FR2999722A1 - LOCATION OF ONE OR MORE DEFECTS IN AN ELECTROCHEMICAL ASSEMBLY. - Google Patents

Abstract

The object of the present invention is to provide a method for probing one or more defects in an assembly comprising an electrochemical device and a peripheral system for operating said electrochemical device, said method comprising: a decomposition step (S1) of applying a wavelet transform on a signal measured on the electrochemical device to decompose said signal into a plurality of sub-signals, a quantization step (S2) of estimating the energy distribution (Ej) of each sub-signal, a determining step (S4) of calculating an entropy (ER) for determining a signature of the signal, and d) a locating step (S5) of analyzing the signature so as to allow the location of a defect in said set.

Description

The present application relates to the field of the detection of one or more defects in so-called electrochemical assemblies, that is to say assemblies comprising at least one electrochemical device and one or more electrochemical assemblies. A peripheral system allowing the operation of said electrochemical device More specifically, the present invention relates to the detection, location and identification of defects (or malfunctions) in such electrochemical assemblies. The present invention thus makes it possible to improve the reliability and the lifetime of the electrochemical devices, alone or in an electrochemical system. The present invention thus has many advantageous industrial applications in particular to avoid irreversible damage to the electrochemical assemblies or a stop of their operation. By peripheral system of an electrochemical device is meant here all the mechanical, electrical and electronic components for the operation of an electrochemical device.

Finally, by default or malfunction in the sense of the present invention, it is necessary to understand all types of defects, such as for example a point failure or a long-term degradation, likely to generate a drop in performance in an electrochemical device and / or in a Peripheral system as defined above. State of the art There are today many devices called "electrochemical". A first category of this type of device concerns devices converting chemical energy into electrical energy, in order to supply this energy to electrical devices or to store it for later provision. As explained above, such devices can be for example batteries, fuel cells or super-capacitors. A second category of this type of device relates to devices using various methods using electricity to perform chemical reactions, or to separate products or reagents between them. Such devices commonly use so-called "electrochemical" processes such as, for example, electrolysis, electroplating, electro-erosion or electro-flotation. It is known that the reliability and the life of these devices are limited by various phenomena.

For fuel cells for example, two main phenomena are identified and lead to the fall in their performance, or even a complete failure: - on the one hand, cycling and / or operation in transient mode (or by accumulation of stops and starts, either by requested power variation) can lead to a reduction in the life of these devices; - on the other hand, certain incidents such as the lack of control of certain parameters of the electrochemical process used (interruption of supply of reagents, mismanagement of the products and by-products of the reaction, etc.), poisoning of the medium , the failure of a component or a module for example, can occur during their operation. These deleterious phenomena require the use of diagnostic methods to enable their detection and possible correction. The methods for detecting classic defects are most often based on the knowledge of a certain number of parameters, external or internal. These methods generally require specific instrumentation, such as the insertion of internal sensors into the electrochemical device itself.

Such an instrumentation is not always desirable because it is often expensive and is not always easy to set up, especially given the geometry of electrochemical devices rarely adapted to the installation of sensors. In addition, the sensors induce, because of their insertion, an alteration of the device that can increase the probability of occurrence of defects and leading to false diagnoses. Finally, in the case of the use of an electrochemical device for embedded mobile applications, the size of the device must be reduced to a minimum, as well as that of the diagnostic instrumentation, which makes it impossible to use the methods classics. WO 2010/149935 proposes a solution to overcome these various disadvantages. The technical teaching of this document makes it possible to detect in real time and non-intrusively a defect in an electrochemical device, this using minimal instrumentation. Nevertheless, the technical teaching of this document WO 2010/149935 does not make it possible to distinguish the faults located at the level of the electrochemical device itself from those situated at the level of the peripheral system. Indeed, in addition to the defects related to the electrochemical device that can cause a decrease in performance or a stop in the operation of said electrochemical device, certain defects may also occur at the peripheral system. These defects that occur at the peripheral system can also vary the operating conditions of the electrochemical device.

Take the example of an electrochemical assembly of the "fuel cell" assembly type. Such an assembly essentially consists of the following elements: - the fuel cell module (s) as such, which corresponds to the electrochemical device in the sense of the invention, and - the peripheral system comprising in particular the different subsystems which are respectively configured to supply the reactive gases and to control the operating temperature of the module. Such a set therefore makes it possible to transform directly (that is to say without combustion) the chemical energy contained in the combustible gases such as hydrogen into electrical energy. A failure of the peripheral system which causes a gas leak at the inlet of a fuel cell causes a reduction of the reactive gas flow Under these operating conditions, the battery undergoes a depletion of gas which causes a degradation in the material catalytic fuel cell. It is therefore clear in this example that a fault located at the peripheral system can directly influence the proper operation of the fuel cell and its performance. It is therefore important to be able to distinguish the faults located at the level of the electrochemical device from those located at the level of the peripheral system. This distinction (or "location") makes it possible in particular to: - on the one hand, to refine the maintenance operations by making it possible to acting directly on the element (s) concerned, and (ii) on the other hand, to refine the corrective actions and to target them in order to prolong the lifetime of the electrochemical device and / or the system. SUMMARY AND OBJECT OF THE PRESENT INVENTION The present invention aims at improving the present situation described above.

One of the objectives of the present invention is to allow this location, this without adding additional sensors neither at the level of the electrochemical device nor at the level of the peripheral system.

One of the concepts underlying the present invention is based on a specific process for extracting and analyzing a signature of a signal measured at the level of the electrochemical device itself. For this purpose, the object of the present invention is to provide a method for probing one or more defects in an assembly comprising an electrochemical device and a peripheral system of the electrochemical device. According to the present invention, the method comprises a decomposition step of applying a wavelet transform on a signal measured on the electrochemical device.

This wavelet transform can be, for example, either a discrete wavelet transform or a continuous wavelet transform. Preferably, the signal measured on the electrochemical device is an electrical signal relative for example to a voltage measured at the terminals of the electrochemical device. Of course, it is understood here that other signals such as a current or pressure signal can also be measured in the context of the present invention. This wavelet transformation makes it possible to decompose the signal into a plurality of sub-signals each represented by wavelet coefficients, each sub-signal corresponding to a determined resolution level of the signal.

Advantageously, the probe method according to the present invention comprises a quantization step which consists in particular in estimating the distribution of the energy of each sub-signal from the wavelet coefficients. This step thus makes it possible to quantify the level of information contained in each sub-signal.

Advantageously, the probe method according to the present invention further comprises a determination step which consists in particular in calculating at least one physical quantity, such as for example a relative entropy of wavelets, for selected sub-signals.

In an advantageous embodiment of the present invention, the physical quantity is calculated for the distribution of the energy of the selected sub-signals. In another embodiment, the physical quantity is calculated for the wavelet coefficients of each selected sub-signal.

By selected sub-signals, sub-signals containing a predetermined significant level of information are to be understood here. These sub-signals may, for example, be selected during a selection step during which the sub-signals having wavelet coefficients containing an information level satisfying a determined selection criterion are selected.

The determination of this selection criterion can be done for example according to a statistical analysis of the energy distribution for the sub-signals. This criterion can be static or dynamic. In other words, this criterion can be determined as a function of the variations of the energy distribution for each sub-signal over time. It can also be determined statically for example according to a level of energy to be achieved. This determination step thus makes it possible to obtain a signature of the signal, this signature constituting a relevant indicator for locating and identifying a fault in the electrochemical device and / or in its peripheral system. This signature preferably corresponds to an estimate of the difference between the energy distribution of the measured signal and that of a reference signal (this reference signal may for example be a measured signal corresponding to a normal operating state of the signal). electrochemical assembly). Then, the method according to the present invention comprises a location step which consists in particular of analyzing the signature of the signal to locate a defect in said set. Thus, thanks to this succession of technical steps, characteristic of the present invention, it is possible to precisely locate the possible defects or defects in an assembly composed of an electrochemical device and a peripheral system.

The approach developed here therefore makes it possible to monitor the evolution of the operating conditions of an electrochemical assembly such as a "fuel cell" assembly in order to be able to evaluate the state of health of said assembly and to detect early the occurrence of defects in the electrochemical device as such and the peripheral system. More specifically, the approach developed here is based on the determination of characteristic values of the state of health of the electrochemical element and its peripheral system. For this purpose, a wavelet transform is used to decompose the 1 () signals from simple measurements on the electrochemical device (for example the measurement of a voltage or current). This transformation has the advantage of conserving the signal energy and redistributing it in a more concentrated form on a limited number of components at different resolution levels. The Applicant observes that at different operating conditions (normal, deviant, abnormal, etc.), the distribution of wavelet energy in the signal from the electrochemical device theoretically has different characteristics. It is on the basis of the analysis of these differences that the fault indicators are identified and determined in order to differentiate the normal and abnormal states of the operation of an electrochemical device and its system. An identification of the defect is also possible using a pre-established knowledge base. This location and this identification is therefore allowed in particular by the specific analysis of the signature of the signal. Preferably, this analysis provides for a comparison of the signal signature variations with one or more predetermined discrimination thresholds. We speak of a location by discriminative thresholding.

Thanks to this analysis, the method according to the present invention makes it possible to define a diagnostic tool capable of discriminating defects related to the electrochemical device itself from those related to the peripheral system. In this process, the electrochemical device is at the same time used as a defect sensor for itself and for the peripheral system. Due to the limitation of the number of sensors used, the diagnosis can be embedded, and is executed in real time. In addition, the diagnosis made uses signals measured at a relatively low sampling frequency. The acquisition of such signals does not require the use of high-end equipment; the use of standard equipment is sufficient here and reduces the cost of diagnosis. In an advantageous variant embodiment, when the comparison made between the variations of the signature of the signal and the predetermined discrimination threshold did not make it possible to detect a defect (for example if the relative entropies of wavelets are not discriminative), the method comprises the reiteration of the selection step by widening the determined selection criterion, for example to select other sub-signals which comprise less significant information. Correlatively, the object of the present invention relates to a computer program which includes instructions adapted for the execution of the steps of the probe method as described above, this in particular when said computer program is executed by a user. computer Such a computer program can use any programming language, and be in the form of a source code, an object code, or an intermediate code between a source code and an object code, such as only in a partially compiled form, or in any other desirable form. Likewise, the subject of the present invention relates to a computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the probe method as described above.

On the one hand, the recording medium can be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM, for example a CD-ROM or a microelectronic circuit-type ROM, or a magnetic recording means, for example a diskette of the type " floppy disc "or a hard drive. On the other hand, this recording medium can also be a transmissible medium such as an electrical or optical signal, such a signal can be conveyed via an electric or optical cable, by conventional radio or radio or by self-directed laser beam or by other ways. The computer program according to the invention can in particular be downloaded to an Internet type network. Alternatively, the recording medium may be an integrated circuit in which the computer program is incorporated, the integrated circuit being adapted to execute or to be used in the execution of the method in question. The subject of the present invention also relates to a fault probe system which is configured to carry out the steps of the method described above. The probe system according to the present invention comprises in particular the following computer means: an acquisition module configured to acquire a measurement of a signal on the electrochemical device; a computer calculator configured to: apply a wavelet transform on the signal for splitting the signal into a plurality of sub-signals each represented by wavelet coefficients, each sub-signal corresponding to a resolution level of said signal; - estimating the distribution of the energy of each sub-signal from the wavelet coefficients for quantizing the level of information contained in each sub-signal, and - calculating at least one physical quantity for selected sub-signals to determine a signature of the signal, the selected sub-signals being sub-signals containing a predetermined significant level of information, and - a processing circuit configured to analyze the signature signal to allow the location of a defect in said set. Thus, the object of the present invention, by its different functional and structural aspects, allows a location and a precise identification of defects in an electrochemical assembly such as a fuel cell.

Brief description of the appended figures Other features and advantages of the present invention will emerge from the description below, with reference to Figures la to 8 attached which illustrate an embodiment having no limiting character and in which: - the figure 1a is a flowchart illustrating the probe method according to an advantageous embodiment; FIG. 1b schematically shows a probe system according to the present invention with an electrochemical assembly; FIG. 2 schematically represents the decomposition of a signal received from an electrochemical device into a plurality of wavelet coefficients according to a determined resolution level; FIG. 3a represents a histogram relating to the Energy Percentages of the Wavelength Coefficients of four details associated with the degradation of a solid oxide fuel cell according to an advantageous exemplary embodiment; FIG. 3b represents a graph showing the evolution of the relative entropy of the voltage signal measured at the terminals of a solid oxide fuel cell according to FIG. 3a; FIG. 4 represents a histogram relating to the Energy Percentages of Wavelength Coefficients of six details associated with a solid oxide fuel cell in good health and in normal operating condition; FIG. 5 represents in example 1 the energy distribution of the signals of the six cells of a fuel cell according to sixteen different operating conditions; FIG. 6 represents, in example 1, the evolution of the relative entropy value of the voltage signals of each cell of a fuel cell according to FIG. 5 as a function of the operating conditions; FIG. 7 represents in example 2 the relative entropy value of the voltage signal of each cell of a fuel cell for a normal condition; and FIG. 8 represents the evolution of the relative entropy values of the voltage signals of a degraded cell and of a healthy cell under different operating conditions. DETAILED DESCRIPTION OF AN ADVANCED EMBODIMENT A probe method and the computer system 100 associated therewith will now be described in the following with reference to Figures 1a-8. Allow the location of one or more defects in an electrochemical assembly ENS such as a "fuel cell" assembly is one of the objectives according to the present invention. The example described here relates to a "fuel cell" assembly. It is understood here that this example is purely illustrative, and would in no case be limiting in nature. Previously, the probe method according to the present invention selects one or more relevant signals which must be informative and representative for the behaviors (normal and abnormal) of the set ENS studied, namely here a set "fuel cell" composed of a fuel cell module PAC consisting of six cells C1 to C6 and PER peripheral system that allows its proper operation (see Figure lb). These signals make it possible to establish a reference signal, this signal representing a raw signal relative to the nominal operation of the set ENS. This reference signal is used for localization, as is understood in the rest of the text below. In the example described here, and as illustrated in FIG. 1a, the probe method according to the present invention provides a measurement SO of one or more signals f coming from the fuel cell module PAC. In the example described here, the electrochemical device PAC being used as a sensor of the state of health of the peripheral system PER as well as of itself PAC, a signal f relating to a voltage or a current is a suitable signal capable of to reflect the state of health of the ENS. As mentioned above, these measurements can be performed by a conventional acquisition module 10 (shown in FIG. 1b).

Then, as illustrated in FIG. 2, the computer computer 20 of the system 100 decomposes, during a decomposition S 1, the signals f measured according to several levels of resolution j, j being included here between 0 and 5. The applicant submits that this variable j is a function, in particular, of the nature of the signal f measured, the nature of the electrochemical device PAC, the recording duration of the signal, or the type of decomposition performed. In the exemplary embodiment described here, this decomposition uses a discrete 1P wavelet transform. The choice of the wavelet 1P is also a function, in particular, of the nature of the signal f measured and the set of ENSs studied (the wavelet 1P should preferably have as many similarities as possible with the signal f). This discrete 1P wavelet transform therefore makes it possible to decompose the signal f into a plurality of sub-signals that can be separated into two categories: the approximations and the details.

The last category of the sub-signals (ie the signals relating to the details) covers the most dynamic aspects of the original signal f and therefore contains so-called "significant" information which is representative of the state of the signal. health of the electrochemical device PAC and its peripheral system PER.

The study of these details is characteristic of the invention. The discrimination details / approximations of the sub-signals to be studied is carried out according to a predetermined selection criterion, which makes it possible to separate the sub-signals relating to sub-signal approximations relating to details (which contain significant information ).

As with the approximation, each detail is represented by a vector of wavelet coefficients. In the example illustrated here in FIG. 2, the wavelet coefficients relative to the approximations, denoted cA, are thus distinguished in FIG. 2 (where j is between 1 and 5) of the wavelet coefficients relating to the details, denoted cD. in Figure 2 (where j is between 1 and 5).

As mentioned above, the wavelet coefficients that will be selected and processed are those relating to the details in that they contain the significant information that one seeks to detect and locate. The approach proposed in document WO 2010/149935 only provides an analysis of the wavelet coefficients of the frequency band that is significant for the defect. Thus, in the case where the sampling frequency of the signal is low (for example 1 Hertz), the approach proposed in WO 2010/149935 is no longer applicable because the significant information required to indicate the defect is unavailable in the measured signal. In contrast, the approach proposed here provides an analysis of wavelet coefficients on all frequency bands. Thus, according to the present invention and as illustrated in FIG. 2, for each level of resolution j, the wavelet coefficients cD are included in a vector in the following form: [c ,,,; c ,, k]. For the sake of clarity, the wavelet coefficients relating to the details are noted in the remainder of the description as follows c 1 (k), where c (k) = (f ItP) It is understood here that k is a positive integer between 1 and N, N being a function, in particular, of the signal f and of its recording duration.

Then, in the example described here, the calculator 20 estimates, during a quantization step S2, the distribution of the energy Ej of each sub-signal from the wavelet coefficients c (k). This calculation is performed according to the following formula: = c1 (k) 2 The calculator 20 thus performs a calculation on the sum of the wavelet coefficients squared (energy) to quantify the level of information contained in each detail. In this calculation, the wavelet coefficients c (k) are normalized to make the distribution of the energy Ej more clear over different frequency bands. Each vector of coefficients c (k) is transformed into a scalar quantity describing the fraction of the energy Ej of the signal f contained on the corresponding frequency band. The set of energy fractions represents the distribution of the signal energy over all the frequency bands. This distribution of the energy Ej has proved to have a close relationship with the operating state of the electrochemical assembly ENS. Indeed, FIG. 3a shows the evolution of the degradation of a solid oxide fuel cell (also known as the "SOFC" anagram) which could be reflected by the evolution of the PECO (Percentage of Energy of the Wavelet Coefficients) present in the voltage signal. In this example, we represent the evolution over time of the energy distribution (PECO) for four sub-signals corresponding to four details: dl, d2, d3 and d4.

In this figure 3a, there is notably an increase of about 20% of the PECO for the fourth detail d4, which resulted in a reduction of 20% for the first detail dl. This figure 3a thus shows the close relationship existing between the defects (abnormal conditions) and the distribution of the energy. Then, the computer 20 makes it possible to select, during a selection S3, the sub-signals (that is, the details) that are significant. This selection provides an analysis of the energy distribution using statistical approaches to determine levels of detail that are significant. By way of illustration, FIG. 4 shows a group of histograms of the CEECs present in the voltage signals of a SOFC fuel cell at the different periods of operation. In this example, we represent the evolution over time of the energy distribution (PECO) for six sub-signals corresponding to six details: dl, d2, d3, d4, d5 and d6. It can be observed in this FIG. 4 that the sixth detail d6 in the first two periods occupies a relatively large portion of the energy relative to that in the other periods, which leads to a reduction of PECO of the first detail d1.

This is due to the disruption of the detail d6 which in fact does not associate with the state of health of the fuel cell. Therefore, the detail at the sixth level should not be taken into account in the CEEC calculation. This selection S3 thus makes it possible to filter the details which are not significant. The selection S3 on the sub-signals being carried out, the computer 20 then proceeds to a determination S4 during which a physical quantity is calculated for these sub-signals. This physical quantity is in the example described here an ER relative entropy of wavelets which is calculated according to the following formula: ER = -Epk * ln (/ A) qk where pk is a percentage of energy of the coefficients of wavelets equal at: Ek Pk = LE, and q is a vector of the energy percentages of the reference signal mentioned above. This physical quantity, here the relative entropy of the wavelet coefficients, thus makes it possible to determine a signature of the signal which in the example described here is a measure of the difference between the energy distributions of two signals (the signal f and the reference signal).

The relative entropy calculated here makes it possible to obtain measurements between 0 and 1, which is more easily comparable than measurements made directly between the signal studied and the reference. Taking again the example described above for a SOFC which is degraded (FIG. 3a), it can be seen in FIG. 3b that the relative entropy of the voltage signal increases in time with the degradation. Then, the processing circuit 30 of the system 100 allows a location S5 of the defects in the set ENS. During this location S5, the signature of the signal is analyzed so as to allow the location of a defect in said set ENS. In this step, discrimination thresholds are determined from which a defect can be considered as present. If the ER entropy values are not discriminative enough to differentiate the normal situations from the abnormal situations, it is necessary to return to the selection step S3 and redo the selection of the significant details by resuming a less selective selection criterion. PAC electrochemical device has occurred, the energy distribution of the signal f remains more disorderly than that with a defect of the peripheral system PER. It is this difference that allows us to distinguish these two kinds of defect. In fact, the energy distribution of a signal f characterizes the frequency spectrum shape of the signal. It is this analysis, called discriminative thresholding, which is carried out to determine this location of the defect in the assembly. The document WO 2010/149935 proposes only the use of a wavelet packet transformation, which makes it possible to exploit the frequency content of the signal more deeply than the discrete wavelet transformation, without allowing a lease. In contrast, in the present invention, it is the discrete wavelet transformation that is used which allows the frequency content of a signal to be distributed for frequency bands of different widths. The entropy analysis for the relevant sub-signals makes it possible to locate the defects. In the remainder of this description, various examples of applications are described with reference to the corresponding figures. Example 1: In this example, an assembly composed of six solid oxide fuel cells or cells (denoted c1 to c6) is studied in nominal condition according to four repetitions (as illustrated in FIG. 1b). These tests in nominal condition are named respectively C0-1, C0-2, C0-3 and CO-4. Then, sixteen "deviant" operating conditions were studied; these tests in deviant conditions are named respectively from C1 to C16. Of these sixteen deviant operations, C4 and C12 are considered as "abnormal operating conditions to be detected" (gas use is high level and low current). At each condition, the voltage signals from cells C1 through C6 were studied according to the process described above. The CEECs for each signal of these cells c6 to c6 are calculated and presented in Figure 5.

It can be seen from this FIG. 5 that the value distributions of the CEECs for the signals of each stack C1 to C6 under the abnormal operating conditions (C4 and C12) are different from those at the other conditions: the energy percentages present on the details (detail D1, detail D2, detail D3, detail D4 and detail D5) are not in descending order.

The relative entropy ER of the CEECs for each signal studied was then calculated using the reference signal CO-1, these calculations are represented in FIG. 6. As illustrated in FIG. 6, for each stack, the ER entropies for the conditions C4 and C12 have a peak compared to the other computed entropies. These calculations show that the relative entropy ER makes it possible to distinguish abnormal operating conditions (C4 and C12) from tolerant / normal ones. Example 2: In this example, the test was conducted on the same fuel cell module as that of Example 1 above. On the other hand, in this example, the cell c3 in the middle of the module is seriously degraded.

FIG. 7, which represents the relative entropy of each stack with the condition CO-3, shows that the stack c3 has greatly deteriorated: the relative entropy of this stack c3 is much greater than that of the other stacks. Comparing FIG. 7 to FIG. 8, it is observed that this value (greater than 0.7) is also greater than the relative entropy values when the cl and c3 stacks (before the degradation) operate at abnormal conditions. C4 and C12 (respectively less than 0.5 and greater than 0.15). Consequently, it is possible here to define a threshold, noted here Threshold 1, for the value 0.5 in order to differentiate the fault of the system PER from that in the stack PAC.

In addition, another threshold, noted here Threshold 2, can also be set to the value 0.15 in order to distinguish the operation mode with a fault from the faultless one. FIG. 8 gives the values of the relative entropy ER of the degraded c3 stack (from condition C6) and of the healthy stack C1 under different operating conditions. Before the degradation of this stack c3, the entropy values of these voltage signals at C0-1 and C10 are almost zero. Conversely, after the degradation of this stack c3, the value of the relative entropy ER at each operating condition is greater than 0.5, that is to say, above the threshold of differentiation (Threshold 1) between system faults that of the battery. For the cell cl healthy, when it operates under abnormal conditions C4 and C12, the ER value of its voltage signal is therefore between the two thresholds Threshold 1 and Threshold 2, indicating the occurrence of system fault . Thus, the various examples described here make it possible to appreciate the operation that can be performed of the entropy to determine the precise location of the defects in an electrochemical assembly, by using thresholding discrimination. It should be observed that this detailed description relates to a particular embodiment of the present invention, but in no case this description is of any nature limiting to the subject of the invention; on the contrary, its purpose is to remove any imprecision or misinterpretation of the claims that follow.

Claims (14)

REVENDICATIONS1. A method of detecting a defect in an assembly (ENS) comprising an electrochemical device (PAC) and a peripheral system (PER) of said electrochemical device (PAC), said method comprising: a) a decomposition step (Si) of applying a wavelet transform (Y) on a signal (f) measured on the electrochemical device for breaking down said signal (f) into a plurality of sub-signals each represented by wavelet coefficients c1 (k), each corresponding sub-signal at a resolution level j of said signal, b) a quantization step (S2) of estimating the energy distribution (Ej) of each sub-signal from the wavelet coefficients c1 (k) to quantize the level. information contained in each sub-signal, c) a determination step (S4) of calculating at least one physical quantity (ER) for selected sub-signals containing a predetermined significant information level selected for determining a signature of the signal, and d) a locating step (S5) of analyzing the signature of the signal to locate a defect in said set (ENS). 2. Probe method according to claim 1, characterized in that the distribution of the energy (Ej) of each sub-signal is estimated, during the quantization step (S2), by calculating the sum of the absolute values at square wavelet coefficients c (k) of each sub-signal according to the following formula: = c1 (k) 25 3. Probe method according to one of claims 1 or 2, characterized in that the physical quantity (ER) calculated during the determination step (S4) is a relative entropy (ER). 4. Probe method according to any one of the preceding claims, characterized in that, during the determination step (S4), the physical quantity (ER) is calculated for the distribution of the energy of the selected sub-signals. . The probe method as claimed in any one of the preceding claims, characterized in that, in the determination step (S4), the calculated physical quantity (ER) is a relative entropy (ER) for the distribution of the energy of the selected sub-signals, said relative entropy (ER) being calculated according to the following formula: ER = * ln (/ A) qk where pk is a percentage of energy of the wavelet coefficients (ck) equal to: Ek EkP k = -'- total E E. and q is a vector of the energy percentages of a reference signal. 6. Probe method according to any one of claims 1 to 3, characterized in that, during the determination step (S4), the physical quantity (ES, ER) is calculated for the wavelet coefficients c 1 (k) of each selected sub-signal. 7. Probe method according to any one of the preceding claims, characterized in that it comprises a selection step (S3) of selecting according to a statistical analysis the sub-signals having wavelet coefficients c 1 (k) containing a level of information satisfying a particular selection criterion. 8. The probe method as claimed in claim 7, characterized in that the selection criterion is static or dynamic. 9. Probe method according to any one of the preceding claims, characterized in that the analysis of the signal signature consists, during the locating step (S5), of comparing the variations of the signal signature with a predetermined discrimination threshold (Threshold, Threshold2). 10. The probe method as claimed in claim 9, characterized in that, when the comparison performed during the locating step (S5) has not made it possible to detect a fault, it comprises the reiteration of the selection step (S3). ) by modifying the selection criterion determined. A computer program (PG) having instructions adapted for performing the steps of the probe method according to any one of claims 1 to 10 when said computer program (PG) is executed by a computer. A computer-readable recording medium (CI) on which a computer program (PG) is recorded including instructions for executing the steps of the probe method according to any one of claims 1 to 10. 13. In-assembly fault sensor system (100) comprising an electrochemical device (PAC) and a peripheral system (PER) of said electrochemical device (PAC), said sensor system (100) comprising: acquisition (10) configured to acquire a measurement of a signal (f) on the electrochemical device (PAC), - a computer calculator (20) configured to: - apply a wavelet transform (Y) on said signal (f) to decompose said signal (f) into a plurality of sub-signals each represented by wavelet coefficients c 1 (k), each sub-signal corresponding to a resolution level j of said signal, - to estimate the distribution of the energy (Ej) of each sub-signal from the wavelet coefficients c 1 (k) to quantize the level of information contained in each sub-signal, and - calculate at least one physical quantity (ER) for sub-signals selected to determine a signal signature, the selected sub-signals being the sub-signals containing a predetermined significant information level, and - a processing circuit (30) configured to analyze the signature of the signal to enable the location of a defect in said set (ENS) . 14. Probe system (100) according to claim 13, characterized in that it comprises computer means configured to allow the implementation of the steps of the probe method according to any one of claims 1 to 10.