DEVICE, SYSTEM AND METHOD FOR IMAGING DEFECTS IN A STRUCTURE BY TRANSMITTING AND RECEIVING MECHANICAL WAVES THROUGH THE STRUCTURE

Description

[0001] The present invention relates to a device for imaging defects in a structure by emitting and receiving mechanical waves in this structure. It also relates to an imaging method implemented by such a device and an installation comprising such a device.

[0002] It applies in particular, but not exclusively, to the fields of non-destructive testing and structural health monitoring of thin structures using guided elastic waves. In these applications, the measured elastic mechanical waves can, possibly compared to a reference state, provide information on the state of the structure and identify, locate and sometimes quantify defects such as holes, delamination, cracks, corrosion, etc.

[0003] A large number of different algorithms, based on capturing mechanical waves sequentially emitted in the structure, for example using piezoelectric transducers, can be implemented by such an imaging device. Some algorithms provide very accurate results but generally require knowledge a priori of the geometry of the structure, or at least of the characteristics of wave propagation in the structure which are derived from its mechanical properties. They then pose a problem for the imaging of complex or little-known structures such as those found in the field of aeronautics. There are also simple algorithms which do not require knowledge a priori of the geometry of the structure and its mechanical properties, but the results obtained are generally less good.

[0004] The invention applies more particularly to an imaging device implementing a simple algorithm and comprising: a plurality of N transmitters of mechanical waves intended to propagate in the structure, to be distributed over at least one surface of the structure, a plurality of P receivers of mechanical waves after their propagation in the structure, to be distributed over said at least one surface of the structure, a central unit for controlling the N transmitters and P receivers programmed to record Q ≤ N×P signals obtained from electrical signals supplied respectively by the receivers of Q different transmitter / receiver pairs of the plurality of transmitters and receivers, after reception of mechanical waves emitted by the transmitters of said Q different pairs, means for storing Q first corresponding reference signals, representative of the structure without defects, and a central unit for processing the Q signals obtained, programmed to correlate each signal obtained with the first reference signal which corresponds to it, so as to obtain Q first correlation values,and to construct an image of defect probabilities in the structure from the first Q correlation values.

[0005] Such a device is for example described in the article by Zhao et al, entitled "Active health monitoring of an aircraft wing with embedded piezoelectric sensor / actuator network: I. Defect detection, localization and growth monitoring", published in Smart Materials and Structures, IOP Publishing, volume 16, pages 1208-1217 (2007). It implements an algorithmic method called RAPID (from the English "Reconstruction Algorithm for Probabilistic Inspection of Defects") according to which each correlation value associated with a pair consisting of one of the emitters and one of the receivers is used to construct an intermediate image of probable localization of one or more defects in a predetermined elliptical zone around the emitter and the receiver considered. This zone is delimited by an ellipse whose foci are for example the emitter and the receiver of the pair considered.The probability is inversely proportional to the calculated correlation value and decreases towards 0 as one moves away from the transmitter-receiver segment towards the ellipse. We then construct the complete image of defect probabilities in the structure by summing the Q intermediate images.

[0006] It is clear that such a device offers a particularly simple imaging solution that is independent of the geometry, even complex, and the material of the inspected structure. The computation time is negligible. On the other hand, by exploiting the correlation values ​​using an elliptical model applied around each pair of transmitter and receiver, it introduces a bias clearly visible in Figure 11 of the article by Zhao et al. This bias is highly dependent on the arrangement of the transmitters and receivers on the inspected structure and filtering, which is difficult to parameterize or reproduce systematically, is essential. The result is then approximate and unreliable.In particular, the risk of detecting non-existent defects (generally called "false alarm") is significant: for example, when the transmitters and receivers are arranged in a circle around an area to be imaged, the probability of detecting a false defect in the center of the circle, where many segments made up of the different possible transmitter / receiver pairs intersect, is high.

[0007] The article by Li et al, entitled "Combination of nonlinear ultrasonics and guided wave tomography for imaging the micro-defects", published in Ultrasonics, Elsevier, volume 65, pages 87-95 (2016), proposes an improvement of the RAPID algorithm. This involves optimizing the determination of the decay parameter of the correlation values ​​in the elliptical zones considered. However, this does not solve the problem of bias and the resulting false alarms.

[0008] The article by Sakai et al., titled "Defect detection method using statistical image processing of scanning acoustic tomography", published in 2016 in IEEE 23RD INTERNATIONAL SYMPOSIUM ON THE PHYSICAL AND FAILURE ANALYSIS OF INTEGRATED CIRCUITS (IPFA), IEEE, pages 293-296, proposes a different acquisition and processing technique to construct a defect image of a structure.

[0009] It may therefore be desirable to provide an imaging device which makes it possible to overcome at least some of the aforementioned problems and constraints.

[0010] A device is therefore proposed for imaging defects in a structure by emitting and receiving mechanical waves in this structure, comprising: a plurality of N transmitters of mechanical waves intended to propagate in the structure, to be distributed over at least one surface of the structure, a plurality of P receivers of mechanical waves after their propagation in the structure, to be distributed over said at least one surface of the structure, a central unit for controlling the N transmitters and P receivers programmed to record Q ≤ N×P signals obtained from electrical signals supplied respectively by the receivers of Q different transmitter / receiver pairs of the plurality of transmitters and receivers, after reception of mechanical waves emitted by the transmitters of said Q different pairs, means for storing Q first corresponding reference signals, representative of the structure without defects, a central unit for processing the Q signals obtained, programmed to correlate each signal obtained with the first reference signal which corresponds to it, so as to obtain Q first correlation values,and to construct an image of defect probabilities in the structure from the first Q correlation values, , further comprising means for storing Q corresponding second reference signals, also representative of the structure without defects but differing from the Q first reference signals by random noise, the unit central processing unit for the Q signals obtained being further programmed to: correlating each first reference signal with the second reference signal corresponding to it, so as to obtain Q second correlation values, constructing a noisy reference image from the Q second correlation values, and subtracting the noisy reference image from the defect probability image.

[0011] Thus, the noisy reference image takes up the bias introduced by the exploitation of the Q second calculated correlation values, whatever the model used to exploit these values, and it turns out that this bias is essentially the same as that introduced by the exploitation of the Q first correlation values. It has also been noted that this bias is ultimately relatively little dependent on the way in which the first and second reference signals representative of the structure without defects are obtained, by measurements or by adding simulated noise.

[0012] Optionally, at least some of the N transmitters are part of the P receivers.

[0013] Also optionally, each of the N transmitters and P receivers is an actuator or sensor chosen from the set consisting of: a piezoelectric transducer, an electromagnetic-acoustic transducer, a microelectromechanical or nanoelectromechanical transducer, a fiber optic sensor, and a laser sensor.

[0014] Also optionally, each of the N transmitters and P receivers is capable of emitting or capturing guided waves in the structure, in particular Lamb or Rayleigh waves.

[0015] In particular, when piezoelectric transducers are used, they can all be advantageously used as both transmitters and receivers. Furthermore, if one (e.g., the thickness) of the characteristic dimensions of the inspected structure is much smaller than its other two characteristic dimensions (i.e., the length and the width), Lamb-guided ultrasonic waves propagate in the structure. If the three aforementioned characteristic dimensions of the structure are of the same order of magnitude, Rayleigh-guided ultrasonic waves propagate on the surface of the structure.

[0016] Also optionally, the Q second reference signals are differentiated from the Q first reference signals by random noise, simulated or measured, the standard deviation of which is between that of a measurable noise in the Q first reference signals and that of a useful impulse or harmonic signal also measurable in the Q first reference signals, preferably closer to that of the measurable noise than to that of the measurable useful impulse signal.

[0017] Also optionally, Q = NxP, or N = P and Q = N×(N-1). In this case, all possible different transmitter / receiver pairs are used.

[0018] Also optionally, the construction of the defect probability image and the construction of the reference noisy image each comprise the construction of Q intermediate images, each intermediate image corresponding to one of the Q different transmitter / receiver pairs and taking into account the first or second corresponding correlation value in a predetermined zone, in particular an elliptical zone, defined around the segment formed between the transmitter and the receiver of this pair.

[0019] An installation for imaging defects in a structure by emitting and receiving mechanical waves in this structure is also proposed, comprising: the structure in which an area to be imaged is located, an imaging device according to the invention, in which the N transmitters and P receivers are arranged circularly around the area to be imaged for construction of the defect probability image corresponding to this area.

[0020] An installation for imaging defects in a structure by emitting and receiving mechanical waves in this structure is also proposed, comprising: the structure in which an area to be imaged is located, an imaging device according to the invention, in which the N transmitters and P receivers are distributed within the area to be imaged for construction of the defect probability image corresponding to this area.

[0021] A method of imaging defects in a structure by emitting and receiving mechanical waves in this structure is also proposed, comprising the following steps: distributing, on at least one surface of the structure, a plurality of N transmitters of mechanical waves intended to propagate in the structure, distributing, on said at least one surface of the structure, a plurality of P receivers of mechanical waves after their propagation in the structure, controlling the N transmitters and P receivers to sequentially transmit and receive mechanical waves by the transmitters and receivers of Q ≤ N×P different transmitter / receiver pairs of the plurality of transmitters and receivers, recording Q signals obtained from electrical signals supplied respectively by the receivers of the Q different pairs, after reception of said mechanical waves, processing the Q signals obtained by correlating them respectively with Q first corresponding reference signals, representative of the structure without defects, so as to obtain Q first correlation values,construct an image of defect probabilities in the structure from the first Q correlation values, , further comprising the following steps: correlating the first Q reference signals respectively with Q corresponding second reference signals, also representative of the structure without defects but differing from the first Q reference signals by random noise, so as to obtain Q second correlation values, constructing a noisy reference image from the second Q correlation values, and subtracting the noisy reference image from the defect probability image.

[0022] The invention will be better understood with the aid of the following description, given solely by way of example and with reference to the appended drawings in which: there figure 1 schematically represents the general structure of an installation for imaging defects in a structure, according to a first embodiment of the invention, the figure 2 schematically represents the general structure of an installation for imaging defects in a structure, according to a second embodiment of the invention, the figure 3 illustrates the successive stages of an imaging process implemented by the installation of the figure 1 ou 2 , THE figures 4 et 5 represent, in front and rear views, an example of a structure to be imaged, and the figures 6A, 6B et 6C respectively illustrate an initial defect probability image, a noisy reference image and a corrected defect probability image, obtained by applying the method of the figure 3 on the structure of figures 4 et 5 .

[0023] Installation 10 illustrated on the figure 1 comprises a structure 12 having a circular area 14 to be imaged shown in gray. The structure 12 is for example a thin structure of an aeronautical part. But it could also be any part in the aerospace, automotive, oil, or more generally any industrial field in which one questions the structural state of manufactured parts, in particular thin parts, that is to say whose lateral dimensions are very large (i.e. at least greater than 10 times) compared to the thickness.

[0024] The installation 10 further comprises an imaging device 16 capable of detecting and locating one or more defects in the structure 12 by emitting and receiving mechanical waves. For this purpose, the imaging device 16 comprises several transducers 18 capable of emitting mechanical waves intended to propagate in the structure 12 and an electronic control and processing circuit 20. In the particular example of the figure 1 , the emitting transducers 18 are represented by black dots. They number N = 8 and are distributed in a circle around the circular area 14 to be imaged, this distribution being able to be regular without this being an obligation. They also fulfill a function of receiving mechanical waves after their propagation in the structure 12. In this example also, they are electrically connected to the electronic circuit 20 which controls their emissions of mechanical waves and processes the electrical signals which they provide upon reception of mechanical waves. But more generally, the wave emitters are not necessarily receivers and vice versa. Furthermore, the number N of emitters is not necessarily the same as the number P of receivers. On the other hand, at least some of the N emitters can be part of the P receivers. Also as a variant, it is possible to envisage moving the control electronics to the level of each transducer.In this case, the electrical connection between the electronic circuit 20 and the transducers 18 is no longer necessary, the transmission of the signals provided by the receivers being able to be done by telecommunication, in particular by wireless telecommunication (for example using a wireless protocol of the Wi-Fi type).

[0025] As is well known in imaging devices using mechanical waves for transmission and reception, the transducers 18 are advantageously piezoelectric transducers: such transducers perform both the functions of actuators and sensors and are very inexpensive. They act in the field of ultrasonic waves which propagate very well, in the materials generally inspected, in the form of guided Lamb or Rayleigh waves. The advantage of Lamb waves is that they interrogate the entire thickness of the structure, allowing the detection of invisible defects on the surface, but they propagate easily only in thin structures. Rayleigh waves are suitable for thick structures, but they propagate only on the surface, thus interrogating only a limited depth of the structure.More generally, each of the N transmitters and P receivers is an actuator or sensor which can be chosen from a piezoelectric transducer, an electromagnetic-acoustic transducer EMAT (from the English "ElectroMagnetic Acoustic Transducer"), a microelectromechanical transducer MEMS (from the English "MicroElectroMechanical Systems") or nanoelectromechanical transducer NEMS (from the English "NanoElectroMechanical Systems"), a fiber optic sensor, a laser sensor, or other.

[0026] Furthermore, the circular arrangement of the transducers 18 of the figure 1 around the circular area 14 to be imaged is generally referred to as a "tomography configuration", by analogy with tomographic imaging in which the transmitters and receivers are also arranged in a circle. This configuration, generally expensive in terms of transducers because it actually requires a large number of them, makes it possible to obtain fairly precise information on the health of the structure in the area surrounded by the transducers.

[0027] As previously indicated, in the non-limiting example of the figure 1 the electronic circuit 20 is connected to the transducers 18 in order to transmit to them control signals C for the emission of ultrasonic pulses and in order to receive measurement signals M in response to the emission of these pulses. The electronic circuit 20 is for example that of a computer. It has a central processing unit 22, such as a microprocessor designed to transmit to the transducers 18 the control signals C, to receive from the transducers 18 the measurement signals M and to process the latter, as well as a memory 24, for example a RAM memory, in which computer program instructions and processing data are recorded.

[0028] The computer program(s) instructions are illustrated on the figure 1 as belonging to a first portion 24A of the memory 24 and as defined: in two computer programs 26 and 28 of a first functional module 30 which, in cooperation with the microprocessor 22, forms a central unit for controlling the transducers 18, and in three computer programs 32, 34, 36 of a second functional module 38 which, in cooperation with the microprocessor 22, forms a central unit for processing the measurement signals M.

[0029] The computer program 26 of the central control unit 22, 30 comprises instructions defined for sequentially controlling the transducers 18 in emission and reception of guided waves in the structure 12. More precisely, several different transmitter / receiver pairs can be formed from the N = 8 transducers 18: at most N 2 < = 64 since each transducer 18 can be both transmitter and receiver. In particular, even if the interest is limited in terms of image quality, it is possible to measure what the transmitter receives from a wave. But in general, and in particular in the aforementioned state of the art, at most N×(N-1) = 56 pairs are considered because each transmitter is advantageously used in transmission, that is to say that we do not measure what the transmitter receives from a wave.Furthermore, in theory, a transmitter / receiver pair should give the same results as the reversed pair in which the transmitter becomes receiver and the receiver becomes transmitter, but in practice this is never the case, in particular because of the noise generated by the measurement. Considering all of the N×(N-1) = 56 possible pairs from the N = 8 transducers 18 therefore presents an advantage in averaging the results. More generally, for an imaging device 16 with N transmitters and P receivers, N×P different pairs can be formed at most. Thus, in accordance with the present invention, the computer program 26 includes instructions for successively controlling in transmission and reception Q ≤ N×P (or Q ≤ N 2< or Q ≤ N×(N-1), if N = P, in particular when the transmitters are also the receivers) different transmitter / receiver pairs. At best, Q = N×P pairs can be formed.In practice, a pulse or harmonic wave of finite duration is emitted by the transmitter of each couple when it is actuated by the control unit 22, 30, propagates in the structure 12, then is received by the receiver of the actuated couple.

[0030] Q electrical signals are thus sequentially supplied by the receivers of the Q different transmitter / receiver pairs formed. They are digitized, then processed by executing the computer program 28 to be recorded in a second portion 24B of the memory 24 dedicated to the processing data of the computer programs 32, 34, 36. Alternatively, as is generally practiced in the technical field considered, it is advisable to control each transmitter / receiver pair several times in transmission and reception, for example 10 to 1000 times depending on the quality of the electrical signals received, to receive several electrical signals per pair and obtain a signal to be recorded in memory 24B by an average calculation. This makes it possible to increase the signal-to-noise ratio of the measurement.

[0031] In particular, when the structure 12 is healthy, that is to say without defects, Q first reference signals S REF1 can be obtained and recorded in memory 24B using a first sequence of measurements executed by the control unit 22, 30, with or without repetitions of transmissions and receptions for each transmitter / receiver pair formed. They are representative of the structure 12 without defects. A second sequence of measurements executed by the control unit 22, 30 under the same conditions can make it possible to obtain and record Q second reference signals S REF2 , also representative of the structure 12 without defects, but differing from the Q first reference signals S REF1 by random noise, this random noise being inherent to any sequence of measurements. Alternatively, the Q second reference signals S REF2 can also be obtained by adding simulated random noise to the Q first reference signals S REF1 .

[0032] When it is desired to know subsequently whether the structure 12 remains healthy or whether it has one or more defects, a new sequence of measurements executed by the control unit 22, 30 makes it possible to obtain and record in memory 24B Q signals noted S which are then to be compared with the reference signals S REF1 and / or S REF2.

[0033] For this purpose, the computer program 32 of the central processing unit 22, 38 comprises instructions defined for correlating each of the Q signals S with the first reference signal which corresponds to it in the set of the Q first reference signals S REF1 , so as to obtain Q first correlation values ​​V 1 , ..., VQ . These correlation values ​​are obtained in a manner known per se, for example according to the teaching of the document by Zhao et al previously cited: they are scalar values, advantageously normalized between 0 and 1. A correlation value of 1 indicates a total correlation between one of the signals S and the first reference signal which corresponds to it, that is to say a minimum probability of presence of at least one defect in a neighborhood (to be defined) of the segment constituted by the transmitter / receiver pair considered.On the contrary, a correlation value of 0 indicates a zero correlation between one of the signals S and the first reference signal which corresponds to it, that is to say a maximum probability of presence of at least one defect in a neighborhood (to be defined) of the segment constituted by the transmitter / receiver pair considered. We can therefore deduce Q first probability values ​​P 1 , ..., PQ , defined as follows: ∀i, 1 ≤ i ≤ Q, P i = 1 - V i .

[0034] The computer program 34 of the central processing unit 22, 38 comprises instructions defined for constructing an image Im INIT of probabilities of defects in the structure 12 from the first Q correlation values ​​V 1 , ..., VQ . In a simple embodiment as inspired by the teaching of the RAPID algorithmic method of the document by Zhao et al cited above, the image Im INIT is more precisely constructed from the first Q probability values ​​P 1 , ..., PQ . Each first probability value P i associated with the i-th transmitter / receiver pair formed is used to construct an intermediate image of probable location of at least one defect in a predetermined elliptical zone around the transmitter and receiver considered.This area is for example delimited by an ellipse whose foci are the transmitter and the receiver of the couple considered, or by an ellipse whose major axis is the segment formed between the transmitter and the receiver of the couple considered. The pixel values ​​in each intermediate image are maximum and a function of the first probability value P i corresponding to the segment formed by the transmitter and the receiver considered. Then these pixel values ​​decrease towards 0 when moving away from the transmitter-receiver segment towards the ellipse. The complete image of defect probabilities Im INIT is then constructed by summing the Q intermediate images. The teaching of the aforementioned document by Li et al can also advantageously be used to improve the instructions of the computer program 34.

[0035] The instructions of the computer program 32 also make it possible to correlate each of the Q first reference signals S REF1 with the second reference signal which corresponds to it in the set of Q second reference signals S REF2 , so as to obtain Q second correlation values ​​V' 1 , ..., V' Q and to deduce therefrom Q second probability values ​​P' 1 , ..., P' Q . Similarly, the instructions of the computer program 34 also make it possible to construct a noisy reference image Im REF from the Q second correlation values ​​V' 1 , ..., V' Q or, more precisely, from the Q second probability values ​​P' 1 , ..., P' Q .

[0036] Finally, the computer program 36 of the central processing unit 22, 38 comprises instructions for constructing a corrected image Im CORR of defect probabilities by subtracting, from pixel to pixel, the noisy reference image Im REF from the image Im INIT of defect probabilities: Im CORR = Im INIT - Im REF.

[0037] The 10' installation illustrated on the figure 2 includes the same electronic control and processing circuit 20 as the installation 10 of the figure 1 , but differs in the number and location of the transducers. The imaging device 16' of the installation 10' in fact comprises a number N = 9 of transducers 18', represented by black dots, distributed within the area 14' to be imaged, this distribution being able to be regular without this being an obligation. The area 14' is for example rectangular or square and the distribution of the transducers 18' is for example matrix-like. Such a matrix arrangement is generally referred to as an "imaging configuration", according to which the transducers 18' are scattered in the area 14' to be imaged of the structure 12, forming several elementary cells making it possible to globally identify in which cell of transducers a defect is located.

[0038] The electronic control and processing circuit 20 has been presented as being a computer device with a processor executing a set of computer programs 26, 28, 32, 34, 36 stored in memory 24A. These computer programs are themselves presented as distinct but this distinction is purely functional. They could just as well be grouped according to all possible combinations in one or more software programs. Their functions could also be at least partly microprogrammed or microwired in dedicated integrated circuits. Thus, as a variant, the electronic circuit 20 could be replaced by an electronic device composed solely of digital circuits (without a computer program) for carrying out the same actions.

[0039] An imaging method that can be implemented by the device 16 or 16' of the figure 1 ou 2 will now be detailed.

[0040] During a first step 102 of a first acquisition phase 100, the N transducers 18 or 18' are distributed over a surface of the structure 12, for example around the area 14 or inside the area 14' to be imaged. More generally, the first step 102 consists of distributing N transmitters of mechanical waves intended to propagate in the structure 12 and P receivers of mechanical waves after their propagation in the structure 12. More generally also and according to the envisaged application, the transmitters and receivers can be distributed over several surfaces of the structure 12 if it actually comprises several, in particular on two opposite faces. This first step 102 is executed when the structure 12 is considered to be without defects, so as to obtain and record reference signals according to the methods defined previously.Alternatively, if the structure 12 without defects cannot be previously obtained, step 102 can be carried out on a sample similar to what the structure 12 without defects should have been, preferably identical or almost identical in all respects to the structure 12 without defects, including in terms of instrumentation. It is thus also possible to obtain the reference signals according to the methods defined previously.

[0041] During a following step 104, the N transducers 18 or 18' are actuated sequentially by the central control unit 22, 30, in particular by executing the programs 26 and 28, to obtain and record the Q first reference signals S REF1 corresponding to the actuation of Q transmitter / receiver pairs among the N×(N-1) different possible pairs, with or without repetitions of transmissions and receptions for each transmitter / receiver pair formed.

[0042] During a following step 106, which may represent a very long duration compared to the execution times of steps 102 and 104, the structure 12 undergoes constraints such that one or more defects are likely to appear in its interior volume or on the surface. This may for example occur during a manufacturing or use step involving the structure 12. Alternatively, if the structure 12 without defects is not previously available, this step 106 may not exist.

[0043] Finally, during a last step 108 of the acquisition phase 100, the N transducers 18 or 18' are again actuated sequentially, according to the same sequence as that of step 104, by the central control unit 22, 30, in particular by executing the programs 26 and 28, but this time on the structure 12 with one or more possible defects, to obtain and record the Q signals S corresponding to the actuation of the aforementioned Q transmitter / receiver pairs, with or without repetitions of transmissions and receptions for each transmitter / receiver pair formed. For this step 108, it may be advantageous to respect as much as possible the same inspection conditions, for example the same environmental conditions, as those of step 104, or at least to approach them. But this is not an obligation.It has in fact been found that even if there is a significant deviation from the initial inspection conditions (those of step 104) at step 108, an imaging method according to the invention remains capable of providing usable results in terms of detection and location of defects.

[0044] The acquisition phase 100 is followed by a processing phase 200 executed by the central processing unit 22, 38.

[0045] A preliminary simulation step 202 consists of artificially obtaining the Q second reference signals S REF2 by adding simulated random noise to the Q first reference signals S REF1 . The Q first reference signals S REF1 being made up of both continuous measurement random noise and useful impulse or harmonic signals of finite duration, it is very easy to extract therefrom certain statistical parameters of noise and useful signal, in particular the mean and the standard deviation. The simulated random noise added to the Q first reference signals S REF1 to obtain the Q second reference signals S REF2 is thus for example of zero mean and of standard deviation possibly between that of the noise measurable in the Q first reference signals S REF1 and that of the useful impulse signal also measurable in the Q first reference signals S REF1 .Preferably, the standard deviation of the simulated random noise is closer to that of the measured noise than to that of the measured useful pulse signal. It is even advantageously of the same order of magnitude as that of the measured noise. Alternatively, as discussed previously, it is possible to replace the simulation step 202 with two successive sequences of measurements in step 104, to successively obtain and by measurement the reference signals S REF1 and S REF2 .

[0046] A first correlation step 204, carried out by executing the computer program 32 on the signals S and S REF1 , makes it possible to obtain the first Q correlation values ​​V 1 , ..., VQ .

[0047] A first construction step 206, carried out by executing the computer program 34 on the first Q correlation values ​​V 1 , ..., VQ , makes it possible to obtain the image Im INIT of defect probabilities.

[0048] A second correlation step 208, carried out by executing the computer program 32 on the signals S REF1 and S REF2 , makes it possible to obtain the Q second correlation values ​​V' 1 , ..., V' Q .

[0049] A second construction step 210, carried out by executing the computer program 34 on the Q second correlation values ​​V' 1 , ..., V' Q , makes it possible to obtain the reference noisy image Im REF .

[0050] Finally, a final subtraction step 212, carried out by executing the computer program 36 after steps 202 to 210, makes it possible to obtain the corrected image Im CORR.

[0051] There figure 4 illustrates in front view 12A an example of structure 12 to be imaged. This is for example a rectangular composite panel with rivet holes and stiffeners, in particular a portion of an airplane wing. The rivet holes 40 are visible on the figure 4 and represented by small circles arranged in two columns and a line at the bottom edge. The panel itself may be slightly curved, which is not visible on the figure 4 .

[0052] There figure 5 illustrates the same composite panel 12 in rear face view 12B. The stiffeners are visible in this figure: they comprise two vertical thick stiffeners 42 and seven horizontal thin stiffeners 44. The two thick stiffeners 42 cover the two columns of rivet holes up to near the bottom line of rivet holes.

[0053] Structure 12 of the figures 4 et 5 is a good example of a complex thin structure, because the presence of the diffraction rivet holes 40 and the stiffeners 42, 44 strongly disturbs the propagation of the guided volume or surface mechanical waves in the structure 12. It is very difficult to model the exact propagation of the guided mechanical waves in this type of structure due to its complexity. State-of-the-art algorithms known to provide very accurate results subject to knowledge a priori of the geometry of the structure, or at least of the propagation of waves in the structure, are therefore inapplicable for this type of complex structure or require a prohibitive simulation effort.

[0054] THE figures 6A, 6B et 6C are images obtained in the reference frame of the front face 12A of the composite panel of the figures 4 et 5 by arranging sixteen transducers distributed according to the black points indicated on these figures 6A, 6B et 6C . According to the reference mark indicated in these figures, the composite panel is of width equal to approximately 100 cm and of height equal to approximately 60 cm. A defect 46 on the centimeter scale was artificially created for the purposes of the experiment, for example using two small magnets arranged opposite each other on both sides of the structure 12: the advantage of this type of artificially created defect is that it is removable while exerting mechanical constraints observable as a defect, in particular in terms of creation of reflected waves.

[0055] There figure 6A is the Im INIT image of defect probabilities obtained at the end of step 206. It corresponds to what can be obtained by direct application of the RAPID algorithm. The lighter the pixels, the higher the probability of the presence of a defect. On the contrary, the darker they are, the lower the probability of the presence of a defect. More precisely, the value of the pixels represents an arbitrary scale proportional to the probability of the presence of a defect, as detected by the method detailed previously up to step 206. It appears that the figure 6A does not allow the location and characterization of defect 46. It would even be likely to orient the observer towards a false detection in the central part of the image. This is due to the bias introduced by the chosen image construction method.

[0056] There figure 6B is the noisy reference image Im REF obtained at the end of step 210 from the measurements carried out on the structure 12 without the defect 46. It fully incorporates the bias introduced by the chosen image construction method, linked in particular to the geometry of the structure 12 and the placement of the transducers 18. We again see a non-zero probability of the presence of a defect in the central part of the image, although this image was obtained from S REF1 and S REF2.

[0057] There figure 6C is the corrected image Im CORR of defect probabilities obtained at the end of step 212. The bias introduced by the chosen image construction method is almost completely eliminated by subtracting the image Im REF from the image Im INIT, which makes it possible to reveal a lighter area effectively centered around the defect 46. This lighter area of ​​detection of the defect 46 was absolutely not visible on the figure 6A . Again, the pixel value scale of the figure 6C only provides relative information: the probability of a defect being present is much greater in the light area than in the rest of the image.

[0058] It is clear that an imaging device such as one of those, 16 or 16', described previously makes it possible to eliminate the bias generated by the image construction method from the calculated correlation values, this bias being highly dependent on the positioning of the transducers, or more generally of the transmitters and receivers, and the geometry of the inspected structure. This ultimately makes it possible to ensure robust and much more reliable and precise detection of the defect(s), while considerably reducing the risk of false detections, all this in a context of algorithmic processing which remains particularly simple compared to other imaging methods requiring knowledge of the geometry and mechanical properties of the inspected structure.

[0059] It will also be noted that the invention is not limited to the embodiments described above.

[0060] In particular, the construction of intermediate images by elliptical zones could be adapted according to different possible shapes of zones around the transmitter / receiver pairs formed and according to different possible empirical settings. In other words, the implementation of the present invention is not limited to the context of the RAPID algorithm as taught in the document by Zhao et al. cited above or improved in the document by Li et al. cited above.

[0061] It will more generally appear to those skilled in the art that various modifications may be made to the embodiments described above, in light of the teaching just disclosed to them. In the following claims, the terms used should not be interpreted as limiting the claims to the embodiments set forth in the present description, but should be interpreted to include all equivalents which the claims are intended to cover by their wording and the prediction of which is within the reach of those skilled in the art by applying their general knowledge to the implementation of the teaching just disclosed to them.

Claims

1. A device (16; 16') for imaging defects (46) in a structure (12) by transmitting and receiving mechanical waves in this structure, including: - a plurality of N transmitters (18; 18') of mechanical waves intended to propagate in the structure (12), to be distributed over at least one surface of the structure, - a plurality of P receivers (18; 18') of mechanical waves after their propagation in the structure (12), to be distributed over said at least one surface of the structure, - a central unit (22, 30) for controlling the N transmitters (18; 18') and P receivers (18; 18') programmed to record Q ≤ N×P signals (S) obtained from electrical signals respectively provided by the receivers of Q different transmitter / receiver pairs of the plurality of transmitters and receivers, after reception of mechanical waves transmitted by the transmitters of said Q different pairs, - means (24B) for storing Q corresponding first reference signals (SREF1), representative of the structure (12) without defects, - a central unit (22, 38) for processing the Q signals (S) obtained, programmed to correlate (32) each signal (S) obtained with the first reference signal (SREF1) that corresponds to it, in such a way as to obtain Q first values of correlations, and to construct (34) an image (ImINIT) of probabilities of defects in the structure (12) from the Q first values of correlations, characterized in that it further includes means (24B) for storing Q corresponding second reference signals (SREF2), also representative of the structure (12) without defects but differing from the Q first reference signals (SREF1) by random noise, and in that the central unit (22, 38) for processing the Q signals (S) obtained is further programmed to: - correlate (32) each first reference signal (SREF1) with the second reference signal (SREF2) that corresponds to it, in such a way as to obtain Q second values of correlations, - construct (34) a reference noisy image (ImREF) from the Q second values of correlations, and - subtract (36) the reference noisy image (ImREF) from the image (ImINIT) of probabilities of defects.

2. The imaging device (16; 16') according to claim 1, wherein at least a part of the N transmitters (18; 18') is part of the P receivers (18; 18').

3. The imaging device (16; 16') according to claim 1 or 2, wherein each of the N transmitters (18; 18') and of the P receivers (18; 18') is an actuator or a sensor chosen from the set consisting of: - a piezoelectric transducer, - an electromagnetic acoustic transducer, - a microelectromechanical or nanoelectromechanical transducer, - a fiber-optic sensor, and - a laser sensor.

4. The imaging device (16; 16') according to any one of claims 1 to 3, wherein each of the N transmitters (18; 18') and of the P receivers (18; 18') is capable of transmitting or receiving guided waves in the structure (12), in particular Lamb or Rayleigh waves.

5. The imaging device (16; 16') according to any one of claims 1 to 4, wherein Q = N×P, or N = P and Q = N×(N-1).

6. A system (10) for imaging defects (46) in a structure (12) by transmitting and receiving mechanical waves in this structure, including: - the structure (12) in which a zone (14) to be imaged is located, - an imaging device (16) according to any one of claims 1 to 5, wherein the N transmitters (18) and P receivers (18) are disposed circularly around the zone (14) to be imaged for a construction of the image (ImINIT) of probabilities of defects corresponding to this zone (14).

7. A system (10') for imaging defects (46) in a structure (12) by transmitting and receiving mechanical waves in this structure, including: - the structure (12) in which a zone (14') to be imaged is located, - an imaging device (16') according to any one of claims 1 to 5, wherein the N transmitters (18') and P receivers (18') are distributed inside the zone (14') to be imaged for a construction of the image (ImINIT) of probabilities of defects corresponding to this zone (14').

8. A method for imaging defects (46) in a structure (12) by transmitting and receiving mechanical waves in this structure, including the following steps: - distributing (102), over at least one surface of the structure (12), a plurality of N transmitters (18; 18') of mechanical waves intended to propagate in the structure (12), - distributing (102), over said at least one surface of the structure (12), a plurality of P receivers (18; 18') of mechanical waves after their propagation in the structure (12), - controlling (108) the N transmitters (18; 18') and P receivers (18; 18') to sequentially transmit and receive mechanical waves by the transmitters and receivers of Q ≤ N×P different transmitter / receiver pairs of the plurality of transmitters and receivers, - recording (108) Q signals (S) obtained from electrical signals respectively provided by the receivers of the Q different pairs, after reception of said mechanical waves, - processing the Q signals (S) obtained by respectively correlating (204) them with Q corresponding first reference signals (SREF1), representative of the structure (12) without defects, in such a way as to obtain Q first values of correlations, - constructing (206) an image (ImINIT) of probabilities of defects in the structure (12) from the Q first values of correlations, characterized in that it further includes the following steps: - respectively correlating (208) the Q first reference signals (SREF1) with Q corresponding second reference signals (SREF2), also representative of the structure (12) without defects but differing from the Q first reference signals (SREF1) by random noise, in such a way as to obtain Q second values of correlations, - constructing (210) a reference noisy image (ImREF) from the Q second values of correlations, and - subtracting (212) the reference noisy image (ImREF) from the image (ImINIT) of probabilities of defects.

9. The method for imaging device according to claim 8, wherein the Q second reference signals (SREF2) differ from the Q first reference signals (SREF1) by random noise, simulated or measured, the standard deviation of which is between that of a noise measurable in the Q first reference signals (SREF1) and that of a useful pulse or harmonic signal also measurable in the Q first reference signals (SREF1), preferably closer to that of the measurable noise than to that of the measurable useful pulse signal.

10. The method for imaging device according to claim 8 or 9, wherein the construction (216) of the image (ImINIT) of probabilities of defects and the construction (210) of the reference noisy image (ImREF) each include the construction of Q intermediate images, each intermediate image corresponding to one of the Q different transmitter / receiver pairs and taking into account the corresponding first or second correlation value in a predetermined zone, in particular an elliptical zone, defined around the segment formed between the transmitter and the receiver of this pair.