WO2004031719A1

WO2004031719A1 - Method for detecting and locating at least one source of noise in a pipe transporting a fluid and installation therefor

Info

Publication number: WO2004031719A1
Application number: PCT/FR2003/002886
Authority: WO
Inventors: Joël MILLET; Pascal Simonie
Original assignee: Metravib R.D.S.
Priority date: 2002-10-04
Filing date: 2003-10-02
Publication date: 2004-04-15
Also published as: AU2003282193A1; FR2845474B1; FR2845474A1

Abstract

The invention concerns a method for detecting and locating at least one source of noise (S) in a pipe (2) transporting a fluid. The invention is characterized in that it consists in: providing on the pipe (2) a first series of at least three vibration and acoustic wave sensors, sampling the time signals delivered by each sensor and corresponding each to a signal emitted by the source of noise, determining from said sampled signals whether the source of noise corresponds to an impact on the pipe or to a leak in the pipe, and if the source of noise corresponds to an impact: determining a mathematical estimator such as F(X0, C)= sensorsS(i,j)[C?ti,j-f(X0, Dk(i,j)]2; and in minimizing the mathematical estimator F to determine the estimated values of the position Xo of the impact and the celerity of the sound C as follows: (X0,estimated, Cestimated)=arg[min (x0,C) F(x0,C)].

Description

DETECTION AND LOCATION METHOD

AT LEAST ONE SOURCE OF NOISE IN ONE

DRIVING VEHICLE AND

INSTALLATION FOR ITS IMPLEMENTATION The present invention relates to the technical field of the detection and localization of sources of noise and vibrations, such as leaks or shocks appearing in a pipeline or a pipe for transporting a gaseous or liquid fluid. .

Numerous techniques are known in the prior art suitable for locating, for example, a leak in a pipe. It is thus known for example a technique consisting in mounting on the pipe to be monitored, at least two sensors of acoustic or vibratory magnitude, separated from each other by a known distance D. A leak generates a noise which propagates in creating a pressure fluctuation in the fluid and / or a vibration of the structure of the pipeline. From the measurement of the difference in the time Δt of propagation of the leakage noise to the two sensors and as soon as the propagation speed V of noise propagation in the pipeline is determined, it is possible to deduce the distance d from the leak with respect to one of the sensors according to the relationship: d = (D - V Δ t) / 2.

In order to determine the difference Δ t of the propagation time of the leakage noise towards the sensors, the calculation of intercorrelation functions between the signals from the two sensors is carried out. When an intercorrelation function exhibits a maximum for a given delay, it can be affirmed that a leak exists at a point in the inspected pipe and this leak can be easily located from the difference in propagation time of the noise. leak which corresponds precisely to the delay given.

It turns out that this classic intercorrelation technique is not satisfactory in practice. This method has the drawback of being particularly sensitive to parasitic noise which may come from noise sources located in the environment of the pipeline to be inspected or from shocks occurring on the pipeline and which are liable to affect the integrity of this pipeline. Under these conditions, it turns out that irrelevant sources of noise corresponding to false alarms are detected when relevant sources of noise are not detected.

An object of the invention therefore aims to remedy the above drawbacks by proposing a method allowing the detection of a noise source in a pipe with a minimum of false alarms and the relatively precise location of the position of this source. noise corresponding to a shock on the pipe.

To achieve such an objective, the object of the invention is to propose a method for detecting and locating at least one noise source in a pipe conveying a fluid consisting of:

- to have on the pipeline a first series of at least three vibration or acoustic wave sensors separated from each other by known distances D _k (i, j>,

- taking the time signals Si (t) delivered by each sensor and each corresponding to the signal emitted by the noise source reaching said sensor with an arrival time tj,

- to be determined from the signals taken if the noise source corresponds to an impact on the pipeline or to a leak in the pipeline,

- and in the case where the noise source corresponds to a shock: • to determine a mathematical estimator such as

F (X „0 = _ £ [cΔt. _J - f (x ₀ , D _{k (i> j)} ) j sensors (i, j)

with Δ tj, _j : the difference in arrival times between the sensors i, jf = a function of the geometric arrangement of the sensors, c = the speed of the sound, • and to minimize the mathematical estimator F to determine the estimated values the position Xo of the shock and the speed of the sound C as follows:

Another object of the invention is to propose a method making it possible to detect a source of noise in a pipe with a minimum of false alarms and to locate in a relatively precise way the position of this source of noise corresponding to a leak on the pipe. To achieve such an objective, the object of the invention is to propose a method consisting in the case where the noise source corresponds to a leak:

- determining the reference sensor i that has detected the noise source,

- calculating a first cross-correlation function IVι, i between the signals emitted by the reference sensor i and a neighboring so-called upstream sensor i - 1 and a second cross-correlation function

between the signals emitted by the reference sensor i and a neighboring sensor called downstream i + 1,

- to look for in the Ti-ij intercorrelation functions and

emerging peaks whose number is between 0 and 2,

- and to analyze the position of the emerging peaks of the intercorrelation functions Γ ,, and

to reject the initial detection of the reference sensor i or on the contrary to locate the leak. According to a preferred variant embodiment, the method consists in locating the leak by analyzing the compatibility of the combinations of the peaks, possibly present in the intercorrelation functions Ti-i ,, and

, i ₊ ι. According to another preferred embodiment characteristic, the method according to the invention consists in analyzing the compatibility of the emerging peaks of the intercorrelation functions rVι, i and F, -, j ₊ ι in the following manner:

- if the intercorrelation function T _J - _I has no peak and if the intercorrelation function T-,, _{+ \} has: • either no peak, or a peak corresponding to a position in i + 1, then it is concluded that there was no leak,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that d '' a leak at a position between i and i + 1,

• either a peak corresponding to a position in i, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, i.e. a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak in position i, - if the cross-correlation function Fj.ι, i has a peak corresponding to a position at i and if the cross-correlation function rj, i ₊ 1 has:

• either no peak, or a peak corresponding to a position in i, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, or a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak in position i,

• either a peak corresponding to a position in i + 1, then it is concluded that there is a leak but not localized, • either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1, - if the intercorrelation function Γ _M , _J has a peak corresponding to a position between i-1 and i and if the cross-correlation function Fj, j ₊ ι has:

Either no peak, or a peak corresponding to a position at i + 1, or a peak corresponding to a position at i, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position at i, either a main peak corresponding to a position at i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i-1 and i,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then if the position of the peaks between i -1 and i and between i and i + 1 are equal to an error value close, it is concluded that there is a leak at position i, while otherwise it is concluded that there is a leak but not localized,

- if the cross-correlation function -i ,! has a main peak corresponding to a position in i and a secondary peak corresponding to a position between i-1 and i and if the cross-correlation function Γ _J , _{J + I} has:

• either no peak, or a peak corresponding to a position in i, or a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a flight in position i, • either a peak at a position in i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, then it is concluded that there is a leak at a position between i-1 and i,

Either a peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1,

Either a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak but not localized,

- if the cross-correlation function T, .ι, i has a peak corresponding to a position in i-1 and if the cross-correlation function

present :

• either no peak, or a peak at a position in i + 1, then it is concluded that there is no leak,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, or a main peak corresponding to a position at i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1,

Either a peak corresponding to a position at i, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position at i, then it is concluded that there is a leak but not localized, - if the cross-correlation function Fj.ι, i has a main peak corresponding to a position in i-1 and a secondary peak corresponding to a position in i and if the cross-correlation function ri, i + ι has:

• either no peak, or a peak corresponding to a position in i, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, then it is concluded that there is a leak in position i,

• either a peak corresponding to a position in i + 1, then it is concluded that there is a leak but not localized, • either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position between i and i + 1, i.e. a main peak corresponding to a position at i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that the presence of a leak at a position between i and i + 1,

- if the intercorrelation function ru, i has a main peak corresponding to a position in i-1 and a secondary peak corresponding to a position between i-1 and i and if the intercorrelation function

present :

• either no peak, or a peak corresponding to a position in i + 1, or a peak corresponding to a position in i, or a main peak corresponding to a position in i + 1 and a secondary peak in i, then it is concluded the presence of a leak at a position between i-1 and i,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then if the position of the peaks between i -1 and i and between i and i + 1 are equal to an error value, it is concluded that there is a leak at position i while otherwise it is concluded that there is a leak but not localized,

• either a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak but not localized. Another object of the invention is to propose a method making it possible to detect a source of noise originating from a leak or from a shock while background noise is present in the environment surrounding the pipeline.

To achieve such an objective, the method according to the invention consists in: - having at least a second series of at least two and preferably at least three vibration or acoustic wave sensors on the pipeline, neighbors from each other, separated from each other by known distances of _k ( _p , _q)

- And analyze the signals measured by these sensors to ensure the reduction of background noise as a function of the signal to noise ratio.

According to a preferred alternative embodiment, the method consists in extracting the signal and the noise by a method of sorting waves with limited conditioning consisting of:

- to estimate each frequency component of a vector X according to the following relation: X = (M ⁺ M) ^_I M ⁺ P, with P: vector of the pressures measured on the sensors,

M ⁺ : conjugate transpose of M, M: matrix of propagation operators,

- and to take into account the spectral component of the vector only if the conditioning of M ⁺ M is less than a limit value. According to a preferred alternative embodiment, the method consists in the case where the three sensors each have a bypass with respect to the pipe:

- to model each branch by estimating the frequency transfer function H (f) between the sensor and the input of the branch,

- and to take into account this transfer function in the method of sorting waves with limited conditioning.

According to another preferred embodiment characteristic, the method consists in taking into account the transfer function H (f) of each derivation by using a deconvolution method by reverse filtering of the "Wiener" type such that

H * (f) Y (f)

X (f) =

H * (f) H (f) + ε C (f) with

X (f): lead input Y (f): output of the derivation C (f): constraint operator ε: inversion regularization parameter Another object of the invention aims to propose a method making it possible to reduce the volume of data to be transmitted to a device processing measurement when the noise source corresponds to a leak.

To achieve such an objective, the object of the invention is to propose a method consisting in the case where the noise source corresponds to a leak:

- possibly ensuring Shannon filtering of the time signals measured by the sensors,

- And to ensure binarization of these signals before calculating the cross-correlation function between each pair of these signals thus obtained in order to reduce the volume of data to be processed.

Another object of the invention is to provide a method of detecting a noise source in a pipe making it possible to determine the temperature of the fluid or the internal pressure of the fluid flowing in the pipe.

To achieve such an objective, the method according to the invention consists:

- to measure resonance frequencies of at least one derivation of the sensors, by a spectral estimate, - to calculate the speed of the sound in the pipe,

- And from a characteristic curve of the fluid circulating inside the pipe, to deduce the temperature of the fluid if the pressure of the fluid is known or the pressure of the fluid if the temperature of the fluid is known.

Another object of the invention is to propose an installation allowing the implementation of the location detection method according to the invention.

Various other characteristics will emerge from the description given below with reference to the appended drawings which show, by way of nonlimiting examples, embodiments of the subject of the invention.

Fig. 1 is a diagram explaining the implementation of a detection and localization installation according to the invention.

Fig. 2 is a diagram explaining a part of the installation according to the invention. Figs. 3 to 9 are tables explaining a characteristic phase of the process making it possible to locate a leak.

As shown more precisely in fig. 1, the subject of the invention relates to a method and an installation 1 ensuring the detection and localization of a noise source S likely to appear on a pipe or a pipe 2 for transporting a fluid in the general sense. Such a transport pipe 2 can be made in any suitable manner for the transport of a fluid such as a liquid or a gas and be placed in the open air, buried or immersed.

In a first embodiment, the installation 1 comprises a first series of at least three vibration or acoustic wave sensors bearing the references, for example il, i, i + l. Each sensor is separated from each neighboring sensor, by a distance D _{k (} j _j , which may be of equal or different value for each pair of neighboring sensors thus formed. These sensors can be separated from each other by a distance typically of the order of a few kilometers or a few tens of kilometers. For example, each acoustic or vibratory magnitude sensor consists of an accelerometer, a hydrophone, a geophone, a microphone, etc.

The installation according to the invention 1 also comprises a device for measuring and processing the time signals Sj (t) delivered by each sensor. According to a preferred embodiment, each sensor is part of a station integrating a signal conditioning and processing unit, connected by radio or wire to a common control and processing center 5 making it possible to detect and locate a source noise. In a conventional manner, such a measurement and processing device comprises means for taking the time signals Sj (t) delivered by each sensor and each corresponding to the signal emitted by the noise source S, reaching said sensor with a time d 'arrival tj. In a conventional manner, the processing measurement device according to the invention comprises specific means of the programmed type whose main functionalities allow the implementation of the method which will be described in the following description.

According to an embodiment characteristic of the invention, the processing measurement device comprises means for determining from the sampled signals Sj (t) whether the noise source S corresponds to an impact on the pipe or to a leak from the pipe . In general, a shock constitutes a transient phenomenon while that a leak is characterized by a noise generally persistent over time. The signals measured by the sensors are analyzed so that a stationary signal in a characteristic frequency band can be estimated as corresponding to a leak, while a transient signal in a characteristic frequency band can be characterized as a shock.

According to a first aspect of the subject of the invention, the method according to the invention aims to make it possible to locate shocks occurring on the pipe 2, following for example a hammer blow, an explosion, d '' the impact of a projectile, the fall of an object, the impact of a public works vehicle, etc. Thus, in the case where the noise source S corresponds to a shock, the method according to the invention aims to determine a mathematical estimator and to minimize it in order to locate such a shock with a potential reduction of false alarms.

As shown more precisely in FIG. 2, this method is based on the use of the instants of arrival of the shock signals on at least the three consecutive stations n ° 1, n ° 2 and n ° 3 each comprising a sensor respectively i-1, i, i + 1 . The shock intervenes by hypothesis at time T in Xo between station n ° 1 and station n ° 3 by considering station n ° 2 as the intermediate station, with O≤Xo≤Dπ + D ₂₃ and with Ou e D ₂₃ positive . The location of the shock is for reasons of simplification given by its abscissa Xo measured from station n ° 1 and towards station n ° 3. We therefore have

where tj is the time of arrival of the shock measured at the level of station n ° i. The analytical calculation then provides the estimated values of the position X ₀ of the shock as well as the speed of the sound:

Y - Xn -Xn + C ^ -t X

Λ ₀ with C =!

² ('>-'_) The minimization method of a mathematical estimator is that used in the context of the invention.

The formula chosen below then provides estimated values of the position X ₀ of the shock as well as the speed of the sound C.

and:

( ^X 0, estimated • ^C estimated) ^{- ar} 6 min F (Λr ₀ .C)

. (o- ^c >

In the example illustrated in fig. 2, the first series of sensors comprises three stations, but it is clear that the object of the invention applies to n stations. In this case, the general estimator is such that:

F (X ₀ , C) = X IJCΔt _y - f (x „, D _{k (i; j)} ) j sensors (i, j)

with Δ t, _(j : the difference in arrival times between the sensors i, jf = a function of the geometric arrangement of the sensors, c = the speed of the sound. This mathematical estimator is minimized to determine the estimated values of the position Xo of the shock and the speed of the sound C as follows: {X _{t>, e} , _mée , C _estimated (X ₀ , C)].

This minimization can be carried out by optimization, that is to say by a numerical analysis of the grid, gradient, conjugate gradient, etc. type.

The advantages of this method of locating shocks at at least three positions are as follows: • the analytical method is very simple to implement and provides very good results in most cases and the method based on minimization of the mathematical estimator is also simple to implement and also provides very good results in most cases. • This last method rejects any false alarms. According to a preferred embodiment characteristic, the method according to the invention makes it possible to locate a source of noise corresponding to a leak while reducing false alarms.

According to such a characteristic, the method according to the invention consists in determining the reference sensor i having detected the noise source. The method then consists in calculating a first intercorrelation function rn, i between the signals emitted by the reference sensor i and a neighboring sensor called upstream i - 1 and a second intercorrelation function Ti ₊ i between the signals emitted by the reference sensor i and a neighboring sensor called downstream i + 1. The method then consists in searching in the intercorrelation functions rj.ι, i and Ty ₊ i for the emerging peaks whose number is between 0 and 2. This number peak can be 0 (no peak detected), 1 or 2 (typically a peak generated by the leak and a peak generated by a disturbing noise).

The method according to the invention consists in analyzing the position of the emerging peaks of the intercorrelation functions _7i_ι, i and F + i to reject the initial detection of the reference sensor i or on the contrary to locate the leak.

According to an embodiment characteristic, the method consists in locating the leak by analyzing the compatibility of the combinations of the peaks, possibly present in the intercorrelation functions F- and T-,, i ₊ ι. Figs. 3 to 9 represent in time space T the different possible combinations of the peaks emerging from the intercorrelation functions and the conclusions which should be deduced therefrom for the location of the leak.

The different possible positions for the emerging peaks of the intercorrelation function

are as follows: - no peak emerging in the intercorrelation function Γ, _J (fig. 3).

- a peak emerging from the intercorrelation function

corresponding to a position at i (fig. 4)

- a peak emerging from the intercorrelation function Ti-ι, i corresponding to a position at F, that is to say between i-1 and i (fig. 5), - a main peak emerging from the function d 'intercorrelation Fj.ι, i corresponding to a position at i and a secondary peak at F (fig. 6), - a peak emerging from the intercorrelation function Fi corresponding to a position at i-1 (fig. 7), a main peak from the intercorrelation function Fj_ι, i corresponding to a position at i-1 and a corresponding secondary peak at a position in i (fig. 8), - and a main peak of the intercorrelation function Ti-i corresponding to a position in i-1 and and a secondary peak corresponding to a position in F (fig. 9).

For each of these seven possible cases for the intercorrelation function F-ij, it is necessary to consider the seven possible cases for the intercorrelation function F ₊ i which are symmetrical.

Fig. 3 illustrates the case where the cross-correlation function r ^ u has no peak with the seven possible cases for the cross-correlation function

as stated above.

In case n ° 1, the intercorrelation functions do not have a peak so that there is no search for a secondary peak on these two functions. Initial detection is considered a false alarm so the leak is not confirmed.

In case n ° 2, a parasitic noise propagates from i + 1 to i but does not arrive on i-1. The peak positioned at i + 1 is rejected and the search for secondary peaks on

and rn _, i is negative. Detection is also considered a false alarm.

In case n ° 3, the leak is at F ', that is to say between i and i + 1 and its noise does not propagate until i-1. There is therefore no search for a secondary peak on the intercorrelation functions. A leak is validated in F '.

In case n ° 4, the leak is in F 'and its noise does not propagate in i-1. A parasitic noise propagates from i + 1 to i but does not arrive on i-1. The main peak positioned at i + 1 is rejected. The search for a secondary peak on

is positive while the search for a secondary peak on ru ,, - is negative. The location of a leak is validated in F '.

In case n ° 5, the leak is in i and its noise does not propagate in i-1. There is therefore no search for a secondary peak on the two intercorrelation functions. The leak at i is validated. In case n ° 6, the leak is in i and its noise does not propagate in i-1. A parasitic noise propagates from i + 1 to i but does not arrive on i-1. The main peak positioned at i + 1 is rejected. The search for a secondary peak on is positive, while the search for the secondary peak on -y is negative. The location of the leak is validated in i.

In case n ° 7, the leak is in i and its noise does not propagate in i-1. An unknown noise causes a peak in F '. There is no secondary peak research on intercorrelation functions. The leak is validated in i.

Fig. 4 illustrates the case where the cross-correlation function

presents a peak corresponding to a position in i with the seven possible cases for the intercorrelation function F, j ₊ ι.

In case n ° 8, the leak is in i and its noise does not propagate in i + 1. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in i. In case n ° 9, a parasitic noise propagates from i + 1 to i and i-1. However, a leak in i cannot be totally excluded. The main peak positioned at i + 1 is rejected, while a search for a secondary peak on the intercorrelation functions is negative. The detection of a leak is confirmed but its location is impossible. In case n ° 10, the leak is at F 'and its noise propagates on the three sensors. There is no secondary peak research on intercorrelation functions. The location of the leak is confirmed in F '.

In case n ° 11, the leak is at F 'and its noise eventually propagates on the three sensors. A parasitic noise propagates in the direction i + 1 towards i and can be up to i-1. The main peak positioned at i + 1 is rejected. The search for the secondary peak on, j ₊ ι is positive, while the search for a secondary peak on F _-1; j is negative. The location of the leak is confirmed in F '.

In case n ° 12, the leak is at i and its noise propagates on the three sensors. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in i.

In case n ° 13, the leak is at i and its noise propagates on the three sensors. A parasitic noise propagates in the direction i + 1 towards i. The main peak positioned in i + 1 is rejected. The search for the secondary peak on F, j ₊ ι is positive while the search for the secondary peak on Fi is negative. The location of the leak is validated in i.

In issue 14, the leak is at i and its noise propagates on the three sensors. An unknown noise causes a peak in F '. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in i.

Fig. 5 illustrates the case where the cross-correlation function -y has a peak corresponding to a position at F, that is to say between i-1 and i, combined with the seven possible cases for the cross-correlation function F, j ₊ ι. In case n ° 15, the leak is at F and its noise does not propagate at i + 1. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in F.

In case n ° 16, the leak is at F and a parasitic noise propagates in the direction i + 1 towards i without reaching i-1. The peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation functions is negative. The location of a leak is validated in F.

In case 17, the leak is close to i. There is no secondary peak research on intercorrelation functions. Two different results therefore appear, which can come from propagation speed errors. The ambiguity is removed by comparing the distances from the two peaks. If the position of the peaks between i-1 and i and between i and i + 1 are equal to an error value, the location is validated in position i. Otherwise, it is concluded that there is a leak but not localized.

In case n ° 18, the leak is close to i and a parasitic noise propagates in the direction i + 1 towards i. The main peak positioned at i + 1 is rejected. The search for a secondary peak on F, i ₊ ι is positive while the search for a secondary peak on F- is negative. The solution is reduced to case n ° 17. The leak location is validated in i, if the position of the peaks between i-1 and i and between i and i + 1 are equal to an error value. Otherwise, it is concluded that there is a leak but not localized.

In case n ° 19, the leak is at F and spreads over the three sensors. This case is symmetrical to the case in No. 10. There is therefore no search for a secondary peak on the intercorrelation functions. The location of the leak is validated in F. In case n ° 20, the leak is at F and a parasitic noise propagates in the direction i + 1 towards i. The main peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation function

is positive while the search for a secondary peak on -ι, i is negative. The location of the leak is confirmed in F. In case n ° 21, the leak is in F and its noise propagates on the three sensors.

An unknown noise causes a peak in F '. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in F.

Fig. 6 illustrates the case where the intercorrelation function F-ι, i has a main peak corresponding to a position at i and a secondary peak corresponding to a position between i-1 and i with the seven possible cases for the intercorrelation function F, i ₊ ι.

In case n ° 22, the leak is in i and its noise does not propagate on i + 1. An unknown noise causes a peak far from i. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in i.

In case 23, the leak is at F and a parasitic noise propagates on the three sensors from i + 1 to i-1. The peak positioned at i + 1 is rejected. The search for a secondary peak on F, j ₊ ι is negative, while the search for a secondary peak on.] ₍ I is positive. The location of a leak is validated in F. In case n ° 24 , the leak is at F 'and an unknown noise causes a peak at F. There is no search for a secondary peak on the intercorrelation functions.The location of the leak is validated at F'.

In case 25, the location is uncertain. The main peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. A leak detection is confirmed but its location is impossible.

In case n ° 26, the leak is at i and an unknown noise causes a peak at F. There is no search for a secondary peak on the intercorrelation functions. The location of the leak is validated in i.

In case n ° 27, the leak is at F and its noise propagates on the three sensors. A parasitic noise propagates in the direction i + 1 towards i. The main peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. The location of a leak is validated in F. In case 28, the leak is at i. There is no secondary peak research on intercorrelation functions. The location of the leak is validated in i.

Fig. 7 illustrates the case where the cross-correlation function .ι _, i has a peak corresponding to a position at i-1, with the seven possible cases for the cross-correlation function, j ₊ ι.

In case n ° 29, a parasitic noise propagates from i-1 to i but does not arrive on i + 1. The peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is negative. It is therefore considered that the detection was a false alarm. Case no. 30 is unlikely since the noises propagate from i-1 to i and from i + 1 to i. The peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation functions is negative. Detection is therefore considered a false alarm.

In case n ° 31, the leak is at F 'and a parasitic noise propagates in the direction from i-1 to i without reaching i + 1. The peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is negative. The location of the leak is confirmed in F '.

In case n ° 32, the leak is at F '. This case is unlikely since the noises propagate from i-1 to i and from i + 1 to i. The peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation function

is positive, while the search for a secondary peak on F-ι, i is negative. The location of a leak is validated in F '.

In case 33, detection is a priori due to noise outside the inspected pipe sector. An i-leak cannot however be completely excluded. The peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is negative. The detection of a leak is confirmed but its location is impossible.

In case 34, detection is a priori due to noise outside the sector of the pipe inspected. An i-leak cannot be completely ruled out. The peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation function F, i ₊ ι is positive, while the search for a secondary peak on the cross-correlation function is negative. The detection of a leak is therefore confirmed, but its location is impossible.

In case n ° 35, the leak is at F 'and its noise does not propagate on the three sensors. Spurious noise propagates in the i-1 to i direction on the three sensors. The peak positioned at i-1 is rejected. The search for a secondary peak on the cross-correlation function F.ι, j is negative, while the search for a secondary peak on

Fy ₊ i is positive. The location of a leak is validated in F '.

Fig. 8 illustrates the case where the intercorrelation function ru, has a main peak at i-1 and a secondary peak corresponding to a position at i, with the seven possible cases for the intercorrelation function F-, i + ι •

In case n ° 36, the leak is in i and its noise does not propagate in i + 1. A parasitic noise propagates from i-1 to i without arriving on i + 1. The main peak positioned at i-1 is rejected. The search for a secondary peak on the cross-correlation function F-ι, i is positive, while the search for a secondary peak on the cross-correlation function F, i ₊ ι is negative. This case is symmetrical to case n ° 6. The location of a leak is therefore validated in i.

In case no. 37, detection is a priori due to noise outside the sector of the pipe inspected. An i-leak cannot be completely ruled out. The peak positioned at i-1 is rejected. The search for a secondary peak on the cross-correlation function -ι, j is positive, while the search for a secondary peak on the cross-correlation function Ty ₊ i is negative. We then find ourselves in a symmetrical case in case n ° 34 for which the location is uncertain. But the leak detection is confirmed.

In case n ° 38, the leak is at F 'and a parasitic noise propagates in the direction i-1 towards i. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation function F-ι, i is positive, while the search for a secondary peak in the intercorrelation function, i + ι is negative. This case is symmetrical in case n ° 20. The location of the leak is validated in F '.

In case n ° 39, the leak is at F 'and two parasitic noises propagate in the directions i-1 towards i and i + 1 towards i. The main peak positioned at i + 1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. The location of a leak is therefore validated in F '. In case n ° 40, the leak is in i. Its noise propagates on the three sensors.

Spurious noise propagates in the i-1 towards i direction. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation function

is positive, while the search for a secondary peak on the intercorrelation function F, i ₊ ι is negative. The location of the leak is validated in i.

In case 41, it is unlikely to find an absence of leakage in i. Indeed, it would require a very particularly unfavorable configuration of the background noise. A leak was detected in i only. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. The location of the leak is validated in i.

In case n ° 42, the leak is at F 'and a parasitic noise propagates from i-1 to i on the three sensors. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. This case is symmetrical in case n ° 27. The location is validated in F '. Fig. 9 illustrates the case where the cross-correlation function

has a main peak corresponding to a position in i-1 and a secondary peak corresponding to a position in F, with the seven possible cases for the intercorrelation function F, i ₊ ι-

In case 43, the leak is at F and its noise does not propagate at i + 1. Spurious noise propagates from i-1 to i without reaching i + 1. The main peak positioned at i-1 is rejected. The search for a secondary peak on the cross-correlation function rj.ι, i is positive while the search for a secondary peak on the cross-correlation function ₊ i is negative. This case is symmetrical to case n ° 4. The location of the leak is validated in F.

In case n ° 44, the leak is at F which is unlikely since the noises propagate from i-1 to i and from i + 1 to i. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation function F- is positive, while the search for a secondary peak on the intercorrelation function F, i ₊ ι is negative. This case is symmetrical in case n ° 32. The location is validated in F. In case n ° 45, the leak may be close to i. Spurious noise propagates in the i-1 towards i direction. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation function Γ _M ^ is positive while the search for a secondary peak on the cross-correlation function _, j ₊ ι is negative. This case is symmetrical to case n ° 18. A leak is therefore considered in i if the position of the peaks between i-1 and i and between i and i + 1 are equal to an error value. Otherwise, it is concluded that there is a leak but not localized. In case 46, the leak may be close to i. Two parasitic noises propagate in the directions i-1 towards i and i + 1 towards i. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. This case is identical to case n ° 17. If the position of the peaks between i-1 and i and between i and i + 1 are equal to an error value, it is concluded that there is a leak at the position i. Otherwise, it is concluded that there is a leak but not localized.

In case n ° 47, the leak is at F and a parasitic noise propagates in the direction i-1 towards i. The main peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation function F- is positive, while the search for a secondary peak on the intercorrelation function

is negative. The location of a leak is validated in F.

In case n ° 48, the peak positioned at i-1 and the one positioned at i + 1 are rejected. The search for a secondary peak on the intercorrelation functions is positive. This case is symmetrical in case 39. The location of a leak is validated in F. In case 49, the peak positioned at i-1 is rejected. The search for a secondary peak on the intercorrelation functions is positive. The location is uncertain. This case is symmetrical to case n ° 25. It is therefore concluded that there is a leak but not localized.

As shown in Figs. 3 to 9, the method according to the invention therefore makes it possible to locate a mite or to reject the initial detection by analyzing the compatibilities of the different combinations of the peaks possibly present in the intercorrelation functions. This process thus makes it possible to confirm or not the initial detection of a leak by a sensor, and then to locate the leak.

Another object of the invention is to improve the detection of signals emitted by measurement sensors when background noise is present in the surrounding environment. The objective is therefore to extract the relevant signals (leaks or shocks) from the background noise. To achieve such an objective, the method according to the invention aims to have on the line 2 at least a second series of at least two and preferably at least three sensors 7 of vibration or acoustic waves, mounted close to each other. By neighbor, it must be considered that the vibration or acoustic wave sensors 7 are separated by known distances d _{(P) q)} typically between a few tens of centimeters and a few tens of meters, or even more.

The method according to the invention then consists in analyzing the signals measured by these sensors 7 to ensure the reduction of the background noise as a function of the signal to noise ratio.

According to a preferred embodiment characteristic, the method consists in extracting the signal and the noise by a method of sorting waves with limited conditioning consisting of:

- to estimate each frequency component of a vector X according to the following relation: X = (M ⁺ M) ^_1 M ⁺ P, with

P: vector of the pressures measured on the sensors, M ⁺ : conjugate transpose of M,

M: matrix of propagation operators,

- and to take into account the spectral component of the vector X only if the conditioning of M ⁺ M is less than a limit value.

This method consists in taking into account in the reconstruction of the signals only the frequencies for which there is a high reliability of the inversion result. This method proves to be robust and stable in most cases encountered in practice for which the sensors are placed near a source of parasitic noise.

According to an embodiment characteristic, the sensors 7 of the second series are each mounted on a branch 8 of the pipe 2. In the case where these sensors 7 are mounted on such a branch 8, provision is made to take this branch 8 into account. in the limited conditioning wave sorting method. According to this characteristic, the method consists in modeling each derivation 8 by estimating the frequency transfer function H (f) between the sensor 7 and the input of the derivation 9 and in taking this transfer function into account in the sorting method waves with limited conditioning. Preferably, the method consists in taking into account the transfer function H (f) of each derivation 8 using a deconvolution method by reverse filtering of the "Wiener" type such as:

H * (f) Y (f) H * (f) H (f) + ε C (f) with X (f): input of the derivation

Y (f): output of the derivation C (f): constraint operator ε: regularization parameter of the inversion. Another characteristic of the invention is to propose a method making it possible to reduce the volume of data to be transmitted in the case where the noise source corresponds to a leak.

To this end, a one-bit correlation method is implemented which makes it possible to envisage a reduction in the sampling frequency (decimation factor).

According to this method, the time signals measured by the two sensors surrounding the noise source are taken into account. Optionally, a pre-filtering of these signals is carried out.

The method then optionally consists in ensuring Shannon filtering of the signals measured by the sensors via the implementation of a spectrum anti-aliasing filter. The method then consists in ensuring binarization of these signals by therefore proceeding to one-bit coding. This coding consists in characterizing the dynamic variations of this signal around its point of rest in terms of positive or negative signs. The method then consists in calculating the cross-correlation function between these signals thus obtained in order to reduce the volume of the data to be processed.

This decimation is possible without deteriorating performance in terms of location accuracy by one-bit cross-correlation method. Furthermore, the anti-aliasing filtering of the Shannon spectrum also makes it possible to obtain good results in terms of absolute value of the correlation coefficient. Without such filtering, it is observed in most cases, a degradation of this value. A decimation by a factor of 5 to 10 and the correlation to a bit can for example generate a gain of a factor of 100 on the volume of data to be processed.

According to another aspect of the invention, the object of the invention is to propose a method making it possible to deduce a measurement of the temperature of the fluid or of the internal pressure of the fluid circulating inside the pipe 2. According to a such a method, provision is made to measure the resonance frequencies of at least one derivation 8 of the sensors 7 by a spectral estimation. The speed of the sound is then calculated in the pipe 2. The method then consists in starting from a characteristic curve of the fluid circulating inside the pipe 2 to deduce the temperature of the fluid if the pressure of the fluid is known, or the fluid pressure if the fluid temperature is known.

The invention is not limited to the examples described and shown since various modifications can be made without departing from its scope.

Claims

1 - Method for detecting and locating at least one noise source (S) in a pipe (2) conveying a fluid, characterized in that it consists:

- to have on the pipe (2) a first series of at least three vibration or acoustic wave sensors separated from each other by known distances D _k (i, j>,

- taking the time signals S, (t) delivered by each sensor and each corresponding to the signal emitted by the noise source reaching said sensor with an arrival time tj, - determining from the signals picked up if the noise source corresponds to an impact on the pipeline or a leak from the pipeline,

- and in the case where the noise source corresponds to a shock:

• to determine a mathematical estimator such as

F (X ₀ , C) = _ £ [cΔt ^ - f (x ₀ , D _{k (i)} ) j sensors (i, j)

with Δ t; _{; j} : the difference in arrival times between the sensors i, jf = a function of the geometric arrangement of the sensors, c = the speed of the sound, • and minimize the mathematical estimator F to determine the estimated values of the position Xo shock and speed of sound

C as follows:

(o _, ^ _> C _es ,,,) = arg [min F (N ₀ , C) 1.

2 - Method according to claim 1, characterized in that it consists, in the case where the noise source corresponds to a leak:

- determining the reference sensor i that has detected the noise source,

- calculating a first intercorrelation function F.ι, j between the signals emitted by the reference sensor i and a neighboring sensor called upstream i - 1 and a second intercorrelation function F, i ₊ ι between the signals emitted by the reference sensor i and a neighboring sensor called downstream i + 1, - to search in the interconnection functions F.ι, i and F, i + ι for emerging peaks whose number is between 0 and 2,

- And to analyze the position of the emerging peaks of the intercorrelation functions F-ι, i and F, i + ι to reject the initial detection of the reference sensor i or on the contrary to locate the leak.

3 - Method according to claim 2, characterized in that it consists in locating the leak by analyzing the compatibility of the combinations of the peaks, possibly present in the intercorrelation functions F.ι, j and F, j ₊ ι.

4 - Method according to claim 2 or 3, characterized in that it consists in analyzing the compatibility of the emerging peaks of the cross-correlation functions Γ _J. _I , _J and

T, - _, i ₊ ι. As follows:

- if the cross-correlation function rn, i has no peak and if the cross-correlation function T [, j ₊ ι has:

• either no peak, or a peak corresponding to a position in i + 1, then it is concluded that there is no leak,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that d '' a leak at a position between i and i + 1, • either a peak corresponding to a position at i, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position at i, or a peak main corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak in position i,

- if the cross-correlation function F.ι, i has a peak corresponding to a position at i and if the cross-correlation function F, j _{+ 1} has:

• either a peak corresponding to a position in i + 1, then it is concluded that there is a leak but not localized, • either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1,

- if the cross-correlation function F-ι, i has a peak corresponding to a position between i-1 and i and if the cross-correlation function T ^ ₊ has:

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, then if the position of the peaks between i -1 and i and between i and i + 1 are equal to an error value, it is concluded that there is a leak at position i, while otherwise it is concluded that there is a leak but not localized,

- if the cross-correlation function -ι, i has a main peak corresponding to a position in i and a secondary peak corresponding to a position between i-1 and i and if the cross-correlation function F, i + ι has:

• either no peak, or a peak corresponding to a position in i, or a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak in position i,

• either a peak corresponding to a position in i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, then it is concluded that there is a leak at a position between i-1 and i,

• either a peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1, • or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak but not localized,

- if the cross-correlation function F-ι, j has a peak corresponding to a position in i-1 and if the cross-correlation function

presents: • either no peak, or a peak corresponding to a position in i + 1, then it is concluded that there is no leak,

Either a peak corresponding to a position at i, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position at i, then it is concluded that there is a leak but not localized,

- if the cross-correlation function rn, ι has a main peak corresponding to a position in i-1 and a secondary peak corresponding to a position in i and if the cross-correlation function F, i ₊ ι has: • either no peak , either a peak corresponding to a position in i, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position in i, then it is concluded that there is a leak in position i,

• either a peak corresponding to a position in i + 1, then it is concluded that there is a leak but not localized,

• either a peak corresponding to a position between i and i + 1, or a main peak corresponding to a position in i + 1 and a secondary peak corresponding to a position between i and i + 1, i.e. a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak at a position between i and i + 1, - if the cross-correlation function F.ι, i has a main peak corresponding to a position in i-1 and a secondary peak corresponding to a position between i-1 and i and if the function d 'intercorrelation F, i ₊ ι presents:

Either no peak, or a peak corresponding to a position at i + 1, or a peak corresponding to a position at i, or a main peak corresponding to a position at i + 1 and a secondary peak corresponding to a position at i, then it is concluded that there is a leak at a position between i-1 and i,

• either a main peak corresponding to a position in i and a secondary peak corresponding to a position between i and i + 1, then it is concluded that there is a leak but not localized.

5 - Method according to one of claims 1 to 4, characterized in that it consists:

- to have on the pipe (2), at least a second series of at least two and preferably at least three sensors (7) of vibration or acoustic waves, adjacent to each other, separated from each other others by known distances of _k ( _P , _q )

6 - Method according to claim 5, characterized in that it consists in extracting the signal and the noise by a method of sorting waves with limited conditioning consisting of:

- to estimate each frequency component of a vector X according to the following relation: X = (M ⁺ M) ^" 'M ⁺ P, with P: vector of the pressures measured on the sensors,

M ⁺ : conjugate transpose of M,

M: matrix of propagation operators,

- and to take into account the spectral component of the vector only if the conditioning of M ⁺ M is less than a limit value.

7 - Method according to claim 5 or 6, characterized in that it consists, in the case where the three sensors (7) each have a bypass (8) relative to the pipe (2):

- to model each derivation (8) by estimating the frequency transfer function H (f) between the sensor and the input of the derivation,

8 - Method according to claim 7, characterized in that it consists in taking into account the transfer function H (f) of each derivation (8) using a deconvolution method by reverse filtering of the "Wiener" type such as

H * (f) Y (l) H * (f) H (f) + ε C (f) with

X (f): input of the derivation Y (f): output of the derivation C (f): constraint operator ε: inversion regularization parameter

9 - Method according to claim 3, characterized in that it consists in the case where the noise source corresponds to a leak:

10 - Method according to claim 7 or 8, characterized in that it consists: - in measuring resonance frequencies of at least one derivation (8) of the sensors, by a spectral estimation, - to calculate the speed of sound in the pipe (2),

- And from a characteristic curve of the fluid circulating inside the pipe, to deduce the temperature of the fluid if the pressure of the fluid is known or the pressure of the fluid if the temperature of the fluid is known. 11 - Installation for detecting and locating at least one noise source (S) in a pipe (2) carrying a fluid, characterized in that it comprises:

a first series of at least three vibration or acoustic wave sensors each separated from each other by known distances D _{k (i} , j),

- and a measuring and processing device comprising:

• means for picking up the time signals Si (t) delivered by each sensor and each corresponding to the signal emitted by the noise source reaching said sensor with an arrival time t ;, • means for determining from the signals picked up if the noise source corresponds to a shock on the pipe or to a pipe leak,

• and calculation means allowing in the case where the noise source corresponding to a shock: * to determine a mathematical estimator such that

F (X „0 = ∑ [cΔt _{i; j} - f (x ₀ , D _{k (i> j)} ) j sensors (i, j)

with Δ ti, _j : the difference in arrival times between the sensors i, j

* minimize the mathematical estimator F to determine the estimated values of the position Xo of the shock and the speed of the sound C in the following manner:

(X _θ , es, _im é> C estimated) = ^{al *} g "Mil F (X ₀ , C)

| _v ^Λ o, c ⁾ 12 - Installation according to claim 11, characterized in that the calculation means allow, in the case where the noise source corresponds to a leak: To determine the reference sensor i having detected the noise source,

• to calculate a first intercorrelation function

between the signals emitted by the reference sensor i and a neighboring sensor called upstream i - 1 and a second intercorrelation function F, i ₊ ι between the signals emitted by the reference sensor i and a neighboring sensor called downstream i + 1 ,

• to search in the intercorrelation functions F-ι, i and F, i ₊ ι for the emerging peaks whose number is between 0 and 2, • to analyze the position of the emerging peaks for the intercorrelation functions -i and

to reject the initial detection of the reference sensor i or on the contrary to locate the leak.

13 - Installation according to claim 12, characterized in that the calculation means locate the leak by analyzing the compatibility of the combinations of the peaks, possibly present in the intercorrelation functions Fu and F, i ₊ ι-

14 - Installation according to one of claims 11 to 13 characterized in that it comprises at least a second series of at least two and preferably at least three sensors (7) of vibration or acoustic waves neighboring them from each other on the pipe (2) being separated from each other by known distances of _k ^ _q) , and the signals of which are picked up by the measuring and processing device which includes signal analysis means measured by these sensors to ensure the reduction of background noise as a function of the signal to noise ratio.

15 - Installation according to claim 14, characterized in that the measurement and processing device comprises means for extracting the signal and the noise by a wave sorting method with limited conditioning.

16 - Installation according to claim 14 or 15, characterized in that the three sensors (7) of the second series are each mounted on a branch (8) relative to the pipe (2) and in that the measuring device and processing includes: • means for modeling each derivation by estimating the frequency transfer function H (f) between the sensor and the input of the derivation (8), • and means to take into account this transfer function in the method of sorting waves with limited conditioning.