latter two. In all of the above cases, out of the successive phase-angle differences, it is alternative¬ ly possible to form a cumulative function and to determine the time shifts as an angle coefficient of the regression line of said function.

If the correlation function also includes other peaks, modulation occurs both in the phase-difference distribution and in the cumulative function. One does, however, not fall off from the correct line if the phase-angle difference remains genuinely within the range of -_-...+_-. The risk of coining above or below these limits is reduced if the average value of the phase-angle difference remains near zero. For this reason, it is advisable to compare the factual travel time, for

example, with the travel time in a zero flow situation in the reference pipe 1, in which case the travel-time difference produced by the flow velocity v is alone seen as a minor deviation from zero in the average of the distribution of the phase-angle difference. Fig. 5 shows the phase difference graphs both for even frequencies and for odd frequencies in a case in which the modulation arises from reflected echoes alone. Fig. 6 shows the corresponding cumulative distribu¬ tions.

In order that the phase-angle method were usable, it is required that the correla- tion function has a clearly distinguishable main peak and that, with every individual frequency, the phase angle of its Fourier transform comes reasonably close to the value that corresponds to the travel time of said peak.

In the following, the patent claims will be given, and the various details of the invention may show variation within the scope of the inventive idea defined in said claims and differ from the details given above by way of example only.