WO2023110749A1 - Method for non-destructive testing of a set of substantially identical structures - Google Patents

Abstract

The invention relates to a method for non-destructive testing of a set of a plurality of structures (401, 402), the structures (401, 402) of said set being substantially identical, each structure (401, 402) being provided with a plurality of identical elastic wave sensors positioned at substantially identical locations on each of the structures, the method comprising the steps of: for each pair of sensors, determining a signal characteristic of the propagation of an ultrasonic elastic wave in the structure along a path connecting the sensors of the pair; grouping the signals obtained into groups corresponding to substantially identical paths between sensors; extracting from each signal a plurality of characteristics representative of the signal; within each group of signals, determining a damage indicator representative of a distance between each vector of characteristics associated with a pair of sensors and an average of said vectors of characteristics over the whole of the signals of the group; and comparing the damage indicator with a predetermined threshold.

Description

DESCRIPTION

Title of the invention: Method of non-destructive testing of a set of substantially identical structures

[0001] The invention relates to the field of non-destructive testing of mechanical structures or integrated health testing and more specifically to methods aimed at detecting the presence of defects on these structures and imaging them.

More specifically, the invention relates to methods based on guided ultrasonic elastic waves and non-destructive testing applied to sets of substantially identical structures, for example a set of blades of an aircraft engine or a set of blades of a wind turbine or even a set of sections of the same pipe. More generally, the invention also applies to ultrasonic waves, such as bulk waves.

[0003] In particular, the invention applies advantageously to structures which are geometrically complex and/or whose materials are anisotropic.

[0004] Non-destructive testing methods for structures, in particular methods using guided waves, are generally based on a comparison between an initial measurement, carried out when the structure is considered sound, and a subsequent measurement carried out at a time when the structure is potentially damaged.

[0005] A problem arises when the environmental or operational conditions of the control vary with respect to those of the reference state. Environmental conditions concern, for example, temperature or humidity. Operational conditions relate to mechanical changes impacting the structure during its operational functioning. Such mechanical changes can affect wave propagation. For example, in the case of the blades of a rotating engine, the centrifugal force affects the propagation of the waves in a non-uniform way, but in a substantially identical way on all the blades.

[0006] This problem arises particularly for integrated health monitoring applications for structures for which sensors integrated into a structure to be monitored are used. An advantage of this technology is to be able to carry out the check the structure frequently and without the need to stop the operation of the structure.

[0007] In such a case, the comparison between the initial measurement and the subsequent measurement is distorted because it is not possible to differentiate the contribution, in the measurement, of a defect, from that of a variation of environmental conditions.

This phenomenon is illustrated in FIG. 1 which represents an ST structure to be controlled by means of a pair of ultrasonic sensors E, R.

[0009] To check the state of the structure between the sensors E and R, an ultra-sound wave UL is emitted by the sensor E and propagates in the structure ST until it is measured by the sensor R.

[0010] At the reference instant (a), the structure ST is subjected to temperature conditions T° _o .

At a later time (b), a fault appears between sensor E and sensor R. The temperature conditions are identical to those of the reference state.

The signals measured by the sensor R at the reference instant (a) and at the instant (b) are represented superimposed on the diagram 101. The diagram 102 represents the result of the subtraction of the two signals from the diagram 101 which reveals the signature of a defect since the other variations of the signal linked to the environmental conditions are identical for the two measurements.

[0013] Conversely, when the temperature conditions evolve towards a value T°i at a time c), the variations of the signal linked to the environmental conditions are no longer identical between the reference state and the measurement at instant c). Diagram 103 represents the superposition of the signals at times a) and c). Diagram 104 represents the difference between the two signals. Note on diagram 104 that the signature of the fault can no longer be differentiated with respect to the differences in signal variations linked to changes in temperature.

[0014]There is therefore a need for a method of non-destructive testing of a structure equipped with integrated sensors which overcomes the evolution of the environmental and operational conditions of the structure to be controlled.

[0015] Reference [1] describes a non-destructive testing method which consists of positioning several pairs of transmitter-receiver sensors on the structure to be control in order to define several substantially identical propagation paths between the different pairs of sensors. Substantially identical propagation paths are understood to mean paths between two points of the structure having the same distance, the same geometry, the same mechanical properties and being subjected to the same environmental and operational conditions. By comparing the measurements obtained on the different identical paths, it is possible to detect a fault on one of the paths when the corresponding measurement differs from that of the other paths.

[0016] This method has the advantage of not depending on the evolution of the environmental conditions affecting the structure over time since the similarity of the paths chosen implies the assumption that they are all impacted in the same way by the changes in state of the structure.

The aforementioned method is illustrated in Figure 2 which shows an example of a structure equipped with three pairs of sensors defining three identical paths 201, 202, 203. Paths 201, 202 are healthy while path 203 is impacted by a fault. The measurements 211 ,212 carried out using the sensors defining the healthy paths 201 ,202 make it possible to generate a reference measurement 213 characterizing the healthy state of the structure. By calculating the difference between this reference measurement 213 and a measurement 214 taken on the path 203, a signature 215 of the fault is obtained.

[0018] An application of this method to a composite panel is described in reference [2],

[0019] A disadvantage of this method is that it requires defining strictly identical paths from the point of view of the propagation of the ultrasonic waves along these paths and of the operational conditions of operation of the structure. In practice, this constraint cannot always be respected for all structures, for example if the material of the structure is anisotropic or if the geometry of the structure is complex and presents variations in thickness and curvatures or even because of variability related to the sensors and their coupling to the structure. Another disadvantage of this method is that it requires comparing the raw measurements of the signals with each other, which does not always lead to a reliable result.

The invention proposes to solve the limitations of the aforementioned methods by exploiting several substantially identical structures or several substantially identical parts of the same structure which are subject to the same environmental conditions. The invention is particularly suitable for the non-destructive testing of systems composed of several substantially identical structures such as the sets of blades of an aircraft engine.

The subject of the invention is a method for the non-destructive testing of a set of several structures supporting ultrasonic elastic wave propagation modes, the structures of said set being substantially identical, each structure being equipped with a plurality of identical elastic wave sensors positioned at substantially identical locations on each of the structures, the method comprising the steps of:

- For each pair of sensors, determine a signal characteristic of the propagation of an elastic ultrasonic wave in the structure along a path connecting the sensors of the couple,

- Group the signals obtained into groups corresponding to paths between sensors that are substantially identical from the point of view of the propagation of the wave along the path,

- Extract from each signal, a plurality of characteristics representative of the signal in the form of a vector of characteristics,

- Within each group of signals, determine a damage indicator representative of a distance between each characteristic vector associated with a pair of sensors and an average of said characteristic vectors over all the signals in the group,

- Compare the damage indicator to a predetermined threshold and conclude that there is a fault on the path between a pair of sensors if the damage indicator associated with this pair is greater than said threshold. [0023] In a variant embodiment, the method also comprises a preliminary step of defining several groups of paths between sensors that are substantially identical from the point of view of the propagation of the wave along the path, taking into account at least one of the constraints from among: a constraint of geometrical similarity, a constraint of similarity of the composition of the materials forming the structures, a constraint of similarity of the environment to which the structures are subjected, a constraint of similarity of the operational conditions to which the structures are subjected.

[0024] According to a particular aspect of the invention, the representative characteristics extracted from the signal are taken from among: the energy of the signal, the maximum amplitude of the signal, an instant corresponding to a particular point of the signal, the frequency of the maximum of the signal, the value of the maximum of the signal in frequency, the coefficients of a decomposition into wavelets applied to the signals, frequency and time centroids.

[0025] In a variant embodiment, the method further comprises the steps of:

- Spatially sample each structure at a plurality of points,

- Calculate an ellipse around each pair of sensors,

- Assign to each point located inside an ellipse the value of the damage indicator calculated for the pair of sensors,

- For each point located inside several ellipses, sum the corresponding damage indicator values,

- Compare the damage indicator to a predetermined threshold and conclude that there is a defect around the point if the damage indicator associated with this point is greater than said threshold.

[0026] In a variant embodiment, the method further comprises: a prior step of determining initial damage indicators for the same structures in a healthy initial state, a step of subtracting the initial damage indicators from the damage indicators determined for said structures in a subsequent state.

In a variant embodiment, the method comprises a step of calculating, for each group of signals, statistical moments applied to the feature vectors and comprising at least one mean calculation and one covariance matrix.

[0028] In a variant embodiment, the step of calculating statistical moments is carried out iteratively by deleting, at each iteration, the vector of characteristics having a distance greater than a highest predetermined threshold with respect to the average of said vectors.

According to a particular aspect of the invention, the step of determining a signal characteristic of the propagation of an ultrasonic elastic wave in the structure along a path connecting the torque sensors is carried out by emitting an elastic wave ultrasound from the first torque sensor and measuring with the aid of the second torque sensor the ultrasonic elastic wave propagated in the structure.

According to a particular aspect of the invention, the step of determining a signal characteristic of the propagation of an ultrasonic elastic wave in the structure along a path connecting the torque sensors is carried out by:

- Using the two torque sensors, measuring the propagation of an ultrasonic elastic wave in the structure, resulting from mechanical vibrations impacting the structure in its operational behavior;

- Calculating a correlation between the signals measured by the two torque sensors.

- Convoluting the result of the correlation with a predetermined spectrum signal.

The invention also relates to a non-destructive testing system for a set of several structures supporting modes of guided propagation of ultrasonic elastic waves, the structures of said set being substantially identical, the system comprising a plurality of sensors elastic waves intended to be positioned at substantially identical locations on each of the structures and a calculation unit configured to execute the steps of the non-destructive testing method according to the invention from the signals measured by said sensors.

In a variant embodiment, the system according to the invention further comprises a display interface for displaying a map of said structures from the calculated damage indices.

According to a particular aspect of the invention, the elastic wave sensors are chosen from among piezoelectric transducers, electromagnetic acoustic transducers or fiber optic Bragg grating sensors.

The invention also relates to a set of several structures supporting modes of guided propagation of ultrasonic elastic waves, the structures of said set being substantially identical and being equipped with the non-destructive testing system according to the invention, the sensors of elastic waves of said system being at substantially identical locations on each of the structures.

According to a particular aspect of the invention, said assembly being constituted by a set of blades of an aircraft engine.

According to a particular aspect of the invention, said assembly being constituted by a set of identical cylindrical sections connected to form a pipe.

Other characteristics and advantages of the present invention will appear better on reading the following description in relation to the following appended drawings.

[0038] [Fig. 1] illustrates the principle of a non-destructive testing method based on a reference state according to the prior art,

[0039] [Fig. 2] illustrates another non-destructive testing method based on a comparison between identical paths on the same structure,

[0040] [Fig. 3] illustrates, on a diagram, the principle on which the invention is based,

[0041] [Fig. 4] illustrates an example of arrangement of sensors on two substantially identical structures, [0042] [Fig. 5] schematizes, on a flowchart, the steps for implementing a non-destructive testing method according to the invention,

[0043] Fig. 6] illustrates an example of mapping of a structure obtained with the invention.,

[0044] [Fig. 7] represents a diagram of a preferential application of the invention to control the state of health of the blades of an aircraft engine.

Subsequently, the invention is described in the case of application of guided ultrasonic waves, however the invention applies identically using bulk waves.

The principle of the invention consists in carrying out several measurements of guided ultrasonic waves between pairs of transmitter-receiver sensors located at the same positions on substantially identical structures in such a way that the wave propagation travels identical paths. vis-à-vis the geometry of the part, the properties of the material but also the environmental and operational conditions. For example, in the case of a structure corresponding to a pipe, it can be considered that the pipe, of constant section, is formed of several identical parts between them from a geometric point of view and being composed of the same materials. However, these different parts of the same pipe can be subjected to different environmental conditions. This is the case, for example, when part of the pipe is in the sun and another part in the shade. In this case, the different parts of the pipe are subjected to different temperatures which can impact the propagation of the waves even on geometrically identical paths. In addition, the operational conditions of the structure's operation can also have an impact on wave propagation. For example, if part of the pipe has hot content and another part of the pipe has cold content, the temperature differences will also impact signal propagation. Wave propagation variabilities can also come from vibrations impacting different parts of the pipe differently.

Subsequently, the notion of substantially identical paths must be understood as meaning paths having substantially the same geometry, traversing structures composed of the same materials and being subjected to the same environmental and operational conditions. Of course, the expression “substantially identical” must be understood as meaning identical to the tolerances ready for manufacture, integration of the sensors on the structure and measurements of the impact of the environmental and operational conditions.

Furthermore, the sensors positioned on the structures to define the paths must also be substantially identical to limit the variability of the operating conditions of the sensors.

FIG. 3 schematizes N substantially identical structures which undergo the same operational and environmental conditions (in particular temperature, humidity) over time. On each of these structures, a pair of sensors E, R is positioned at the same positions. An identical propagation path is therefore defined on all the structures. In the absence of a defect, the measured signals are supposed to be very similar, whatever the instant of measurement since the influence of the environmental conditions on the signals is identical on each structure.

[0050] Consequently, the appearance of a fault on one of the paths (for example on the structure i identified in FIG. 3 modifies the corresponding signal. A comparison between all the measurements makes it possible to identify the one that stands out of the others and which corresponds to a potential defect. To do this, a damage indicator calculation is carried out for each structure and compared to a defect detection threshold as illustrated in the bottom diagram of figure 3.

[0051] Figure 4 shows a top view of two substantially identical structures 401, 402 of the type described in Figure 3. These structures have a variation in thickness and a curvature in the lateral direction and an invariance in the direction longitudinal as can be seen in Figure 3.

In the example of Figure 4, twelve sensors are positioned on each of the structures 401, 402 at identical positions. On the right of FIG. 4, a few identical paths have been shown between pairs of sensors in view of the mechanical properties and the geometry of the structures. For example, the 8 horizontal propagation paths 0-1, 3-4,6-7,9-10,12-13, 15-16, 18-19 and 21-22 are mutually identical.

The 8 horizontal propagation paths 1-2, 4-5,...22-23 are identical to one another. They are different from the previous paths because the structures 401, 402 are not symmetrical with respect to their longitudinal axis.

The 6 vertical propagation paths 0-3, 3-6, ... 18-21 are identical to one another.

The 6 diagonal propagation paths 0-4, 3-7, ... 18-22 are identical to one another.

The 6 diagonal propagation paths 0-5, 3-8, ... 18-23 are identical to each other.

The above list is not exhaustive and other groups of identical propagation paths can be defined.

The flowchart in Figure 5 details the steps for implementing the non-destructive testing method according to one embodiment of the invention.

The first step 501 consists in acquiring a set of signals by means of the sensors positioned on the structures 401, 402. For example, waveform acquisition 501 is a multi-channel simultaneous acquisition. Within each structure 401, 402, each sensor is excited in turn to emit an ultrasonic elastic wave which propagates in the structure. The other sensors measure the propagated ultrasonic signals. For example, the sensors are piezoelectric transducers or electromagnetic acoustic transducers or fiber optic Bragg grating sensors.

The excitation signals are, for example, signals comprising a series of square pulses centered at a chosen frequency or sinusoid cycles.

In a pre-processing step, a filtering time window is applied to the measured signals in order to keep only the useful part of the signal and to possibly eliminate the reflections of the waves at the end of the part as well as the beginning of the signal which corresponds to the electronic coupling with the acquisition system. In a variant embodiment of the invention, no excitation signal is used but the sensors each carry out a passive acquisition of the ultrasonic waves generated by mechanical disturbances of the structure. These disturbances come for example from ambient noise, or from air friction on the structure or from vibration of the structure. All the sensors positioned on the structures 401, 402 then operate in reception to measure the propagation of these signals in the structure. The acquisition step 501 then comprises additional steps. For each pair of sensors defining a path, the signals measured by the two sensors of the couple are correlated, then the result of the correlation is convolved with an excitation signal to obtain a response over a limited frequency band. By carrying out these two additional processing steps, a signal is obtained comparable to that obtained with the first embodiment (sensors operating as transmitter and receiver successively). Those skilled in the art will find more information on the implementation of passive acquisition using ultrasonic wave sensors, for example in patent applications FR3060743A1, FR3105554A1, W02019091705A1.

At step 502, the different signals measured are then separated into groups of signals corresponding to a family of identical paths for which the signals can be compared. The notion of identical paths depends on the geometric specificities of the structure as well as on the material that composes it. For example, horizontal and vertical paths may be comparable if the part is isotropic and not comparable otherwise. In the example of figures 3 and 4, the structure considered has an absence of left-right symmetry, but an invariance according to the height. In such an example, the following paths constitute a family: 0-1, 3-4, 6-7, 9-10, 12-13, 15-16, 18-19, 21-22. A second distinct family is formed by 1-2, 4-5... 22-23 due to the lack of symmetry.

At the end of step 502, groups of comparable signals are therefore obtained.

An additional criterion to be considered in the definition of the propagation paths adopted is the number of identical paths. Indeed, when the number of identical paths in a group is too low, this group of paths is not retained because the statistical processing applied to the signals are subject to too many biases. In the example of FIG. 4, the four paths 0-8, 3-11, 12-20 and 15-23 are identical to one another, but only four in number. The minimum number of paths in a group depends on the degree of similarity between the paths.

The more the paths are similar to each other, that is to say the fewer variations there are in terms of the environmental and operational conditions, the material and the sensors, the more the signals acquired for these paths are similar. Consequently, the impact of structural damage on one of these signals will then be easily detectable by comparison. If the paths have significant variations between them, then it is necessary to have a statistical sample of paths in greater number to be representative of these variations.

[0068] In step 503, for each signal, a set of characteristics representative of the signals are generated. They can be, without being limited thereto, the energy of the signal calculated in all or part of the useful frequency band and/or in a time window of interest, the maximum amplitude of the signal, instants corresponding to particular points of the signal (for example the maximum or a change to 0), the frequency of the maximum and the value of the maximum in frequency, the coefficients of a decomposition into wavelets, the frequency and temporal centroids. Document [3] lists a set of possible characteristics.

At the end of step 503, each signal indexed by i is represented by a vector of characteristics Xi which corresponds to a point in a multidimensional space.

In step 504, for each group of signals determined in step 502, a set of first statistical moments is then calculated from the characteristic vectors X of the group. More precisely, the mean and the covariance matrix of the vectors X are calculated for each group.

[0071] Optionally, the calculation of the first statistical moments is carried out on a subset of points by filtering out certain points deemed to be aberrant. Indeed, due to the limited number of paths in a group, it is possible that the presence of a fault alters the statistical moments of an entire group. So the calculated statistical moments can be relatively far from those corresponding to a healthy set, which can have the effect of detecting as damaged a healthy signal because the statistic is “moved” towards the damaged case. To avoid this scenario, in a particular embodiment of step 504, an iterative calculation of the moments is carried out. Once the statistic has been calculated for the first time, it is checked whether a point Xi is located at too great a distance from the mean in terms of standard deviations in the associated direction (for example 3 standard deviations). Note that, as the statistic is multi-dimensional, it is necessary to calculate the standard deviation in the direction of the point considered. More precisely, if C is the covariance matrix, Xi the point considered and X _m the mean of the points, the standard deviation or standard deviation in the direction of the point Xi is given by the relation: a =

The furthest point is removed from the set to recalculate the moments. We iterate this process until we have converged on a set, and therefore a stable statistic, i.e. until there are no more points beyond a predetermined distance to the average or until the number of points remaining in the group falls below a minimum threshold.

At the end of step 504, for each group of signals, an average of the characteristic vectors and a covariance matrix are obtained. At step 505, a damage indicator equal to a distance from the point to the average calculated at step 503 is then calculated for each point Xi of each group of signals. The distance X _m is normalized by the deviation -type o calculated in the direction of the point.

According to a first embodiment of the invention, the damage indicators calculated in step 505 are compared (step 507) with a predetermined fault detection threshold. If a damage indicator exceeds the detection threshold, a fault is detected in the vicinity of the path connecting the two torque sensors associated with the point considered. The fault detection threshold is, for example, equal to three times the value of the standard deviation calculated on the damage indicators. This standard deviation can be calculated per group of signals or globally for all the signals for which the damage indicator has been calculated. In order to improve the probability of correct detection and to reduce the probability of false alarm, an additional optional step 506 is envisaged in a second embodiment. Step 506 consists in mapping the structures 401, 402 from the damage indicators calculated in step 505. In other words, a spatial sampling grid of each structure 401, 402 is defined, the grid defining a set of points of the structure for which one wishes to calculate an indicator of damage. Next, for each damage indicator calculated in step 505 and associated with a pair of sensors, an ellipse is determined around the pair of sensors, this ellipse defining a zone in which points of the spatial sampling grid are located. Ellipses are determined for each pair of sensors of the structure for which a damage indicator was calculated in step 505. The damage indicator calculated for the pair of sensors is assigned to the entire ellipse. Then, when a point of the spatial sampling grid is located in several ellipses, a damage indicator is calculated for this point equal to the sum of the damage indicators of the different ellipses. If a point is located in only one ellipse, its damage indicator is equal to that of the ellipse.

FIG. 6 schematizes, on the left, an example of an ellipse 600 defined around a pair of sensors E, R. The size of the ellipse 600 is defined so as to take into account indirect paths of the wave 602 in addition to the direct path 601 .

At the end of step 506, a map of the structures 401, 402 is obtained which gives, for each point of the structure, a damage indicator value which is then compared with a fault detection threshold at step 507. An example of mapping 610 is represented on the right of FIG. 6 where it is possible to identify the presence of a defect 611 at the intersection of several ellipses. The fault detection threshold is for example equal to half the maximum possible value for a point of the map. It can be obtained more generally by calibration in a preliminary phase. The step 507 of comparison with a fault detection threshold can be replaced by an automatic learning engine trained to recognize the fine characteristics of the maps resulting from step 506 or of the damage indicator values resulting from the step 505. The learning engine is, for example, a classifier implemented by an artificial neural network or a support vector machine type algorithm or a random forest type algorithm. The learning engine is trained from real or simulated data to recognize the characteristics representative of a defect. In this variant, the invention can be seen as a dimensionality reduction operation carried out upstream of the automatic learning step.

In another embodiment variant, in the case where the measurements carried out on the structures show great variability linked for example to the types of materials, or to the instrumentation of the sensors, an acquisition having been carried out when the structures were healthy, for example at an initial moment of commissioning. The same process described in figure 5 is applied, making it possible to determine the indicators of damage for the structures in their reference state, which makes it possible to take into account the variability of the structures and the instrumentation. By subtracting these initial damage indicators from the damage indicators calculated at a later time, measurement variability related to instrumentation and parts is reduced. Contrary to a usual reference state as envisaged in the state of the art, the invention does not directly compare the signals with each other but uses statistical indicators calculated from certain characteristics of the signals, this strategy is therefore more robust (in particular if we make the assumption that the variabilities are insensitive to the environmental conditions modifying the signals). This embodiment variant corresponds to a calibration of the degree of similarity between the paths. It has the advantage of not requiring identical environmental or operational conditions for the reference state and for the subsequent state.

The invention is implemented by means of elastic ultrasonic wave sensors.

The sensors are chosen from among piezoelectric transducers such as polyvinylidene fluoride films or ceramics based on lead titano-zirconates, electromagnetic acoustic transducers (for example of the EMAT type or magnetostrictive patches) or network sensors of Bragg on fiber optics. [0081] The sensors are positioned at predetermined locations of the structures 401 and 402. The sensor arrangements are identical on each of the structures 401, 402 as similarly as possible given the manufacturing constraints. For example, if the wavelengths of interest of the waves used are of the order of 10 millimeters, then a positioning error of the sensors of the order of 1 millimeter leads to a change in the signals measured by a tenth of a cycle. , which may be significant.

The sensors are identical to one another within the limits of manufacturing constraints. In particular, variations may exist between the sensors due to calibration faults or manufacturing imperfections.

The sensors can be fixed on the structure or directly integrated into the material of the structure.

Each sensor is connected to a signal acquisition chain and all of the sensors are connected to a processing unit which is configured to execute the damage index calculation and fault detection method described below. above.

The processing unit can be located at a distance from the structures to be controlled. Each structure is equipped with sensors and a signal acquisition chain which also performs the extraction of the characteristics of the signals (step 504). These characteristics represent low bit rate data compared to the measured raw signals and can therefore be transmitted to the remote processing unit by means of a low bit rate wireless technology, for example LoRa ^tm technology.

The processing unit can be produced in software and/or hardware form from a processor and a memory. The processor can be a generic processor, a specific processor, a GPU (Graphics Processing Unit) graphics processor, an application-specific integrated circuit (also known as an ASIC for “Application-Specific Integrated Circuit”) or a in situ programmable gate array (also known as FPGA for "Field-Programmable Gate Array")

The results provided by the processing unit can be displayed on a computer screen or directly on an interface forming part of the device. Each sensor can combine both the function of transmitter and receiver or only one of the two functions. In this second case, there is at least one sensor having the transmitter function and at least one sensor having the receiver function among the set of sensors. Advantageously, the set of sensors comprises half of sensors having the transmitter function and half of sensors having the receiver function.

An advantage of the invention is that it constitutes a physically relevant dimensionality reduction. The calculated damage indicator values correspond to a compression of the raw measurements and therefore, this is advantageous for transferring data in reduced format and/or for storing them and/or training machine learning algorithms.

FIG. 7 schematizes a preferred application of the invention which relates to the health check of a set of blades P1, P2, P3 of an aircraft engine R. As can be seen in Figure 7, the blades of a jet engine constitute substantially identical structures from a geometric point of view but also subject to similar operational and environmental conditions. Furthermore, this type of structure has a complex geometry with curvatures and variations in thickness which do not make it possible to use health monitoring methods according to the prior art. The invention advantageously makes it possible to control the appearance of defects on this type of structure.

[0091 ] References

[0092][1] Anton, S.R., Inman, D.J., & Park, G. (2009). Reference-free damage detection using instantaneous baseline measurements. AIAA journal, 47(8), 1952-1964.

[0093][2] Salmanpour, M.S., Khodaei, Z.S., & Aliabadi, M.H. (2016). Instantaneous baseline damage localization using sensor mapping. IEEE Sensors Journal, 17(2), 295-301.

[0094] [3] Morizet, Nicolas, et al. "Classification of acoustic emission signals using wavelets and Random Forests: Application to localized corrosion." Mechanical Systems and Signal Processing 70 (2016): 1026-1037.

Claims

1 . Method of non-destructive testing of a set of several structures (401, 402) supporting ultrasonic elastic wave propagation modes, the structures (401, 402) of said assembly being substantially identical, each structure (401, 402) being equipped with a plurality of substantially identical elastic wave sensors and positioned at substantially identical locations of each of the structures, the method comprising the steps of: - For each pair of sensors, determine (501 ) a characteristic signal of the propagation of an ultrasonic elastic wave in the structure along a path connecting the sensors of the couple, - Group (502) the signals obtained by groups corresponding to paths between sensors that are substantially identical from the point of view of the propagation of the wave along the path, - Extract (503) from each signal, a plurality of characteristics representative of the signal in the form of a vector of characteristics, - Within each group of signals, determining (505) a damage indicator representing a distance between each vector of characteristics associated with a pair of sensors and an average of said vectors of characteristics over all the signals of the group, - Compare (507) the damage indicator with a predetermined threshold and conclude that there is a fault on the path between a pair of sensors if the damage indicator associated with this pair is greater than said threshold. 2. Non-destructive testing method according to claim 1 further comprising a prior step of defining several groups of paths between sensors that are substantially identical from the point of view of the propagation of the wave along the path by taking into account at least one of the constraints among: a geometric identity constraint, an identity constraint of the composition of the materials forming the structures, an identity constraint of the environment to which the structures are subjected, an identity constraint of the operational conditions to which the structures are subjected. Non-destructive testing method according to any one of the preceding claims, in which the representative characteristics extracted from the signal are taken from among: the energy of the signal, the maximum amplitude of the signal, an instant corresponding to a particular point of the signal, the frequency of the maximum of the signal, the value of the maximum of the signal in frequency, the coefficients of a decomposition into wavelets applied to the signals, of the frequency and temporal centroids. Non-destructive testing method according to one of the preceding claims further comprising the steps (506) of: - Spatially sample each structure at a plurality of points, - Calculate an ellipse around each pair of sensors, - Assign to each point located inside an ellipse the value of the damage indicator calculated for the pair of sensors, - For each point located inside several ellipses, sum the corresponding damage indicator values, - Compare the damage indicator to a predetermined threshold and conclude that there is a defect around the point if the damage indicator associated with this point is greater than said threshold. A non-destructive testing method according to any preceding claim further comprising: - a preliminary step to determine initial damage indicators for the same structures in a healthy initial state, - a step of subtracting the initial damage indicators from the damage indicators determined for said structures in a subsequent state. Non-destructive testing method according to any one of the preceding claims, comprising a calculation step (504), for each group of signals, of statistical moments applied to the feature vectors and comprising at least one mean calculation and one covariance matrix. Non-destructive testing method according to Claim 6, in which the step (504) of calculating statistical moments is carried out iteratively by deleting, at each iteration, the feature vector having a distance greater than a highest predetermined threshold by relative to the mean of said vectors. Non-destructive testing method according to any one of the preceding claims, in which the step (501) of determining a signal characteristic of the propagation of an ultrasonic elastic wave in the structure along a path connecting the torque sensors is carried out by emitting an ultrasonic elastic wave from the first torque sensor and measuring with the aid of the second torque sensor the ultrasonic elastic wave propagated in the structure. Non-destructive testing method according to any one of Claims 1 to 7, in which the step (501) of determining a signal characteristic of the propagation of an ultrasonic elastic wave in the structure along a path connecting the sensors of the torque is made in: - Using the two torque sensors, measuring the propagation of an ultrasonic elastic wave in the structure, resulting from mechanical vibrations impacting the structure in its operational behavior; - Calculating a correlation between the signals measured by the two torque sensors. - Convoluting the result of the correlation with a predetermined spectrum signal. System for the non-destructive testing of a set of several structures supporting modes of guided propagation of ultrasonic elastic waves, the structures of said set being substantially identical, the system comprising a plurality of elastic wave sensors intended to be positioned at locations substantially identical on each of the structures and a calculation unit configured to execute the steps of the non-destructive testing method according to any one of the preceding claims from the signals measured by said sensors. Non-destructive testing system according to claim 10 further comprising a display interface for displaying a map of said structures from the calculated damage indices. Non-destructive testing system according to either of Claims 10 and 11, in which the elastic wave sensors are chosen from among piezoelectric transducers, electromagnetic acoustic transducers or fiber optic Bragg grating sensors. Set of several structures supporting guided propagation modes of ultrasonic elastic waves, the structures of said set being substantially identical and being equipped with the non-destructive testing system according to any one of claims 10 to 12, the elastic wave sensors of said system being at substantially identical locations on each of the structures. Set of several structures according to claim 13, said set being constituted by a set of blades of an aircraft engine. A set of several structures according to claim 13, said set being constituted by a set of identical cylindrical sections connected to form a pipe.