METHOD FOR DETECTING DEFECTS IN A STRUCTURE

Description

TECHNICAL FIELD

[0001] The invention relates to a method for detecting defects in a structure.

[0002] One field of application of the invention relates to structures used in the field of aeronautics. STATE OF THE ART

[0003] Devices for detecting structural defects are known, for example from documents US 7,937,248 and US 2015 / 0308920.

[0004] Document US 2006 / 287842 describes a method for detecting a defect in a structure by transducers, in which a graph including paths between the transducers is generated, weights are calculated from the signals of the transducers and routes among the paths between the transducers are deleted or selected.

[0005] The paper DA TIBADUIZA et al.: “Damage classification in structural health monitoring using principal component analysis and self-organizing maps” DAMAGE CLASSIFICATION IN SHM USING PCA AND SOM”, STRUCTURAL CONTROL AND HEALTH MONITORING, US, (20121213), vol. 20, no. 10, ISSN 1545-2255, pages 1303 - 1316, describes the comparison of transducer signals to a reference.

[0006] Document US 2006 / 164919 describes an ultrasonic device for detecting underwater objects by transmitting and receiving transducers.

[0007] US 4,893,025 describes a sensor for automated equipment, comprising light emitters and receivers.

[0008] Document US 2005 / 094490 describes an array of ultrasonic transducers for the non-destructive evaluation of materials.

[0009] Document US2009 / 192727 D6 describes a system for monitoring defects in a structure using surface wave transmitting and receiving transducers, which are positioned on a structure to monitor defects.

[0010] Document US 8,707,787 describes a system for detecting defects on an aeronautical structure, comprising transmitting and receiving transducers.

[0011] Document FR 2 262 303 describes a device for locating sources of stress wave emissions, comprising emitting or receiving acoustic transducers.

[0012] Document US 2014 / 123758 describes a system for monitoring the integrity of a structure using acoustic sensors distributed over a furnace to detect its deformations.

[0013] The primary objective of airlines is to ensure regular and punctual flights at competitive prices, in order to position themselves favorably against their competitors. As such, they are attentive to the maintenance costs of their equipment and are interested in working with equipment manufacturers who are aware of the importance of a product's availability and maintainability throughout its life cycle.

[0014] Today, most purely structural elements of aircraft are only inspected during periodic maintenance to which said aircraft are subjected (typically C-check inspections), for example every 24 months, or every 10,000 flight hours. More precisely, the maintenance of the IFS (internally fixed structure) type structure of an aircraft thrust reverser (nacelle) is today inspected by an operator who does not have decision-making support tools available.

[0015] We are looking for two areas of improvement: the first is related to security and the second is related to economic aspects.

[0016] The fact that structural integrity verification is performed only during periodic maintenance and by an unassisted human operator leaves the door open to the possibility of flying with a system that does not have 100% integrity. Here, the safety aspect would require that the structural system be replaced just before the loss of integrity, which implies the need to push preventive maintenance further.

[0017] Replacing components during Type C inspection also has two major drawbacks: firstly, systems are often replaced preventively, while they are still flight-ready; secondly, they are sometimes oversized to meet planned maintenance intervals without compromising safety. As a result, there is a cost for the operator that could be avoided or at least significantly reduced.

[0018] The aim is to continuously and / or periodically monitor the state of health of aeronautical structures. The objective is to resolve the problems mentioned above and, at the same time, to space out, simplify, or even reorganize maintenance operations.

[0019] Structural health monitoring (SHM) systems are known. An SHM system is essentially a non-destructive testing system that permanently integrates sensors. The collected data can be used to detect the onset and track the evolution of damage / defects in the structure. Ideally, the SHM system should be able to: (i) detect the presence of a defect, (ii) locate it, (iii) determine its size, and (iv) provide a prognosis on the life of the structure.

[0020] The economic challenges are, firstly, to reduce aircraft downtime, because repairs can be predicted or diagnostics automated. Secondly, the challenge is to replace only the structures or components that really need to be replaced, and therefore to have fewer elements to replace, and therefore to produce.

[0021] Instrumentation of the structure allows measurements to be obtained, for example acoustic or guided wave measurements (e.g. Lamb waves) which must be processed and analyzed in order to define whether or not the structure has a defect.

[0022] According to the state of the art, the analysis is currently carried out by comparing the measurements taken on a structure to be analyzed (called the sample under analysis) with those taken on a reference structure (called the reference sample), which is by definition in a healthy state.

[0023] Measurements made on the reference structure are often taken in well-defined reference thermal states. This choice implies a constraint, which is that measurements on the structure to be analyzed cannot be made before it has reached a thermal state identical to the reference state in which the measurements on the reference structure were made.

[0024] This problem involves costs linked to delays in analysis and therefore delays in aircraft availability. SUMMARY OF THE INVENTION

[0025] The invention aims to obtain a method for detecting defects in a structure, in particular an aeronautical structure, making it possible to carry out an evaluation of the defects without resorting to the reference state, but which is based only on the information at its disposal (of the current state) to identify the presence of a defect.

[0026] For this purpose, a first object of the invention is a method for detecting defects in a structure according to claim 1.

[0027] Claims 2 to 14 relate to embodiments of the detection method. QUICK DESCRIPTION OF THE FIGURES

[0028] The invention will be better understood on reading the description which follows, given solely as a non-limiting example with reference to the appended drawings, in which: there figure 1 schematically represents a modular block diagram of a fault detection device according to an embodiment of the invention, the figure 2 schematically represents an example of the installation of transducers of the method and of the fault detection device in an area of a structure to be monitored of a first type, according to an embodiment of the invention, the figure 3 schematically represents an example of the installation of transducers of the method and of the fault detection device in an area of a structure to be monitored of a second type, according to an embodiment of the invention, the figure 4 schematically illustrates the overlap between two bordering emission circles of transducers, the figure 5 schematically represents a first example of meshing of a structure by transducers of the method and of the fault detection device according to an embodiment of the invention, the figure 6 schematically represents a second example of meshing of a structure by transducers of the method and of the fault detection device according to an embodiment of the invention, the figure 7 schematically represents a third example of meshing of a structure by transducers of the method and of the fault detection device according to an embodiment of the invention, the figure 8 schematically represents a flowchart of a first mode of implementation of the fault detection method according to the invention, the figure 9 schematically represents a flowchart of a second mode of implementation of the fault detection method according to the invention, the figure 10 schematically represents the structuring of a measurement matrix, which can be used according to a mode of implementation of the fault detection method according to the invention, the figure 11 schematically represents a schematic flowchart, illustrating the principle of calculating a fault indicator, which can be used according to an implementation mode of the fault detection method according to the invention, the figure 12 schematically represents an example of a result of fault detection, indicating on the abscissa a test number, and on the ordinate the value of a fault indicator according to the figure 11 , which can be used according to an implementation mode of the fault detection method according to the invention. DETAILED DESCRIPTION OF THE INVENTION

[0029] To figures 1 à 6 , the method for detecting defects in a structure and the device 11 for detecting defects in a structure according to the invention use a plurality of transducers 100 or sensors 100, which are positioned on or in an STR structure, the purpose of which is to detect and locate any defects. The device 11 according to the invention may be a system for diagnosing the condition of the structure. The method according to the invention may be a method for diagnosing the condition of the structure. The STR structure may be made of any type of material, for example composites (for example monolithic type composites and / or sandwich type composites and / or cellular composites which may be honeycomb or other). The STR structure may be a part of an aircraft, such as for example a part of aircraft turbomachines, such as for example aircraft turbojets or aircraft turboprops or a part of a thrust reverser (nacelle).The method and device 11 according to the invention can be used directly on a part of the aircraft itself, the STR structure then being permanently located in the aircraft. The method and device 11 according to the invention can also be used on a part of the aircraft, having been dismantled therefrom. The invention can be applied to an IFS type structure (internally fixed structure, or "Internal Fixed Structure" in English terminology) of a nacelle of an aircraft turbomachine or other, this structure being made of composite materials of monolithic and sandwich types, but it can be extended to any other structure of the aircraft.

[0030] The transducers 100 may be of any type, for example of the ultrasonic type (for example of the piezoelectric type or other, in particular of the PZT type, i.e. Lead Zirconium Titanate) or other, for example of the optical type (for example comprising one or more Bragg gratings of one or more optical fibers). Preferably, in the context of the invention, the plurality of transducers 100 is distributed between first transducers E capable of being in an emission mode (also called excitation mode) on the one hand, and on the other hand second transducers R capable of being in a reception mode. When a transducer 100 is in the emission mode, this first transducer E emits a predetermined excitation signal. When a transducer 100 is in the reception mode, this second transducer R receives a reception signal in response to the excitation signal emitted by a first transducer E.The excitation signals and the reception signals are capable of propagating through the STR structure, for example along the STR structure or in the STR structure, for example along a surface SUR of the STR structure. When the excitation signal emitted by a first transducer E does not encounter defects in the STR structure, the reception signal received by a second transducer R corresponds to this excitation signal or is equal to this excitation signal. When the excitation signal emitted by a first transducer E encounters one or more defects in the STR structure, the reception signal received by a second transducer R does not correspond to this excitation signal and is disturbed with respect to this excitation signal. Advantageously, the first transducers E are of the ultrasonic type (for example of the piezoelectric type or other, in particular of the PZT type ieto Lead Titano-Zirconate), and the second transducers R are of the optical type (for example comprising one or more Bragg grating(s) of one or more optical fiber(s)).

[0031] Each transducer 100 may comprise means 101 for connection to the exterior as shown in figures 2 And 3 , or wireless. Also provided with the transducers 100 are power supplies and the electronics necessary for their implementation. These connection means 101 serve to send the control signal and the electrical power supply from an external unit 102 to each transducer 100 and to send the reception signals and / or the excitation signals to this external unit 102, as shown in figure 1 This external unit 102 is for example an electronic module and is used for acquiring the signals emanating from the transducers 100. This external unit 102 may include a supervisor whose purpose is to repatriate the measurements (reception signals and excitation signals) carried out by the transducers 100.

[0032] The transducers 100 have known or predetermined positions relative to the STR structure. The transducers 100 may be arranged, for example, on the same surface SUR of the STR structure, as shown in figures 2 And 3 . There figure 2 represents transducers 100a, 100b, positioned on the surface SUR of a structure STR formed from a monolithic composite material of the IFS type structure mentioned above. The figure 3 represents transducers 100a, 100b positioned on the surface SUR of a structure STR formed of a honeycomb cell sandwich type composite material of the above-mentioned IFS type structure. On the figures 2 And 3 , the elements 100a correspond to first transducers E of ultrasonic type (for example of piezoelectric type or other, in particular of the PZT type ie Lead Titano-Zirconate), and the elements 100b to second transducers R of optical type which comprise one or more Bragg grating(s) 100b of one or more optical fiber(s) 105. More precisely, as visible in these figures, one or more optical fiber(s) 105 run(s) on, or in, the STR structure, and comprise(s) a multitude of Bragg gratings 100b, which are placed so as to act as second transducers R.

[0033] The transducers 100 may also be arranged in the STR structure, for example by belonging to a SUR surface embedded inside the STR structure. The SUR surface may be immaterial, in that it constitutes the geometric locus grouping together all the positions of the transducers 100. The SUR surface may also be material, for example consisting of a film applied (for example by gluing) to the STR structure, or embedded inside the STR structure. Alternatively, this material SUR surface may form a sheet arranged between two successive sheets of composite materials, before or after the curing of said materials. In any event, the SUR surface may not constitute a two-dimensional plane, for example when the SUR surface matches the particular shape of the STR structure. Nevertheless, it is always possible to project the SUR surface onto a two-dimensional horizontal plane.In fact, when the SUR surface is material, the transducers 100 are first positioned on this flat surface (i.e. two-dimensional), before being integrated into the STR structure. The particular positioning of the transducers 100 in the context of the invention, which will now be described, must therefore be understood as a positioning on the two-dimensional horizontal plane, whatever the position of the SUR surface in space when it is integrated into the STR structure.

[0034] The network formed from the plurality of transducers 100 allows for optimal analysis coverage of the STR structure, while using a minimal number of transducers 100. The propagation of the excitation signals from a first transducer E is of the free-field and omnidirectional wave propagation type. Furthermore, in the context of the invention, it is assumed that this propagation is concentric with respect to the first transducers E. Each first transducer E emitting a mechanical wave which propagates uniformly, the second transducers R are therefore arranged so as to be reachable by at least one transducer 100 in transmission mode. There is thus an overlap between two bordering emission circles, as visible on the figure 4 . In order to guarantee optimal analysis coverage with a minimum number of transducers 100, in a preferred embodiment of the detection device 11 according to the invention, the surface SUR is covered with a minimum number of emission circles E while guaranteeing minimal overlap between two bordering emission circles E.

[0035] In this regard, as visible on the figure 5 , the first E transducers are arranged to delimit between them several M meshes, according to a hexagonal mesh. The first E transducers are therefore not all aligned with each other. The M meshes are bordering each other. The M meshes have known positions relative to the STR structure. The corners of the M meshes are formed by the first E transducers. Each M mesh surrounds an area located between three unaligned first E transducers or more than three unaligned first E transducers. As visible on the figure 5 , each mesh M is delimited by three unaligned first transducers E and is therefore triangular. Each mesh M can have one, two, three, or more than three meshes bordering it. The M meshes are equilateral triangular, so as to form the hexagonal mesh. Each vertex of the equilateral triangular M meshes thus corresponds to the position of a first transducer E. For example, at figure 5 , the ten meshes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, which are equilateral and adjacent to one another, and which are delimited by a first row R1 of three aligned first transducers E, a second row R2 of four first transducers E aligned with each other and not aligned with row R1 and a third row R3 of three first transducers E aligned with each other and not aligned with rows R1 and R3, are represented by fictitious continuous lines between the first transducers E.

[0036] From this hexagonal mesh of the plurality of first transducers E, different embodiments of positioning the plurality of second transducers R have been illustrated on the figures 6 And 7 . On the figure 6 , each first transducer E is surrounded by three second transducers R arranged on an emission circle centered on the first transducer E. On the figure 7 , six second transducers R are distinguished per first transducer E. Advantageously, the distance separating two first transducers is between 1 and 2 times the radius of the emission circle, preferably between 1.5 and 2 times this radius, for example 1.73 times this radius.

[0037] According to an embodiment of the invention, in a first step E1 of acquiring measurements, each first transducer E emits an excitation signal, and the second transducers R which surround it on its emission circle acquire reception signals.

[0038] In a second calculation step E2, subsequent to the first measurement acquisition step E1, a signature S is extracted or determined, from the excitation signals of the first transducers 100 E and the reception signals of the second transducers R, for example by a calculation unit 103, connected to the acquisition unit 102. The signatures S can for example be calculated by a signal processing and multivariate analysis tool, by the fact that the signatures S are extracted from measurement matrices, calculated as a function of the excitation signals and the reception signals. In addition, this calculation unit 103 implements the algorithm described below and can be a fault detection algorithmic module.The calculation unit 103 is automatic, such as for example one or more calculators and / or one or more computers, and / or one or more processors and / or one or more servers and / or one or more machines, which can be programmed in advance by a pre-recorded computer program. The meshes M and / or the position of the meshes M relative to the STR structure and / or the position of the transducers 100 and / or the membership of the transducers 100 to the meshes M are calculated and / or recorded and / or pre-recorded in the calculation unit 103.

[0039] In a third step E3 of defect identification, subsequent to the second step E2 of calculation, the signatures S are compared with each other. Among the meshes M, one or more meshes M are identified as locating a defect in the STR structure, called a faulty mesh (mesh in a damaged state), when the signature S of this or these meshes is different from several other signatures of several other meshes.

[0040] According to one embodiment, the mesh(es) having been identified as faulty mesh(es) and their known position relative to the STR structure are sent in an indication signal to a human-machine interface 104 to be returned to a user, during a fourth indication step E4, subsequent to the third fault identification step E3. The user is thus provided with their known position of the faulty mesh(es) relative to the STR structure and therefore the location of the faults thus detected. The human-machine interface 104 allows for example use by maintenance teams. The human-machine interface 104 is for example connected to the calculation unit 103.

[0041] The device 11 for detecting defects in the STR structure comprises, for example, the transducers 100, the calculation unit 103, the acquisition unit 102 and the human-machine interface 104. The detection device 11 implements the method for detecting defects in the STR structure, in particular thanks to the calculation unit 103 which implements the extraction steps E2 and comparison E3.

[0042] A first embodiment, called majority vote, of this comparison during the third step E3 of fault identification is described below, with reference to the figure 8 The faulty mesh is identified as having a signature S that is different from the signatures S, which are equal to each other or similar to each other, of at least two other meshes. That is, at least two other meshes, independently of each other, are each adjacent to the faulty mesh, for reasons of thermal inertia. The comparison is for example made in several groups of three meshes or more than three meshes, each group being different from the other groups by at least one mesh. Each group has for example an odd number of meshes.

[0043] According to one embodiment, groups of three meshes M are considered below. According to one embodiment, an identification of the defective meshes is thus applied according to a majority criterion 2 out of 3. It is considered that the probability of having two meshes having the same defect in the same places is very low. Of course, each group of meshes could have more than three meshes M, in particular an odd number of meshes. The meshes of each group are adjacent to each other.

[0044] Each group is defined by successively a first mesh, a second mesh and a third mesh bordering each other. Each group therefore has a starting mesh, which is the first mesh. For example, the case of a group of meshes bordering each other is the first group G1 of three meshes, defined by successively the first mesh 1, the second mesh 2, which borders the first mesh 1, and the third mesh 6, which borders the first mesh 1 at the figure 4 or to the second stitch.

[0045] A first function for identifying any group of meshes defined by successively a first mesh, a second mesh and a third mesh, bordering each other, is defined below, taking the example below of the first group G1. The identification function can be applied to the first group G1 and to any group of meshes other than the first group G1, and therefore to any group of meshes which may be other than mesh 1 and / or mesh 2 and / or mesh 6 of the first group G1.

[0046] By the first identification function, the first mesh 1, the second mesh 2 and the third mesh 6 are identified as non-failing meshes 1, 2, 6, when the signatures S of these meshes 1, 2, 6 are equal.

[0047] By the first identification function, when the signature S of the first mesh 1 is different from the signature S of the second mesh 2 equal to the signature S of the third mesh 6, the first mesh 1 is identified as a faulty mesh.

[0048] By the first identification function, when the signature S of the first mesh 1 is equal to the signature S of the second mesh 2 and is different from the signature S of the third mesh 6, the first mesh 1 is identified as an undamaged mesh (or healthy mesh) and the third mesh 6 is identified as a suspect mesh.

[0049] The third identification step E3 may comprise one or more first identification sub-step(s) E31 and one or more second identification sub-step(s) E32.

[0050] Thus, in the first identification sub-step E31, the first identification function is applied to the first group G1 of successively meshes 1, 2 and 6. This first identification sub-step E31 is therefore carried out for the first group G1 having mesh 1 as starting mesh.

[0051] According to one embodiment, in the second identification sub-step E32, subsequent to the first identification sub-step E31, the first identification function is applied to another group G2 of meshes having as starting mesh the second mesh 2.

[0052] For example, this other group G2 of meshes is defined by successively the second mesh 2, then the first mesh 1, if the first mesh 1 has been identified as a non-defective mesh, and a fourth mesh 3, which is adjacent to the starting mesh (mesh 2 in this case), then a fifth mesh, which is adjacent or not adjacent to the first mesh 1, if the first mesh 1 has been identified as a defective mesh, and the fourth mesh 3, which is adjacent to the starting mesh (mesh 2 in this case).

[0053] According to one embodiment, the fourth mesh 3 and the fifth mesh 7 are different from the third mesh 6, which is identified as a suspect mesh. For example, the first mesh 1, the second mesh 2, third mesh 6, the fourth mesh 3 and the fifth mesh 7 are distinct from each other. For example, the fifth mesh, in the case where it is not adjacent to the first mesh 1, may be mesh 7.

[0054] According to one embodiment, the second identification sub-step E32 is repeated one or more times on respectively one or more other groups of meshes. The starting meshes of the successive second identification sub-steps E32 may be different from each other.

[0055] According to one embodiment, the starting mesh of each second identification sub-step E32 is adjacent to the starting mesh of the second identification sub-step E32 preceding it. The identification is thus carried out on groups from one to the next.

[0056] According to one embodiment, the meshes of each group are other than a mesh having been identified as a faulty mesh.

[0057] According to one embodiment, for example, the first identification function is applied to another group G3 of meshes having as starting mesh the fourth mesh 3, bordering the second mesh 2.

[0058] For example, this other group G3 of meshes is defined by successively the fourth mesh 3, a seventh mesh, which is adjacent or not adjacent to the starting mesh and which has not been identified as a faulty mesh, and an eighth mesh, which is adjacent or not adjacent to the starting mesh, a seventh mesh, which is adjacent or not adjacent to the starting mesh, if the second mesh (2) has been identified as a non-faulty mesh, and an eighth mesh, which is adjacent or not adjacent to the starting mesh. The seventh mesh can be adjacent to the starting mesh (fourth mesh 3) and be mesh 4 or mesh 2, if mesh 2 has not been identified as a faulty mesh. The eighth mesh can be adjacent to the starting mesh (fourth mesh 3) and be mesh 8 or mesh 2. One way to make this choice can be to choose the two meshes that were used the least at the time of the analysis, in order to limit the risk of error.

[0059] According to one embodiment, for the third mesh 6 or each mesh, which is identified as a suspect mesh, a second identification function defined in the following manner is applied.

[0060] When the signature of the third mesh 6 or suspect mesh is different from the signature of the first mesh 1 equal to the signature of a sixth mesh 7 bordering the third mesh 6 or suspect mesh, the third mesh 6 or suspect mesh is identified as a faulty mesh.

[0061] When the signature of the third mesh 6 or suspect mesh is different from the signature of the first mesh 1 different from the signature of the sixth mesh 7, the third mesh 6 or suspect mesh is identified as a faulty mesh and the sixth mesh 7 is identified as a suspect mesh.

[0062] When the signature of the third mesh 6 or suspect mesh is equal to the signature of the sixth mesh 7, a second indication signal is sent to the human-machine interface 104, for example to request a decision from the user, this case however occurring rarely.

[0063] When the suspect mesh (for example 6) borders a mesh identified as a faulty mesh (for example 1), the second identification sub-step E32, the first identification sub-step E31 and / or the second identification function are applied to another mesh (for example 7) bordering the suspect mesh.

[0064] According to one embodiment, the second identification sub-step E32, the first identification function and / or the second identification function are applied so that the starting mesh is in turn each of the meshes, without being a mesh having been identified as a faulty mesh.

[0065] A second embodiment, called population classification, of this comparison is described below during the third step E3 of fault identification, with reference to the figure 9 .

[0066] The third identification step E3 may include one or more third classification sub-step(s) E33 and one or more fourth classification sub-step(s) E34.

[0067] During the third sub-step(s) E33 of classification, the meshes M having the same signature are classified into the same respective family F.

[0068] The meshes M, which are classified into the respective family F having the largest number of meshes, called healthy family, are identified as non-failing meshes.

[0069] Each mesh M, which is classified into a respective family F having only one mesh M, called the respective faulty family, is identified as a faulty mesh.

[0070] It is unlikely to have the same type of defect on several meshes, and therefore the population with the most individuals is considered to be a healthy mesh population, while we consider all populations with only one individual to be failing mesh populations.

[0071] According to one embodiment, in the population construction phase, each cell of the structure is classified into a family in order to subsequently analyze the families and deduce whether there are any defective families. Each population is defined by a reference.

[0072] According to one embodiment, to classify the meshes, a first mesh 1 having a first signature S1 is classified into a first family F1, to which a respective reference equal to the first signature S1 is assigned. The reference is thus defined for the first population, the number of individuals of which is equal to 1 (for the first mesh 1).

[0073] Then successively for each other mesh k (for example mesh 2, 3, 4, 5, 6, 7, 8, 9, 10), different from the first mesh 1, we iterate the fourth sub-step E34 of classification described below.

[0074] We compare the signature S2 of the other mesh k to the respective reference of each family F (including family F1), for example by successively comparing the signature S2 of the other mesh k to the respective reference of the families Fj one after the other.

[0075] If the signature S2 of the other mesh k is equal to the respective reference Fj of one of the families F, then this other mesh k is classified in this family Fj. Thus, in this case, the number of individuals of the family Fj is incremented by 1. If the signature S2 of the other mesh k is not equal to the respective reference of the family Fj, then the comparison of the signature S2 of the other mesh k is carried out with the respective reference of the following family Fj, with j incremented by 1.

[0076] If the signature S2 of the other mesh is not equal to any respective reference of the families F, Fj, then this other mesh k is classified in a new family F2, to which we attribute a respective reference equal to the signature of this other mesh k. Thus, in this case, the number of individuals of the new family F2 is equal to 1.

[0077] We then move on to the next mesh, on which we carry out the fourth sub-step E34 of classification.

[0078] This iteration is carried out, as long as there are one or more meshes not yet classified in a family F.

[0079] According to one embodiment, it is determined, by calculating distances, whether the family being analyzed is closer to the healthy family or to a failing family.

[0080] According to one embodiment, each mesh belonging neither to the healthy family nor to the respective defective family(ies), called mesh to be analyzed, is classified into a family to be analyzed.

[0081] According to one embodiment, for each mesh to be analyzed of the family to be analyzed, a first distance is calculated between the signature of the mesh to be analyzed and the signature of the meshes of the healthy family, and a second respective distance between the signature of the mesh to be analyzed and the signature of the meshes of each respective faulty family.

[0082] When the first distance is less than each respective second distance, the mesh to be analyzed is classified in the healthy family or the mesh to be analyzed is identified as a non-defective mesh.

[0083] When the first distance is greater than one or more respective second distances, the mesh to be analyzed is classified in the faulty family having this respective second distance or the mesh to be analyzed is identified as a faulty mesh.

[0084] According to one embodiment, in SHM, the calculated distances are called defect indicators. According to one embodiment, the proposed indicator is calculated using an extraction method based on multivariate analysis. The details of the calculation of this indicator are described in the remainder of this document.

[0085] For example, for all families having 2 or more individuals (meshes), but which are not the population having the most meshes, an indicator is calculated between the main populations and the one having a single individual.

[0086] Then the indication step E4, described above, is carried out.

[0087] Compared to the first embodiment, this second embodiment makes it possible to have a limited number of meshes having the same defect and therefore to overcome the problem of the 2 out of 3 majority vote with 2 defective meshes and one non-defective mesh.

[0088] According to one embodiment, an interrogation wavelength (of the excitation signal) is less than the distance between the transducers of the same mesh.

[0089] One embodiment of feature extraction is described below.

[0090] Either Y s k ∈ ℝ M × n y , the matrix emanating from a structural zone considered without failure (cf. Figure 10 , representing the structuring of the measurement matrix): Y s k = y 11 ⋯ y 1 i ⋯ y 1 n y ⋯ ⋯ … … … y m 1 ⋯ y mi ⋯ y mn y ⋯ ⋯ ⋯ ⋯ ⋯ y M 1 ⋯ y Mi ⋯ y Mn y Or : ny is the number of sensors or transducers C1, C2, ... Ck, Ck+1, ... Cny instrumented in the area, M is the number of acquisitions established to query the area, k is discrete time.

[0091] The definition of the no-failure zone is in our case linked either to the majority vote or to a higher population density, according to the algorithms described previously.

[0092] The extraction of health status characteristics is established through a multivariate analysis method, for example, principal component analysis (PCA). This method allows transforming variables that are related to each other (called correlated) into new variables that are decorrelated from each other, for example according to the document [1] Tibaduiza, DA et al (2015). "Structural damage detection using principal component analysis and damage indices". Journal of Intelligent Material Systems and Structures.

[0093] Mathematically speaking, PCA is based on a decomposition into eigenvalues and eigenvectors of the measurement matrix (see equation 1), thus making it possible to obtain a principal and residual space, defined by the following equations: Λ Σ y _ = Λ ^ n r × n r 0 0 Λ ^ M − n r × M − n r P = P ^ M × n r P ˜ M × M − n r Or : The matrices Λ̂, P̂ are called eigenvalue matrices, eigenvectors, associated with the principal space. The matrices Ã, P̂ are called eigenvalue matrices, eigenvectors, associated with the residual space.

[0094] One embodiment of the fault indicator is described below.

[0095] Let us now consider the matrix emanating from another structural zone considered suspect, and associate with this zone a measurement matrix noted Y ii , constructed in the same way as the diagram of the Figure 10 and according to equation (1).

[0096] An area considered suspect is an area having a different indicator than the area considered healthy.

[0097] To establish the defect indicator necessary for health status decision making, the matrix Y ii is projected into the ACP model (according to equations (2) and (3)) established on the area considered without failure. The fault indicator noted DI is defined by the following equation: E = Y n k I − P ^ P ^ T DI ι = E ι E i T where i corresponds to the number of the acquisition established to query the structure.

[0098] There Figure 11 illustrates a schematic diagram of the calculation of the defect indicator.

[0099] There Figure 12 illustrates the application of majority voting to the fault indicator. By comparison, we can see that the indicator from zone No. 3 is more important than that from zones No. 1 and No. 2, thus indicating that a fault is present in zone No. 3.

Claims

1. A method for detecting faults of a structure (STR), by means of a device (11) for detecting faults of a structure (STR), the device (11) comprising a calculation unit (103) and a plurality of transducers (100) intended to be positioned on or in the structure (STR), - first transducers (E) of the plurality of transducers (100) being capable of being in an emission mode where they emit an excitation signal, - second transducers (R) of the plurality of the transducers (100) being capable of being in a reception mode where they receive a reception signal in response to the excitation signal emitted by a first transducer (E) in the emission mode, the excitation signals and the reception signals being capable of propagating along the structure (STR) or in the structure (STR), the first transducers (E) forming a hexagonal meshing so as to delimit between them several mutually adjacent mesh cells (M), the second transducers (R) being positioned on respective emission circles of the first transducers (E), each emission circle of a first transducer (E) being centered on the first transducer (E), a meshing consisting of a plurality of mesh cells (M) and defined by the plurality of transducers (100) having been previously positioned on or in the structure (STR), the method comprising the steps consisting of: - extracting, (E2), depending on the excitation signals emitted (E1) by the first transducers (E) and on the reception signals received (E1) by the second transducers (R), a signature (S) for each of the mesh cells (M), - comparing (E3) the signatures (S) to one another to identify, among the mesh cells (M), at least one mesh cell as localizing a fault of the structure (STR), called a faulty mesh cell, when the signature (S) of this mesh cell (M) is different from several other signatures (S) of several other mesh cells (M), the extraction (E2) and comparison (E3) steps being implemented by the calculation unit (103), wherein the mesh cells (M) are equilateral triangles of which each vertex corresponds to the position of a first transducer (E), wherein the positioning of the transducers 100 is considered to be a positioning of their projection on a horizontal plane with two dimensions.

2. The method for detecting faults of a structure (STR) according to claim 1, wherein the first transducers (E) are piezoelectric sensors (110a), and the second transducers (R) each comprise a Bragg grating (100b) of an optical fiber (105).

3. The method for detecting faults in a structure (STR) according to one of claims 1 and 2, wherein the distance separating two first transducers (E) within the meshing is 1.73 times the radius of each emission circle.

4. The method for detecting faults in a structure (STR) according to one of claims 1 to 3, wherein the faulty mesh cell is identified as having a signature which is different from the signatures, which are equal to one another, of at least two other mesh cells (M), each independently adjacent to the faulty mesh cell.

5. The method for detecting faults in a structure (STR) according to one of claims 1 to 4, wherein a first function of identifying a group defined by, successively, a first mesh cell (1), a second mesh cell (2) and a third mesh cell (6) is also defined by the fact that: - the first mesh cell (1), the second mesh cell (2) and the third mesh cell (6) are identified as non-faulty mesh cells (1, 2, 6), when the signatures of these mesh cells are equal, - when the signature of the first mesh cell (1) is different from the signature of the second mesh cell (2) equal to the signature of the third mesh cell (6), the first mesh cell (1) is identified as a faulty mesh cell, - when the signature of the first mesh cell (1) is equal to the signature of the second mesh cell (2) and is different from the signature of the third mesh cell (6), the first mesh cell (1) is identified as a non-faulty mesh cell and the third mesh cell (6) is identified as a suspect mesh cell, - in a first identification sub-step (E31), the first identification function is applied to a first group of three mesh cells (1, 2, 6), comprising successively a first mesh cell (1), a second mesh cell (2), which is adjacent to the first mesh cell (1), and a third mesh cell (6), which is adjacent to the first mesh cell (1).

6. The method for detecting faults in a structure (STR) according to claim 5, wherein, in a second identification sub-step (E32) subsequent to the first identification sub-step (E31), the first identification function is applied to another group of mesh cells defined by successively a starting mesh cell formed by the second mesh cell (2) and: - the first mesh cell (1), if the first mesh cell (1) has been identified as a non-faulty mesh cell, and a fourth mesh cell (3), which is adjacent to the starting mesh cell, - a fifth mesh cell (7), which is adjacent or non-adjacent to the first mesh cell (1), if the first mesh cell (1) has been identified as a faulty mesh cell, and the fourth mesh cell (3) which is adjacent to the starting mesh cell.

7. The method for detecting faults in a structure (STR) according to claim 6, wherein the second identification sub-step (E32) is reiterated one or more times on respectively one or more other groups of mesh cells.

8. The method for detecting faults in a structure (STR) according to claim 7, wherein the starting mesh cell of each second identification sub-step (32) is adjacent to the starting mesh cell of the second identification sub-step (E32) preceding it.

9. The method for detecting faults in a structure (STR) according to any one of claims 6 to 8, wherein the fourth mesh cell (3) and the fifth mesh cell (7) are different from the third mesh cell (6), which is identified as a suspect mesh cell.

10. The method for detecting faults in a structure (STR) according to any one of claims 5 to 9, wherein the mesh cells of each group are other than a mesh cell having been identified as a faulty mesh cell.

11. The method for detecting faults in a structure (STR) according to any one of claims 5 to 10, wherein, for the third mesh cell (6), which is identified as a suspect mesh cell, a second identification function is applied, defined by the fact that: - when the signature of the third mesh cell (6) is different from the signature of the first mesh cell (1) equal to the signature of a sixth mesh cell (7) adjacent to the third mesh cell (6), the third mesh cell (6) is identified as a faulty mesh cell, - when the signature of the third mesh cell (6) is different from the signature of the first mesh cell (1) different from the signature of the sixth mesh cell (7), the third mesh cell (6) is identified as a faulty mesh cell and the sixth mesh cell (7) is identified as a suspect mesh cell, - when the signature of the third mesh cell (6) is equal to the signature of the sixth mesh cell (7), a first indication signal is sent to a human-machine interface (104).

12. The method for detecting faults in a structure (STR) according to one of claims 1 to 3, wherein, during at least one classification sub-step (E33, E34), the mesh cells (M) having the same signature are classified in a same respective family: - the mesh cells (M) which are classified in the respective family having the greatest number of mesh cells, called the healthy family, being identified as non-faulty mesh cells, - each mesh cell (M) which is classified in a respective family having only a single mesh cell, called a respective faulty family, being identified as a faulty mesh cell.

13. The method for detecting faults in a structure (STR) according to claim 12, wherein, to classify the mesh cells (M): - a first mesh cell (1) having a first signature (S1) is classified in a first family (F1), to which is attributed a respective reference equal to the first signature (S1), - then successively for each other mesh cell (k) different from the first mesh cell (1), the classification sub-step (E34) is iterated according to which ∘ the signature (S2) of the other mesh cell (k) is compared to the respective reference of each family (F1), ∘ if the signature (S2) of the other mesh cell (k) is equal to the respective reference of one (Fj) of the families (F), then this other mesh cell (k) is classified in this family (Fj), ∘ if the signature (S2) of the other mesh cell (k) is not equal to any respective reference of the families (F), then this other mesh cell (k) is classified in a new family (F2), to which is attributed a respective reference equal to the signature (S2) of this other mesh cell (k).

14. The method for detecting faults in a structure (STR) according to claim 12 or 13, wherein: - for each mesh cell belonging neither to the healthy family, nor to the respective faulty family(ies), called the mesh cell to be analyzed, a first distance is calculated between the signature of the mesh cell to be analyzed and the signature of the mesh cells of the healthy family, and a respective second distance between the signature of the mesh cell to be analyzed and the signature of the mesh cells of each respective faulty family is calculated to: ∘ when the first distance is less than each respective second distance, classify the mesh cell to be analyzed in the healthy family or identify the mesh cell to be analyzed as a non-faulty mesh cell, ∘ when the first distance is greater than one or more respective second distances, classify the mesh cell to be analyzed in the faulty family having this respective second distance, or identify the mesh cell to be analyzed as a faulty mesh cell.