EP1966580A2

EP1966580A2 - Detection of a fluid leakage in a circuit by intercorrelation

Info

Publication number: EP1966580A2
Application number: EP06847154A
Authority: EP
Inventors: Jaouad Zemmouri; Bruno Garnier
Original assignee: Optical System and Research for Industry and Science OSYRIS SA
Current assignee: Optical System and Research for Industry and Science OSYRIS SA
Priority date: 2005-12-28
Filing date: 2006-12-28
Publication date: 2008-09-10
Also published as: FR2895508A1; WO2007074244A3; WO2007074244A2; FR2895508B1

Abstract

The inventive method for automatically controlling a circuit containing a fluid flowing therein consists (a) in detecting the flowing fluid by means of input and output sensors which are positioned at the circuit input and output, respectively, and transmit two respective measuring signals E(t) and S(t) characteristic for the fluid flow or flow variation, (b) in processing the measuring signals E(t) and S(t) for detecting, where appropriate, a fluid leakage in the circuit, in carrying out an offset adjustment by retreating from the measuring signals E(t) and S(t) the mean values (Emoy; Smoy) thereof and in calculating the intercorrelation of the thus corrected two measuring signals. More particularly, in an embodiment, the inventive method consists in detecting central intercorrelation peaks (P), in comparing the amplitude variations of the central intercorrelation peak (P) with at least one predefined threshold and in detecting the leak appearance when said amplitude variation absolute value is negative and greater than said threshold.

Description

DETECTION OF A FLUID LEAK IN A CIRCUIT BY

correlation

Technical Field The present invention relates to the automatic detection of fluid leaks (liquid or gas) in a circuit. It relates more particularly, but not exclusively, to the detection of a leakage of fluid that may have a very low flow rate in a circuit that may be subject to disturbances, in particular thermal disturbances, which have an influence on the flow rate of the fluid transported. For example, the invention advantageously allows a reliable detection and with a very good sensitivity of the appearance of a leak in a coolant circuit, such as a fluid circuit used to cool or to heat an enclosure. One of the preferred, but not exclusive, applications of the invention lies in the detection of water leaks in the cooling circuits of the enclosures of electric arc furnaces used in the iron and steel industry for the melting of scrap or the refining of steels. Another application of the invention is the detection of leaks in pipelines for transporting and / or distributing fluids (pipelines, pipelines, etc.).

A method widely used to date to detect a leakage of fluid in a circuit, such as for example a cooling circuit or heating an enclosure, consists in placing two flowmeters respectively at the output and the input of the circuit, and continuously calculating, by means of these two flowmeters, the difference between the inlet fluid flow rate and the output fluid flow rate. A leak is detected when this difference in flow rates exceeds a predefined threshold.

This method, hereinafter referred to as the "flow differential method," has at least three major disadvantages. It is limited by the accuracy of the measurement of the flowmeters used

( ^1st disadvantage). Less the accuracy of the flow meters used is good, and the higher the flow rate of the fluid leak must be important for it to be detected. For example, and as an indication, the currently marketed electromagnetic flowmeters have an accuracy of the order of 0.3% to 1%. Thus, with this type of flow meter, the differential flow rate method can best be used today to detect leak rates that are at least greater, depending on the case, from 0.6% to 2% of the flow rate of the fluid. in the circuit.

Also, any disturbance of the circulating fluid resulting in a change in the volume of transported fluid is likely to falsify the detection (second disadvantage). It is for example and mainly a thermal disturbance (heating or cooling of the fluid), or a hydraulic disturbance (for example closure or opening of a pipe in a circuit forming a more or less complex network of several pipes in parallel). It is understood that in the case of an increase in the volume of fluid transported, for example under the effect of an increase in the temperature of the fluid between its input and its output of the circuit, the output flow rate becomes greater than the flow rate. input, which prevents the detection of any leak below this increase in flow. Conversely, in the case of a reduction in the volume of fluid transported, for example under the effect of a sharp drop in the temperature of the fluid between its inlet and its outlet of the circuit, the output flow rate becomes lower than the input flow, which can trigger a false leak detection, while no leak is present.

The presence of EMI can also disrupt the flow measurement signals, and thereby distort the leak detection ^(3rd drawback).

In general, the three aforementioned drawbacks make this flow differential method unreliable and, where appropriate, unsuitable for detecting fluid leaks having a very low flow rate and / or leakage of a fluid subjected to disturbances, in particular thermal, hydraulic, affecting the volume of fluid transported, or electromagnetic.

However, there are numerous industrial applications in which a fluid is circulating in a circuit, and it is necessary to reliably and sufficiently early detect any leak, even very low, and moreover in an environment in which the fluid can to undergo non-stationary disturbances, and in particular more or less significant thermal disturbances over time.

Among these many industrial applications, we can particularly mention all industrial applications in which a fluid flows in a cooling circuit or in a heating circuit of an installation.

More particularly, in the field of iron and steel, it is usual to cool the tank and more specifically the vault and the panels of arc furnaces, which are used for the melting of scrap or the refining of steels. In this type of furnace, the furnace tank is charged with the various ferrous materials to be melted, then the temperature of the scrap is raised to their melting point (typically between 1500 ^° C. and 2000 ° C.) mainly by means of electric arcs generated inside the tank. The molten steels are then poured out of the furnace for further processing.

The cooling of this type of furnace is obtained by means of at least one cooling circuit formed by a more or less complex network and more or less long tubing, inside which is circulated a liquid heat transfer fluid (generally water) which allows to effectively cool the tubings. Typically, an arc furnace comprises in practice several independent cooling circuits, generally including at least one cooling circuit for the side panels of the tank and a cooling circuit for the vault of the furnace tank. This type of steel furnace is particularly vulnerable to liquid leaks in its cooling circuit, for at least two reasons:

- if due to a leak in the cooling circuit, water flows inside the oven, it can reach the bottom of the furnace which is lined with refractory bricks; in this case, the refractory bricks deteriorate very rapidly, and this deterioration results in a piercing of the bottom of the furnace tank; in this case, the molten metal escapes out of the tank with potentially serious human and financial consequences;

- if, because of a leak in the cooling circuit, water falls inside the tank of a furnace in operation, and comes into contact with the molten metal, it can decompose into oxygen and hydrogen under certain conditions because of the very high temperature. This decomposition generates a significant risk of accidental explosion. It is therefore understandable that in this particular type of application the stakes for finding an automatic leak detection that is reliable and early are extremely important, both on the human level to avoid accidents that can be fatal in certain cases, and on the plane. to avoid long and costly stopping of the furnace in the event of an accident.

Today, the leak detection methods that are implemented industrially on these steel furnaces are essentially methods based on a measurement of the differential of the flow rates of entry and exit, and therefore do not give entirely satisfactory considering the disadvantages mentioned above inherent to this type of detection method.

To try to correct the effects of thermal disturbances, the volume of fluid transported, and try to overcome the aforementioned problems of false detection or missed detection that may result, improvements of the aforementioned method of differential flow rates have been proposed to this day. These improvements are based on a correction of the flowmeter measurements, in particular from the temperature or the pressure of the fluid flow.

Temperature correction solutions are proposed by the manufacturers of flow meters, and are implemented in particular in the field of arc furnaces mentioned above. For example also, in another field of application described in the French patent application FR 2,509,839, there is provided a method for detecting nitrogen leakage in a circuit forming a barrier between the primary and secondary spaces for thermal insulation. a cryogenic tank, said method being based on the differential nitrogen flow rate method at the outlet and the inlet of the circuit. In this publication, it is taught to provide temperature corrections and pressure corrections to flow measurements by measuring the temperature and pressure of the flowing fluid.

On the one hand this type of corrections is ultimately not very reliable and also imposes a calibration of temperature and pressure measurements that can be difficult to achieve, and must be performed regularly. On the other hand and above all, these corrections of temperature or pressure do not make it possible to overcome the first disadvantage mentioned above related to the low measurement accuracy of the flowmeters. It is furthermore proposed in the European patent application EP

0 188 911 a detection method which is similar to the aforementioned method of the flow differential, and which in this document is used to detect a leakage of fluid, and in particular a gas leak, in a pipe of the pipeline type. This method is essentially based on the calculation of the FD parameter below:

FD

Q _1n is the amount of gas entering the pipeline and Q _out being the amount of gas leaving the pipeline. In this publication, it is also proposed, to reduce the errors on the measurement signal and improve the accuracy of the measurement, to increase the integration time and to remove the average value of the input signal over the period of time. integration by calculating the following FD * parameter:

The calculation of the parameter FD or the parameter FD * has all the aforementioned drawbacks of the differential flow rate method.

Moreover, in this publication, it is taught to detect a leak when the calculated parameter (FD or FD *) exceeds a threshold value calculated by simulation. This simulation takes into account many parameters including the diameter of the pipe in which the fluid circulates, and any changes of section, the pressure of the fluid flow, fluid friction, gravity, etc.

This method of detection by simulation has the disadvantage of being very complicated, and unsuitable for complex fluid circuits comprising several possible fluid paths, such as for example the cooling circuits of iron-arc furnaces. is performed only at the end of the integration period, which disadvantageously induces a delay in the detection. But in a prejudicial way this delay is all the more important as the period of integration chosen is great for reasons of precision of the measurement.

It has also been proposed in German patent application DE 26 03 715 to perform a leak detection, by measuring, for example, the pressure at the inlet of a circuit and the pressure at the outlet of the circuit, by withdrawing at each measurement. input and output the same reference value, which is calculated from the input and output signals, and calculating a cross-correlation of the signals. On the one hand, this publication does not describe the means by which inter-correlation automatically detects a leak. On the other hand, it seems that this method does not allow to obtain a sufficient sensitivity of detection, and at the Applicants know this relatively old method (published in 1977) would never have been used industrially.

It has also been proposed to date to locate a leakage of fluid in a circuit, by detecting the noise of the leakage by means of three acoustic sensors (which do not make it possible to detect the flow or the variations of flow of the fluid in the circuit), by calculating intercorrelations of the acoustic measurement signals, and by determining the physical position of the leak from the relative positions of the larger amplitude intercorrelation peaks. This method of locating a leak is described, for example, in US Pat. No. 5,544,074. This method has at least two major disadvantages. It can be implemented only in a simple fluid circuit having a single path for the fluid, and this leak detection is unusable in circuits subjected to strong acoustic disturbances. Objectives of the invention

The present invention has the general objective of proposing a new technical solution for automatically detecting leakage of a fluid (liquid and / or gas) flowing in a circuit.

Another more particular object of the invention is to propose a new technical solution which makes it possible to automatically detect leakage of a fluid (liquid and / or gas) flowing in a circuit, with improved detection sensitivity.

Another more particular objective of the invention is to propose a new technical solution which makes it possible to automatically detect a leak of a fluid (liquid and / or gas) flowing in a circuit subjected to disturbances (in particular thermal disturbances and / or hydraulic systems) affecting the volume of the flowing fluid.

Another more particular object of the invention is to propose a new technical solution which makes it possible to automatically detect a leakage of a fluid (liquid and / or gas) in flow in a circuit subjected to electromagnetic disturbances. Another more particular object of the invention is to propose a new technical solution which makes it possible to automatically detect a leakage of a fluid (liquid and / or gas) flowing in a circuit subjected to acoustic disturbances. Another more particular object of the invention is to propose a new technical solution which makes it possible to automatically detect a leakage of a fluid (liquid and / or gas) in flow in a complex circuit comprising several different possible paths for the flowing fluid. . Summary of the invention

All or some of these objectives, including at least the general objective mentioned above, are achieved by the invention, whose first object is a method of controlling a circuit in which a fluid circulates, said method comprising the following steps: a) the circulating fluid is detected by means of an input sensor and an output sensor respectively positioned at the input and at the output of the circuit, and respectively delivering two measurement signals E (t) and S (t) characteristics of the flow or fluid flow variations,

(b) the two measurement signals E (t), S (t) are processed for the purpose of detecting, if appropriate, a fluid leak in the circuit.

According to a first aspect of the invention, in step (b) of processing the signals E (t) and S (t), an offset correction is performed by removing the measurement signals [E (t) or E _k ; S (t) or S _k ] their mean value (Emoy;

Smoy), and. an intercorrelation of the two corrected measurement signals is calculated.

More particularly, the method comprises the following additional and optional features, taken alone or, where appropriate, in combination: in step (b) of processing the signals E (t) and S (t), the measurement signals E (t) and S (t) are sampled, and the cross-correlation function Icc is a function discrete calculated using the following formula:

or :

E _k are the values resulting from the sampling of the measurement signal E (t);

S _k are the values resulting from the sampling of the measurement signal S (t); n is the total number of samples E _k and S _k in a predefined calculation window; k is an integer from 0 to (n-1); j is an integer taking the values between - (n-1) and (n-1);

Iccj is a vector of size 2n-1;

in an alternative embodiment, in step (b) for processing the measurement signals, an intercorrelation of the two measurement signals is calculated, minus the autocorrelation of the output measurement signal;

in an alternative embodiment, in step (b) of processing the measurement signals, the amplitude variations of the central peak (P) of intercorrelation are detected; more particularly, the variations of the amplitude of the central cross-correlation peak (P) are compared with at least one predefined threshold, and the appearance of a leak is detected when said variation of the amplitude is negative and becomes, in absolute value, greater than this threshold;

said threshold is preferably auto-adaptive and is calculated as a function of the amplitudes of the previous central cross-correlation peaks; in another variant embodiment, in step (b) of treatment measurement signals, the integral of the inter-correlation is calculated, and the result of this integral is compared with a predefined threshold (s). According to a second aspect of the invention, in step (b) of processing the signals E (t) and S (t), an intercorrelation of the two measurement signals (with or without prior offset correction) is calculated, and the amplitude variations of the central intercorrelation peak (P) are detected, or the integral of the cross-correlation is calculated, after possibly subtracting the autocorrelation of the output or input measurement signal. More particularly, in the variant with detection of the amplitude variations of the central cross-correlation peak (P), the variations of the amplitude of the central intercorrelation peak (P) are compared with at least one predefined threshold, and detection is made the appearance of a leak when said variation of the amplitude is negative and becomes, in absolute value, greater than this threshold.

More particularly, in the variant with calculation of the integral of the intercorrelation, to detect the appearance of a leak, the result of this integral is compared with a threshold (s) predefined.

The invention also has for its second object the use of the method referred to above to detect the appearance of a leak in a cooling circuit or in a heating circuit, and more particularly, to detect the appearance of a leak in a cooling circuit of an electric arc furnace.

The third object of the invention is also the use of the method referred to above, for detecting the appearance of a leak in a pipe or a network of pipes for the transport and / or distribution of at least one fluid.

The invention also has for its fourth object an installation comprising at least one circuit, means for circulating a fluid in this circuit, an input sensor and an output sensor respectively positioned at the input and at the output. of the circuit, and delivering respectively two measurement signals E (t) and S (t) characteristic of the flow or fluid flow variations, and electronic means for processing the measurement signals E (t) and S (t). Said electronic means (32) are designed to implement the measurement signal processing step (b) E (t) and (St) which is defined in the method referred to above.

In a particular variant embodiment, the sensors are preferably non-invasive type flow meters.

In a particular application of the invention, the circuit is a cooling circuit or a heating circuit. More particularly, the installation is for example constituted by an electric arc furnace.

In another application, the circuit is a pipeline or pipe network for the transport and / or distribution of at least one fluid. The fifth subject of the invention is a program recorded on a medium or in a memory and which, when executed by a programmable processing unit, performs the processing of two measurement signals E (t) and S (t). characteristics of the flow or fluid flow variations respectively at the inlet and the outlet of a circuit, said processing of the measurement signals E (t) and S (t) being carried out in accordance with step ( b) previously defined treatment.

Brief description of the drawings

Other characteristics and advantages of the invention will emerge more clearly on reading the following detailed description of several preferred embodiments of the invention, which description is given by way of nonlimiting and non-exhaustive example of the invention. invention, and with reference to the accompanying drawings in which:

FIG. 1 schematically represents an installation of the invention, FIG. 2 schematically represents an installation of the invention, which has been modified for experimental purposes to provoke a fluid leak,

FIG. 3 schematically represents an installation of the invention, which has been modified for experimental purposes to cause thermal disturbances; FIG. 4 schematically represents an installation of the invention, which has been modified to experimental purposes to cause hydraulic disturbances,

FIG. 5 is a functional diagram illustrating the main successive steps of a preferred variant embodiment of step (b) of processing the measurement signals according to the invention,

FIGS. 6 to 26 are curves resulting from tests carried out by means of the installations of FIGS. 2 to 4, and are detailed below,

FIG. 27 is a block diagram illustrating the main successive steps of another variant embodiment of step (b) of processing the measurement signals according to the invention.

detailed description

FIG. 1 diagrammatically shows a circuit 1, which is fed with a fluid (f) (symbolized by arrows), from a set 2 comprising a buffer tank, for the storage of this fluid, associated a pump for forced circulation of this fluid. Circuit 1 comprises:

a main inlet pipe 10, the inlet of which is sealingly connected to the outlet of the assembly 2 (tank / pump),

a complex network 11 of pipes 110 in parallel, whose inlet common to all the pipes 110, is sealingly connected to the outlet of the main inlet pipe 10, said pipes 110 possibly having different geometric characteristics (lengths , diameters, etc.),

a main evacuation pipe 12, whose inlet is sealingly connected to the outlet of the network 11 common to all of the pipes 110, and the outlet of which is connected in a sealed manner at the inlet of the set 2 for forced circulation of the fluid (f).

When the pump of the assembly 2 operates, the fluid (f) circulates in a closed circuit, being introduced into the main inlet pipe 10, circulates in the pipes 110 of the network 11, and is then conveyed at the outlet of the circuit 1 towards the assembly 2, via the evacuation pipe 12.

In the particular example illustrated, the circuit 1 comprises a complex network 11 of tubings 110 defining for the fluid (f) several possible paths between the input and the output of the circuit 1. Although the invention is particularly interesting for automatically detecting a leak in this type of complex circuit 1, it is however not limited to this type of circuit, and can also be applied to a simpler single path circuit structure and constituted for example by a single tube.

In order to automatically detect a fluid leak (f) in the circuit 1, said circuit is equipped with a control device 3 which is in accordance with the invention.

This control device 3 comprises:

a sensor 30 which is mounted at the inlet of the circuit 1, on the main intake duct 10, and which in operation delivers an electrical measurement signal E (t), - a sensor 31 which is mounted at the output of the circuit 1, on the discharge pipe 12, and which in operation delivers an electrical measurement signal S (t),

electronic means 32 for processing the measurement signals E (t) and S (t) delivered by the sensors 30 and 31. The sensors 30 and 31 are identical and may in general be constituted by any sensor, which allows to detect and characterize the flow or fluid flow variations. Depending on the application, a measurement signal E (t) or S (t) characterizes the quantity of fluid per unit of time passing through the sensor and / or characterizes the material waves generated by the fluid flow variations.

The measurement signals E (t) and S (t) delivered by the sensors 30 and 31 can be of analog type, or be of digital type when the sensors integrate an analog / digital converter.

The sensors 30 and 31 may be of the invasive type (for example measurement probe introduced inside the corresponding conduit 10 or 12) or of the non-invasive type (such as in the example of sensors illustrated in FIG. 1). Non-invasive sensors are however preferred for the implementation of the invention, because the invasive sensors are detrimental to the source of a pressure drop on the fluid.

Preferably, and in a nonlimiting and non-exhaustive manner of the invention, the sensors 30 and 31 are flow meters selected from the following list:

- electromagnetic flowmeters

- ultrasonic flow meters

- mass flow meters (using the Coriolis principle) - vortex flowmeters

The first two types of flowmeters above are of the non-invasive type and are therefore preferred over the last two types of flowmeters which are of the invasive type.

More particularly, for the implementation of the invention, among these flowmeters, the electromagnetic flowmeters are preferably used, because they have the best performances to date (precision, measurement range, robustness, etc.).

The electronic processing means 32 may be implemented in different forms, knowing that what is important for the invention lies in the method of processing the measurement signals E (t) and S (t) which is detailed below. For example and without limitation to the invention, the electronic processing means 32 may be implemented:

in the form of a programmable processing unit such as a microcomputer executing a signal processing program

E (t) and S (t) according to the invention and loaded in random access memory, or in the form of a specific electronic card whose electronic architecture comprises a microprocessor or microcontroller capable of executing an on-board signal processing program E (t) and S (t) according to the invention, or - in the form an electronic card comprising a specific electronic circuit ASIC type, specially designed to perform a signal processing program E (t) and S (t) according to the invention.

FIG. 5 shows the main steps of a preferred variant embodiment of the invention for processing the measurement signals E (t) and S (t) delivered by the sensors 30 and 31.

In general, the signal processing is based on an intercorrelation of the measurement signals E (t) and S (t) [block 43] and on a detection [block 44] and a follow-up [block 45] of the central peak ( P) resulting from the cross-correlation of the measurement signals.

The intercorrelation function is a mathematical function that is known per se. In the continuous time domain, this intercorrelation function lcc (t), when applied to the measurement signals E (t) and S (t), is defined by the relation:

(1) I _cc (t) = E (t) S S (t) = + τ) dτ

The cross-correlation ultimately consists of calculating the overlap integral between the signals E (t) and S (t) by temporally shifting one signal relative to the other.

In practice, the intercorrelation which is calculated in the algorithm of FIG. 5 [block 43] is a discretized intercorrelation function.

The discrete intercorrelation function I _C q implemented is preferably defined by the following relation:

(2) Ic _Cj = ΣE _k S _{J + k} k = û where: E _k are the values resulting from the sampling of the measurement signal E (t) [vector of size n];

S _k are the values resulting from the sampling of the measurement signal S (t) [vector of size n]; n is the total number of samples E _k and S _k in a predefined calculation window; k is an integer from 0 to (n-1); j is an integer taking the values between - (n-1) and (n-1) and thus takes the following successive values: - (n-1); - (n-2); ...; -2; -1; 0; 1; 2; ...; (n-1);

Iccj is a vector of size 2n-1.

The algorithm of Figure 5 will now be detailed. Blocks (40): Analog / Digital Conversion

To allow the calculation of a discretized intercorrelation function, the continuous measurement signals E (t) and S (t) are sampled with a predefined sampling frequency (fe).

These sampling operations are usually carried out using analog / digital converters. These converters can be integrated in the sensors 30, 31 or can be integrated in the electronic processing means 32.

At the output of this sampling step, a succession of discrete samples E _k and S _k is obtained at the sampling frequency (f e).

Blocks (41) and (42): Offset Correction When the system speed is set, the input flow rate E (t) and output S (t) signals oscillate around a mean value.

This average value (Emoy and Smoy) is calculated in parallel for each signal [block 41].

The offset correction [block 42] consists in subtracting from each sample E _k the average value Emoy of the corresponding signal, and each sample Sk the average value Smoy of the corresponding signal. We obtains the corrected signals: (E _k - Emoy) and (S _k - Smoy)

If the system does not drift and / or is not strongly disturbed, these Emoy and Smoy average values are calculated once and for all and stored in memory. On the other hand, if the system derives and / or is strongly disturbed, these average values Emoy and Smoy are recalculated as and when in a sliding time window of predefined duration T _moy . The value of this duration T _m0 y must be greater than the duration of occurrence of a leak in the circuit, and must be sufficiently small to take into account drift and / or disturbances of the system. This duration T _moy is fixed case by case by the skilled person. Preferably, this duration is chosen so that in practice the corrected signals (E _k - Emoy) and (S _k - Smoy) oscillate substantially around zero.

This offset correction step is important because it makes it possible, in combination with the cross-correlation of the corrected signals (block 43), to improve the sensitivity of the leak detection.

In the remainder of the description, it will be considered that the samples

E _k and S _k are those directly derived from the sampling (variant embodiment without offset correction) or are the corrected samples obtained after subtraction of the average value of the signal (embodiment variant with offset correction).

In another variant embodiment, this offset correction could be performed on the analog measurement signals E (t) and S (t), (before sampling) by means of an analog subtractor. Blocks 43 and 44: Intercorrelation / Central Peak To calculate the intercorrelation between the two discretized signals E _k and S _k , a predefined number n of successive samples is stored in memory at each calculation, and the intercorrelation function is calculated. I _C CJ discretized by means of the formula (2) previously described, that is to say (2n-1) successive values l _C c- (ni) to lcc (ni). For example, and without limitation to the invention, the computation of the intercorrelation is carried out with 1000 successive samples E _k and S _k (k varying from 1 to n and n being 1000), with a sampling frequency fe of 1 kHz. For each series of n samples, an intercorrelation vector Icq is thus calculated every second. In this example, one calculates every second 1999 values lcc- (ni) to lcc (ni): lcc- (ni) = Eo ^x S _n -I l0C- (n-2) = Eo ^x S _n- 2 + El * S _n- I

So + E _n- I * Si For each series of (2n-1) sample values Icq, the central peak p of the intercorrelation (peak of greater amplitude) is determined automatically and in a manner known per se.

The aforementioned steps of intercorrelation and detection of the central peak, which respectively correspond to the blocks 43 and 44 of FIG. 5, are performed repetitively in successive sliding windows of n samples Ek and S _k . These successive windows (or series) of n samples E _k or S _k can be non-overlapping in time (we take n samples, then the following n samples without overlapping between the series of samples), or can on the contrary overlap in part.

Also, for each cross correlation calculation Icq, the two calculation windows of n samples E _k and S _k that are used can be defined on the same time interval, without time offset between the windows (in this case the samples E _k and S _k successive of each window were all sampled at the same times), or on the contrary can be defined with a more or less significant temporal shift between the calculation windows.

The sampling frequency (fe) and the number n of samples characterize a parameter T, corresponding to the integration period for the discretized intercorrelation. This parameter T is defined by the following relation: T = - fe In the above example, with a sampling frequency of 1KHz, and 1000 samples in each successive window, the integration period T is 1 s.

For each given application, it is up to the skilled person to judiciously set the values of the sampling frequency (fe) and the number n of samples, so that the integration period T is compatible with the times. system features. Block 45: Tracking and alarm

The variations of the central peak P are controlled by comparing the amplitude of the central peak (P) of the intercorrelation with at least one threshold (s).

This threshold (s) may, depending on the case, be a predefined threshold of constant value.

It can also be constituted by a self-adaptive threshold which is a function of the values E _k and S _k , and more precisely successive values calculated for the central peak of the intercorrelation; for example, the self-adaptive threshold corresponds to the average value over a predefined number N of previously calculated central peak values.

A leak is automatically detected when the variation in the amplitude of the central peak of the intercorrelation is negative and is in absolute value greater than a predefined threshold (s), and an alarm (audible, visual, sending) is triggered if necessary. automatic alarm message by any known telecommunication means, etc.)

It is also possible to calculate several auto-adaptive detection thresholds, which are differentiated by the value of the variable N (number of central peak values used to calculate the threshold (s).) For example, a threshold (if ) in the short term is calculated on the last 10 central peaks (N = 10), a middle-term threshold (s2) is calculated on the last 100 central peaks (N = 100), a long-term threshold (s3) is calculated on the last 500 central peaks (N = 500) When the variation of amplitude of the central peak is negative and exceeds in absolute value one of the thresholds (si) to (s3), one triggers a characteristic alarm of this threshold.

Experimental results

In all the examples of implementation of the invention described in the experimental tests below, the fluid (f) flowing in the circuit 1 is a liquid (in this case water); the sensors 30 and 31 used are electromagnetic flowmeters having an accuracy of the order of 0.5%. The duration T _moy for the offset correction was 30 seconds. The sampling frequency (fe) of the measurement signals E (t) and S (t) is equal to 1 KHz. The number of samples (n) for intercorrelation is equal to 1000. The discretized intercorrelation function Icq is calculated on (n) samples Ek and Sk acquired simultaneously and not temporally shifted, and in sliding windows of n samples which are juxtaposed and do not overlap. Leak Detection FIG. 2 shows a fluid circuit 1 modified for purely experimental purposes to cause fluid leakage.

This circuit 1 of FIG. 2 differs from that of FIG. 1 only in that on one of the tubes 110 of this circuit, a bypass 13 has been made, on which a leakage valve 14 is mounted. a receptacle 15. When the valve 14 is open, a very small portion of the fluid (f) is taken from the circuit 1 and is not redirected at the output of the circuit 1 to the assembly 2, but flows into the receptacle 15 , which makes it possible to cause a leak in the circuit 1 in the context of experiments of the invention. Of course, these means 13, 14, 15 for causing a leak in the circuit 1 are provided solely for experimental purposes to test the invention, and are not found in the context of a final operational and non-experimental installation comprising the circuit 1 and its means 2 for supplying fluid.

In a first step, the installation of FIG. 2 was operated without leakage, by supplying the circuit 1 with water with an average flow rate of the order of 295 m ³ / h, the leakage valve 14 being closed. FIGS. 6 and 7 show, over a short period of operation time without leakage, the raw data ER and Sk coming from the sampling of the measurement signals E (t) and S (t) (signals at the output of FIG. block 42 after offset correction). FIG. 8 shows the cross-correlation of the sampled measurement signals ER and S _k of FIGS. 6 and 7.

In a second step, a leak with a flow rate of the order of 1.2 m ³ / h is caused by opening the valve 14.

FIGS. 9 and 10 show, for a short period of operating time taken at the beginning of leakage, the raw data E _k and S _k coming from the sampling of the measurement signals E (t) and S (t). (output signals of block 42 after offset correction). FIG. 11 shows the inter-correlation of the sampled measurement signals E _k and S _k of FIGS. 9 and 10. The comparison on the one hand of FIGS. 6 and 7 (measurement signals E _k and SR without leakage) with FIG. on the other hand, FIGS. 9 and 10 (measurement signals E _k and S _k with leakage) show that it is very difficult to discriminate in these signals a change making it possible to characterize the appearance of a leak. On the other hand, the comparison of FIGS. 8 and 11 resulting from inter-correlation respectively without leakage and with leakage, shows a notable modification of the result of the intercorrelation function, and in particular a negative variation of the minimum, which makes it possible to characterize and detect the appearance from the leak. Sensitivity of leak detection

To characterize the sensitivity of the leak detection, it has caused, by appropriately opening the leakage valve 14 of the installation of Figure 2, five successive leaks having a rising flow.

These successive leaks are characterized in Table I below. Table I

FIGS. 12 and 13 show the measurement signals E (t) and S (t) delivered by the sensors 30 and 31 for the period of time between 470s and 570s. FIG. 14 shows the time course of the amplitude of the central peaks of the intercorrelation (output of block 44) for the period of time between 0s and 730s. In this figure 14, the parts of the curve for discriminating the five successive leaks are indicated by the arrows F1, F2, F3, F4 and F5.

The analysis of FIG. 14 makes it possible to show a sudden negative variation in the amplitude of the central peak of the intercorrelation at each occurrence of a leak. It also makes it possible to show that this amplitude variation, which makes it possible to characterize and detect the appearance of a leak, increases in absolute value with the flow rate of the leak.

FIG. 15 shows the amplitude variation (in absolute value) of the central cross-correlation peak characteristic of a leak as a function of the relative value of the leakage caused (in m ³ / h). This figure shows that the variation of the amplitude of the central cross-correlation peak is proportional to the relative value of the leak. The intercorrelation can therefore not only be used to detect the appearance of a leak in the circuit 1, but also advantageously makes it possible to characterize the relative value of the leak that has been detected.

Finally, this experiment validated that it was possible to detect leakage of very small amplitude with respect to the main flow rate of the fluid at the inlet of the circuit 1. In the case of the first induced leak which has the smallest amplitude, it was possible to detect this leak with a sensitivity of the the order of one thousandth (1.9.10 ^"3 ), whereas comparatively the measurement accuracy of the flow meters used was of the order of 0.5%.

FIG. 3 shows a fluid circuit 1 modified for experimental purposes only to cause thermal disturbances in circuit 1. This circuit 1 of FIG. 3 differs from that of FIG. 2 only in that added on one of the tubes 110 of the circuit 1:

a heating element 16 (for example a heating resistor) which makes it possible to locally heat the tubing 110 and thereby the portion of fluid circulating locally in this tubing 110;

two temperature sensors 17a and 17b making it possible to measure the temperature variations of the fluid flowing in said tubing 110 equipped with the heating element 16.

In a first phase, the installation of FIG. 3 is operated without causing a leak (valve 14 closed).

FIGS. 16 and 17 respectively show the measurement signals E (t) and S (t) during this test with fluid heating and without leakage, and in FIG. 18 the temperature of the fluid as a function of time.

(substantially constant) measured upstream of the heating element 16 by the sensor 17a (curve A) and the temperature of the fluid as a function of time measured downstream of the heating element 16 by the sensor 17b (curve B). FIG. 19 shows the evolution over time of the amplitude of the central cross-correlation peaks. This curve of FIG. 19 shows that, at the start of heating, the variation in the amplitude of the central intercorrelation peaks increases sharply, but in the opposite direction (positive variation) to the variation of the amplitude of the central peaks of FIG. intercorrelation in case of leakage.

In a second phase, the heating of the tubing is stopped 110 for a time sufficient to return to substantially identical fluid temperature conditions throughout circuit 1.

In a third phase, the tubing 110 is heated again by means of the heating element 16, then a leak is generated by opening the valve 14.

The characteristics of the leak are given in Table II below: Table II

FIG. 20 shows the temperature of the fluid as a function of time (substantially constant) measured upstream of the heating element 16 by the sensor 17a (curve A) and the temperature of the fluid as a function of the time measured downstream of the heating element 16 by the sensor 17b (curve B) at the end of the second phase and during the third heating phase.

FIGS. 21 and 22 respectively show the measurement signals E (t) and S (t) during this leakage heating test, and in FIG. 23 the evolution over time of the amplitude of the central peaks of FIG. intercorrelation. Referring to the curve 23, there is a sudden negative variation in the amplitude of the central peaks of the intercorrelation very shortly after the outbreak of the leak. This sudden variation in the amplitude of the central intercorrelation peaks thus makes it possible to always detect the appearance of the leak, despite the local thermal disturbance in the circuit 1, which was caused by the heating of the tubing 110 by means of the heating element 16. Hydraulic disturbances

FIG. 4 shows a fluid circuit 1 modified for purely experimental purposes only to cause hydraulic disturbances in the circuit 1. This circuit 1 of FIG. 4 differs from that of FIG. 2 only in that has added on one of the tubings 110 of the circuit 1 a valve 18 for closing or opening said tubing 110.

To generate a hydraulic disturbance, the valve 18 is closed. The measurement signals E (t) and S (t) are shown in FIGS. 24 and 25 respectively during this test with closure of the valve 18, and in FIG. the evolution over time of the amplitude of the central peaks of the intercorrelation.

This curve of FIG. 26 shows that very shortly after the closing of the valve 18 (hydraulic disturbance in the circuit 1), the variation in the amplitude of the central intercorrelation peaks increases sharply, but in the opposite direction (positive variation ) the variation of the amplitude of the central intercorrelation peaks in case of leakage. This behavior is comparable to what has been previously described for thermal disturbances, with particular reference to FIG. 19. Knowing that the effect of the hydraulic disturbance on the results of intercorrelation (positive variation of the amplitude of the central peaks cross-correlation) is inversed with respect to the effect of a leak (negative variation in the amplitude of the central intercorrelation peaks), it can be concluded that the follow-up of the evolution over time of the amplitude of the Central peaks resulting from the cross-correlation of the measurement signals E (t) and S (t) still make it possible to detect the occurrence of a leak in the circuit, even in the presence of hydraulic disturbances. Other variants of achievements

The invention is not limited to the preferred embodiment which has just been described with reference to FIGS. 1 to 26. In particular and in a non-exhaustive manner, the detection of the central peak of correlation (peak of larger amplitude) (block 44 of FIG. 5) can be replaced by a calculation of the sum of all the samples I _CC j resulting from the intercorrelation (which amounts to calculating the integral of the intercorrelation). In the absence of leakage (see Figure 8), this sum is close to zero. In the presence of a leak (see Figure 11), the result of this sum becomes strongly negative. For leak detection (block 45), a negative threshold (s) is chosen in this case. When the negative result of this sum of the samples _CCj resulting from the intercorrelation becomes lower than this threshold (s) predefined negative, in this case is automatically detected the appearance of a leak. In another variant embodiment, illustrated in FIG. 27, the detection (block 45) is carried out either directly on the result of the intercorrelation, but on the result of the reduced intercorrelation (block 47) of the result of the autocorrelation (block 46) of the corrected output signal (S _k -S _avg ). In another variant, it is also possible to perform the detection (block 45) on the result of the reduced intercorrelation (block 47) of the result of the autocorrelation (block 46) of the corrected input signal (E _k -E _m0 y ). This advantageously makes it possible to improve the signal-to-noise ratio and to improve the leak detection sensitivity. Applications of the invention The invention finds its application in the detection of the appearance of fluid leak (s) in any circuit within which a fluid circulates, said fluid may be a liquid or a gas, or a mixture gas / liquid.

Preferably, the invention is of interest in all applications: where the flow rate of the leaks to be detected can be very small compared to the flow rate of the fluid at the inlet of the circuit, and / or

the fluid undergoes perturbations modifying its volume (in particular thermal, hydraulic disturbances, etc.). A particularly interesting example of application of the invention lies in the detection of leaks in a cooling circuit or in a heating circuit, inside which circulates a coolant.

More particularly, the invention may advantageously be used in the field of iron and steel to automatically detect reliably and quickly the appearance of leaks in the cooling circuits of an electric arc furnace which form a complex network of tubing. In this case, the circuit 1 of Figure 1 is for example the cooling circuit fitted to the vault of an electric arc furnace, or the cooling circuit panels of an electric arc furnace.

A particularly interesting example of application of the invention lies in the detection of leaks in a pipe or in a network of pipes for the transport and / or distribution of fluids (pipelines, pipelines, etc.).

Claims

1. A method of controlling a circuit (1) in which a fluid (f) circulates, said method comprising the following steps: (a) the circulating fluid is detected by means of an input sensor (30) and an output sensor (31) respectively positioned at the input and at the output of the circuit, and respectively delivering two measurement signals E (t) and S (t) characteristic of the flow or flow variations of the fluid,

(b) the two measurement signals E (t), S (t) are processed for the purpose of detecting fluid leakage in the circuit, if appropriate, characterized in that in step (b) of processing of signals E (t) and S (t), an offset correction is performed by removing the measurement signals [E (t) or ER; S (t) or S _k ] their mean value (Emoy; Smoy), and an intercorrelation of the two corrected measurement signals is calculated.

2. Method according to claim 1, characterized in that in step (b) of processing the signals E (t) and S (t), the measurement signals E (t) and S (t) are sampled. ), and the intercorrelation function

Icc is a discrete function calculated using the following formula:

Ic _Cj = ΣE _k S _{J + k}

where: E _k are the values resulting from the sampling of the measurement signal E (t);

S _k are the values resulting from the sampling of the measurement signal S (t); n is the total number of samples Ek and Sk in a predefined calculation window; k is an integer from 0 to (n-1); j is an integer taking the values between ~ (n-1) and (n-

1);

Iccj is a vector of size 2n-1.

3. Method according to one of claims 1 or 2, characterized in that in step (b) of processing the measurement signals, one calculates an intercorrelation of the two measurement signals, less the autocorrelation of the signal of output or input measurement.

4. Method according to one of claims 1 to 3, characterized in that in step (b) of processing the measurement signals, it detects the amplitude variations of the central peak (P) intercorrelation.

5. Method according to claim 4, characterized in that in step (b) of processing of the measurement signals, the variations of the amplitude of the central cross-correlation peak (P) are compared with at least one threshold. predefined (s), and the occurrence of a leak is detected when said variation of the amplitude is negative and becomes, in absolute value, greater than this threshold.

6. Method according to claim 5, characterized in that said threshold (si, s2 or s3) is auto-adaptive and is calculated according to the amplitudes of the previous central intercorrelation peaks.

7. Method according to one of claims 1 to 3, characterized in that in step (b) of processing of the measurement signals, the integral of the cross-correlation is calculated, and the result of this integral is compared. with a predefined threshold (s).

8. Use of the method according to one of claims 1 to 7, for detecting the occurrence of a leak in a cooling circuit or in a heating circuit.

9. Use according to claim 8, for detecting the occurrence of a leak in a cooling circuit of an electric arc furnace.

10. Use of the method according to one of claims 1 to 7, for detecting the appearance of a leak in a pipe or a network. of transport pipes and / or distribution of at least one fluid.

11. Installation comprising at least one circuit (1), means (2) for circulating a fluid (f) in this circuit (1), an input sensor (30) and an output sensor (31). ) respectively positioned at the inlet and the outlet of the circuit (1), and respectively delivering two measurement signals E (t) and S (t) characteristic of the flow or fluid flow variations, and means electronic processing means (32) for the measurement signals E (t) and S (t), characterized in that said electronic means (32) are designed to implement the measurement signal processing step (b) E ( t) and (St) which is defined in the process of one of claims 1 to 7.

12. Installation according to claim 11, characterized in that the sensors (30,31) are non-invasive type flow meters.

13. Installation according to claim 11 or 12, characterized in that the circuit (1) is a cooling circuit or a heating circuit.

14. Installation according to claim 13, characterized in that it consists of an electric oven with arches.

15. Installation according to claim 11 or 12, characterized in that the circuit (1) is a pipe or pipe network for the transport and / or distribution of at least one fluid.

16. A program recorded on a medium or in a memory, and which, when executed by a programmable processing unit (32), automatically performs the checking method referred to in one of claims 1 to 7.