Ultrasonic part inspection method

Abstract

Method for inspecting a part using ultrasound. One aspect of the invention relates to a method for inspecting a part using a signal acquired via an ultrasonic sensor, enabling the determination of geometric irregularities, structural periodicity, and defects in the part.

Description

Title of the invention: Method for inspecting a part using ultrasound TECHNICAL FIELD OF THE INVENTION

[0001] The technical field of the invention is that of non-destructive testing by ultrasound.

[0002] The present invention relates to a method of controlling a part from a signal acquired via an ultrasonic sensor. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0003] Ultrasonic non-destructive testing is a widespread inspection method in industry that allows, in particular, the determination of a part's mechanical properties and the assessment of its condition. The concept of condition here refers to the detection of anomalies in the part's structure. These anomalies include, for example, material defects, material inclusions, cracks, delamination, etc.

[0004] In particular, ultrasonic non-destructive testing is used in a production context, because this type of testing allows for a rapid and accurate evaluation of the properties of a part.

[0005] Various approaches exist to improve the control carried out in order to compensate for specificities of the structure of the part, in particular a periodicity of said structure, which are a source of noise in the acquired signals and penalize their analysis, in order to detect these anomalies.

[0006] It is known, in particular from document EP3574316B1, of methods for determining the elongated structure of a part using ultrasonic sensors. The drawback is that such a method requires the use of several sensors to characterize this structure, which implies a complexity of instrumentation and data acquisition that is not compatible with a production context, particularly in terms of practicality and speed of implementation.

[0007] Another known approach relies on a process for reconstructing an image, via an optimization mechanism, based on non-linear mapping. The drawback is that this approach is only applicable to ultrafast ultrasonic imaging, and not to more conventional imaging, typically using a single-element probe, and therefore requires specific measurement equipment.

[0008] There is therefore a need for an approach compatible with a production context, making it possible to compensate for the noise of an ultrasonic signal generated by a periodicity of structure and / or microstructure of a part, independently of the ultrasonic testing performed to inspect said part. Summary of the invention

[0009] The invention offers a solution to the problems mentioned above, by allowing the noise generated by the periodicity of the structure of a part to be filtered quickly and robustly in an ultrasonic signal acquired via any ultrasonic sensor.

[0010] A first aspect of the invention relates to a computer-implemented method for inspecting a part based on a signal acquired via an ultrasonic sensor, the signal corresponding to at least one ultrasonic wave propagating through the part, the ultrasonic sensor performing a periodic movement during signal acquisition, the part comprising a periodic structure, the method comprising: • Determine a signal from an echo of a background of the room by a regression of a propagation time corresponding to the background echo in the signal; • Determine a first filtered signal from the signal, by removing the background echo signal from the signal; • Determine a second filtered signal from the signal by removing a first periodic content and a second periodic content, the first periodic content corresponding to a frequency content generated by the periodic displacement of the ultrasonic sensor during the acquisition of the signal, the second periodic content corresponding to a frequency content generated by the periodic structure of the part; • Detect, in the second filtered signal, an indication corresponding to a defect in the part.

[0011] The term "periodic displacement" means a displacement performed by the ultrasonic sensor during acquisition, following a trajectory with a periodic pattern (i.e., the trajectory is at least periodic per segment). For example, the trajectory may be linear per segment (typically triangular or with a predefined size increment to perform several linear paths along the part), elliptical around the part, circular with a predefined size increment to perform several circles around said part, or be a zigzag displacement in the same plane, or any other form of displacement exhibiting periodicity in its trajectory.

[0012] The term "periodic structure" means that the internal organization of the material of the part includes one or more periodicities, for example in the form of fibers, grains, or any other periodic property related to the structural organization of the material of the part. By way of example, a part with a periodic structure is a A composite or metal alloy part with elongated, unidirectional or multidirectional grains. The periodic structure may be the same throughout the thickness of the part or may vary, as is the case for a composite part which has a layered periodic structure.

[0013] The term "background echo signal" refers to a portion of the signal captured by the ultrasonic sensor that corresponds to the thickness of the part. It is therefore a set of harmonics whose respective amplitudes depend on the thickness in question. This background echo signal is determined, in particular, from the echo of the wave generated by the sensor after backpropagation along the bottom of the inspected part. It is thus the portion of the signal whose amplitude is substantially constant and whose propagation time corresponds to that of a round trip between the sensor and the bottom of the part. The background echo signal exhibits, for example, variations that depend on variations in the thickness of the part and that are significantly slower (by at least two orders of magnitude) than those of the part's texture.

[0014] Thanks to the invention, it is possible to filter the different sources of noise in the acquired ultrasonic signal, in order to raise the signal-to-noise level and thus improve the detection and characterization of defects present in the inspected part.

[0015] Furthermore, it is possible, in a single acquisition, to determine three properties of the part that are usually determined through several separate inspections and treatments. In particular, the proposed method makes it possible to determine the thickness of the part via the background echo signal, the periodicity of the structure via the second periodic content, and to detect the presence of one or more defects in the part.

[0016] It may be appreciated that the proposed approach does not require specific equipment to be implemented, other than a computer adapted to implement the method according to the invention.

[0017] Furthermore, it can be appreciated that the claimed invention imposes no constraints on the method of acquiring the signal via the ultrasonic sensor. In this respect, the ultrasonic sensor can be any type of ultrasonic sensor, allowing the use of sensors commonly employed in non-destructive ultrasonic testing, typically via a single-element probe, maneuvered by automated movement to inspect all or part of the part.

[0018] Furthermore, the proposed approach allows for signal analysis compatible with a production context. Indeed, the present invention is quick and robust to implement, is compatible with real-time or near-real-time inspection of a part, allows for a high inspection rate, and is compatible with conventional acquisition equipment that is quick and easy to instrument. Moreover, the proposed approach can also be implemented during the manufacturing of the part, which allows its manufacture to be controlled and to be interrupted or corrected, if necessary, before it has finished manufacturing.

[0019] In addition to the characteristics just mentioned, the method according to the first aspect of the invention may have one or more complementary characteristics from among the following, considered individually or according to all technically possible combinations.

[0020] In one embodiment, obtaining the signal includes acquiring said signal via the ultrasonic sensor, the ultrasonic sensor performing the periodic displacement during the acquisition of said signal.

[0021] In one embodiment, the signal corresponds to a juxtaposition of several pluralities of time samples, each plurality of time samples being acquired at an acquisition position among a plurality of acquisition positions of the ultrasonic sensor, the signal comprising a first direction and a second direction orthogonal to the first direction, and wherein the background echo signal is obtained by: • Regression of the signal by a first regression function along the first direction, the first function being periodic; • Regression of the signal by a second regression function along the second direction.

[0022] In the context of the invention, "juxtaposition" means the side-by-side placement of pluralities of time samples to form a two-dimensional or three-dimensional signal, relative to the position of the sensor with respect to the part during the acquisition of each plurality of time samples. In other words, juxtaposition serves to form what is commonly called a B-scan.

[0023] In one embodiment, the second filtered signal is determined by: • Construction of a first vector comprising a concatenation in chronological order of acquisition of the temporal samples included in the first filtered signal; • Determination of a spectrum of the first vector; • Removal of the first periodic content of the first vector to form a second vector; • Determination of the second periodic content; • Removal of the second periodic content from the first filtered signal.

[0024] In the context of the invention, "concatenation" means the joining of pluralities of time samples to form a one-dimensional signal, such that the pluralities of time samples are ordered in chronological order of acquisition. In other words, concatenation is used to form what is commonly called an A-scan.

[0025] In one embodiment, the method further includes the removal of measurement noise in the second filtered signal, the measurement noise being generated by a device adapted to perform the periodic displacement of the ultrasonic sensor.

[0026] In one embodiment, the method further includes the removal of speck noise in the second filtered signal, the speck noise being generated by the bottom of the part.

[0027] In one embodiment, the indication is detected when an amplitude of the second filtered signal corresponding to said indication is greater than or equal to a predefined amplitude.

[0028] In one embodiment, the indication corresponding to a defect in the part is detected relative to a propagation time of an echo generated by said defect.

[0029] In one embodiment, a degree of criticality is determined for the detected indication, the degree of criticality being determined by comparison of the detected indication with a reference indication and / or according to the position of the detected indication in the part.

[0030] A second aspect of the invention relates to a computer program product comprising instructions which, when the program is executed on a computer, lead the latter to implement the steps of the method according to the invention.

[0031] A third aspect of the invention relates to a computer-readable recording medium comprising instructions which, when executed by a computer, lead the computer to implement the steps of the method according to the invention.

[0032] The invention and its various applications will be better understood by reading the following description and examining the accompanying figures. BRIEF DESCRIPTION OF THE FIGURES

[0033] The figures are presented for illustrative purposes only and are in no way limiting of the invention. • Fig. 1 is a synoptic diagram of a method according to one embodiment of the invention. • Fig. 2 is a schematic representation illustrating the movement of a sensor in certain embodiments of the invention. • Fig. 3 represents the variations in thickness of the part, obtained via the method according to Fig. 1, according to one embodiment. • Fig. 4 represents indications of defects in the part, obtained via the method according to Fig. 1, according to one embodiment. DETAILED DESCRIPTION

[0034] Unless otherwise specified, the same element appearing on different figures has a unique reference.

[0035] With the aim of improving ultrasonic testing of parts with a periodic structure, the present invention proposes a method for post-processing a signal acquired via an ultrasonic sensor in order to denoise this signal and thus better distinguish indications of defects in the part. As will be seen later, the method also makes it possible to determine certain properties of the part, including geometric and structural properties, without requiring the acquisition of additional signals.

[0036] A first aspect of the invention therefore relates to a computer-implemented method for inspecting a part from a signal acquired via an ultrasonic sensor. Method 100 and its various steps are described below with reference to [Fig. 1].

[0037] The signal in question corresponds to at least one ultrasonic wave propagating in the room. The wave is generated via any known emission technique.

[0038] In this respect, the sensor technology used is irrelevant for the proposed method 100. It may be a probe based on piezoelectric technology (such as a single-element sensor, a linear multi-element sensor or a multi-element matrix sensor), an electromagnetic acoustic sensor (in English “EM AT” for “Electromagnetic Acoustic Transducer”), a laser sensor, etc.

[0039] The wave propagating in the room and captured by the sensor can be generated by the same ultrasonic sensor, i.e. in "pulse-echo" mode. This wave can alternatively be generated via other techniques, such as using a second sensor for emission and capturing the wave after bouncing off the bottom of the room (in English, the so-called "pitch-catch" mode), or by transmission, i.e. by generating waves via a second sensor positioned opposite the sensor capturing the wave, on the other side of the room.

[0040] The ultrasonic sensor can be placed at a distance from or in contact with the part without changing the method. Similarly, the acquisition can be carried out by immersion or in contact, with or without a shoe, without having to modify the method.

[0041] The sensor is, for example, a single-element piezoelectric transducer, used in pulse-echo mode, whose center excitation frequency is between 100 kHz and 50 MHz, for example between 500 kHz and 10 MHz, or even between 1 MHz and 10 MHz. The sampling frequency of the ultrasonic wave signal backpropagated by the part and captured by the sensor can be between 100 MHz and 1 GHz, for example between 200 MHz and 700 MHz.

[0042] As illustrated in [Fig. 2], the ultrasonic sensor 20 periodically moves around ([Fig. 2](a)) or above (Figures 2(b) and 2(c)) the part 10 during signal acquisition. The sensor 20 is therefore in motion relative to the part during acquisition. In particular, the sensor 20 follows a trajectory 30 which is periodic. For example, in [Fig. 2](a), the trajectory is helical, while in [Fig. 2](b) the trajectory is crenellated and in [Fig. 2](c) the trajectory is triangular. Other forms of periodic trajectories can be used here without modifying the proposed method 100, for example, a periodic trajectory that is linear in segments, typically parallel, and following the same direction along the part.

[0043] The trajectory of movement is predefined, that is to say that it is determined prior to the implementation of method 100, by an operator or via an application, manually or automatically, for example according to the operator's business knowledge, that is to say the knowledge and / or experience that the operator has in the field of non-destructive ultrasonic testing.

[0044] The step size between two acquisition positions of the sensor 20 is between 0.1 mm and 5 mm, for example, between 0.5 mm and 2.5 mm. This step size is also predetermined before the implementation of method 100. The signal thus comprises several time samples acquired at a plurality of acquisition positions of the sensor 20 during its movement. In particular, a plurality of time samples is acquired at each position of the plurality of acquisition positions of the sensor 20. The signal therefore comprises several plurality of time samples, each plurality of samples being acquired at one of the acquisition positions of the plurality of acquisition positions.

[0045] The plurality of acquisition positions therefore corresponds to the set of acquisition positions taken by the sensor during its movement along the periodic trajectory 30. Two successive acquisition positions of the plurality of acquisition positions are thus separated by the predetermined step.

[0046] As mentioned above, the inspected part 10 has a periodic structure. For example, in relation to Figures 3 to 5 showing results obtained by implementing Method 100, the part is made of woven composite. For example, the part is a turbine blade or drive shaft of an aircraft turbomachine.

[0047] For the purposes of control, method 100 includes a step 120 of determining a background echo signal within the signal. This background echo signal corresponds to the echo from the bottom of the part. This signal is sensitive to the thickness of the part 10; that is, its content depends directly on the thickness of the part 10, particularly on variations in this thickness. Indeed, the thickness of the part 10 is not perfectly constant and may include more or less significant variations, thus altering the characteristics of the background echo signal (in terms of propagation time, amplitude, and / or phase).

[0048] Determining the background echo signal allows, on the one hand, for improving the signal-to-noise ratio in the signal, by removing this background echo signal from the signal and, on the other hand, to determine the variations in the thickness of part 10, which makes it possible to detect an anomaly in this thickness.

[0049] The background echo signal corresponds to the regression of a relative propagation time of the background echo in the signal. For example, the background echo signal is determined by regression of the propagation time corresponding to the background echo in the signal, relative to the position of the sensor or the input surface of the part (i.e., the surface of the part that is closest to the sensor at the time of acquisition). The background echo signal is therefore said regression of the signal. The regression can be performed by any known approach.

[0050] The determination of the relative propagation time of the background echo is, for example, carried out in the time domain, such as by evaluating the difference between the propagation time of the background echo and the propagation time of the input echo, typically when the sensor 20 is placed at a distance from the part 10 during acquisition. The input echo corresponds to the echo of the wave after backpropagation on the input surface of the part 10, that is to say, the part of the signal whose amplitude is substantially constant in the acquired signal and whose afferent propagation time corresponds to that of a round trip between the sensor and the first surface (called input surface) of the part 10 that is encountered by the emitted background.Alternatively, the difference is calculated between the propagation time of the background echo and the instant of emission of the ultrasonic wave, typically when the sensor 20 is placed in contact with the entrance surface of the part 10 during acquisition.

[0051] Method 100 also includes a step 130 for determining a first filtered signal. The first filtered signal is obtained by removing the background echo signal from the signal. This removal can be performed by any known approach, whether by simple subtraction or by a more sophisticated technique. Typically, the background echo signal is removed by subtraction from the signal, that is, by subtracting the regression calculated in the previous step. The determination of the first filtered signal is, for example, carried out in the time domain. Since the background echo signal corresponds to low-frequency content, this step serves to remove this low-frequency content from the signal.

[0052] By way of example, the subtraction is implemented by removing amplitudes from the signal whose propagation times correspond to the regression of the background echo signal. The removal of the background echo signal can also involve subtracting amplitudes from the signal whose propagation times fall within a given standard deviation of the regression of the background echo signal. The given standard deviation is, for example, between 5% and 20% of the mean value of the regression; for example, the standard deviation is equal to 10% of this mean value.

[0053] Method 100 also includes a step 140 of determining a second filtered signal. This second filtered signal is obtained by removing a first periodic content and a second periodic content from the first filtered signal.

[0054] The determination of the second filtered signal is, for example, carried out in the spectral domain.

[0055] The first periodic content corresponds to a portion of the first filtered signal, specifically a set of frequencies of the first filtered signal whose respective amplitudes depend on the periodicity of the periodic displacement of the ultrasonic sensor 20 during signal acquisition. Indeed, the periodicity of the trajectory 30 of the sensor 20 induces a periodic measurement artifact characterized by the predominance of certain frequencies, notably a fundamental frequency corresponding to the periodicity of the displacement of the sensor 20 and harmonics of this fundamental frequency. Removing these frequency components reduces the energy of the spectral components that are not related to a fault indication, and thus reveals any fault indications not otherwise visible.

[0056] The first periodic content therefore corresponds to a frequency content generated by the periodic displacement of the ultrasonic sensor during the acquisition of the signal.

[0057] The first periodic content is typically known at the time of inspection since the sensor's travel speed is controlled and predefined. Thus, this first periodic content is typically predefined when implementing Method 100. Alternatively, when the first periodic content is not predefined, it can be determined by any known technique. The first periodic content is, for example, determined in the spectral domain.

[0058] The second periodic content corresponds to a portion of the first filtered signal, specifically a set of frequencies from the first filtered signal whose respective amplitudes depend on the periodicity of the part's periodic structure. Indeed, the periodic structure itself generates, within the signal, a frequency content whose amplitude may exceed that of any potential defect indication, thus masking these indications. Removing this frequency content also improves the signal-to-noise ratio of the first filtered signal and reveals any defect indications that would otherwise be invisible.

[0059] The second periodic content therefore corresponds to a frequency content generated by the periodic structure of the piece.

[0060] The second periodic content can be determined by any known technique. The second periodic content is, for example, determined in the spectral domain.

[0061] This second periodic content also makes it possible to determine the periodic structure of the part, in particular the periodicity in question, for example a grain or fiber orientation of the structure. In the case of the composite part, this second periodic content allows, among other things, the characterization of the orientation of the composite fibers of the material.

[0062] The removal of the first and second periodic contents can be implemented by any known technique. For example, by applying a mask in the frequency domain, said mask corresponding to the frequency components of the first and / or second periodic contents. In other words, it involves subtracting the first and second periodic contents from the first filtered signal. Preferably, the second periodic content is removed from the first filtered signal after the first periodic content has been removed from the first filtered signal.

[0063] Method 100 also includes a step 170 of detecting an indication representative of a defect in the part. This indication is detected in the second filtered signal.

[0064] The indication is representative of a defect in the part in that it corresponds to said defect, that is to say that it corresponds to an echo generated by said defect, contained in the ultrasonic wave captured by the ultrasonic sensor 20.

[0065] The indication corresponding to a fault can be detected in the second filtered signal by any known technique. Said indication is, for example, detected in the time domain.

[0066] In one embodiment, method 100 also includes a step 110 of obtaining the signal acquired via the ultrasonic sensor 20. Obtaining said signal can be implemented by any known technique. Obtaining the signal can be implemented in real time, or quasi-real time, i.e., as the signal is acquired by the sensor 20, or it can be carried out a posteriori, once the acquisition is complete.

[0067] In one embodiment, the signal acquisition step 110 comprises acquiring the signal via the ultrasonic sensor 20. Throughout the acquisition, the ultrasonic sensor 20 is moved along the periodic trajectory 30.

[0068] In one embodiment, step 120 of determining the background echo signal includes a substep 120-1 of determining the propagation time of the background echo in the signal for each acquisition position of the plurality of acquisition positions (i.e., in the plurality of samples acquired and corresponding to said acquisition position among the plurality of acquisition positions). In particular, the propagation time of the background echo can be the propagation time corresponding to the maximum amplitude of the echo signal corresponding to the background echo, for each of the acquisition positions. The position of the sensor, the theoretical geometry of the part, and the propagation speed of a mechanical wave being predefined for each acquisition position, i.e., known during the inspection, the position of the echo corresponding to the bottom of the part (i.e., the surface of the part separated from the input surface by the thickness of the part) in the signal is easily identifiable for each acquisition position.

[0069] Step 120 may also include, where appropriate, a substep 120-2 of determining the propagation time relative to the input echo in the signal for each acquisition position of the plurality of acquisition positions, typically when the sensor 20 is not in contact with the part 10. Similar to substep 120-1, the propagation time relative to the input echo may be the propagation time corresponding to the maximum amplitude of the echo signal corresponding to the input echo, since the position of the sensor, the theoretical geometry of the part and the propagation speed of a mechanical wave are predefined for each acquisition position.

[0070] Step 120 may also include a substep 120-3 for determining the relative propagation time of the background echo for each acquisition position in the plurality of acquisition positions. The relative propagation time is determined by the difference between the propagation time of the background echo and the propagation time of the input echo (when the sensor 20 is positioned at a distance from the part 10), or between the propagation time of the background echo and the time of emission of the ultrasonic wave by the sensor 20 (when the sensor is positioned in contact with the part 10, on the input surface), for each acquisition position. The time of emission of the ultrasonic wave is predefined since it is known at the time of acquisition, for each acquisition position.

[0071] Step 120 may also include a substep 120-4 for constructing the background echo signal by regression of the relative propagation times of the background echo, each of the relative propagation times of the background echo being determined for one of the acquisition positions of the plurality of acquisition positions. In other words, this step 120-4 aims to produce a regression of the relative propagation times determined at the different acquisition positions where the sensor 20 is located.

[0072] In one embodiment, step 140 of determining the second filtered signal includes a substep 140-4 of removing the first periodic content, i.e., the harmonics related to the periodicity of the trajectory, in the first filtered signal.

[0073] Step 140 of determining the second filtered signal also includes a substep 140-5 of determining the second periodic content, i.e., the main periodic components of the structure of the part.

[0074] Step 140 of determining the second filtered signal also includes a substep 140-6 of removing the second periodic content from the first filtered signal, after the first periodic content has been removed.

[0075] In one embodiment, the removal of the first periodic content, in substep 140-4, is carried out by subtracting, according to a first strip width, the frequencies corresponding to the first periodic content. The first bandwidth is proportional to the fundamental frequency of the first periodic content. For example, the first bandwidth is between 1 / 1000 and 1 / 10 of the fundamental frequency of the first periodic content, or even between 5 / 1000 and 5 / 100 of the fundamental frequency of the first periodic content. Typically, the first bandwidth is equal to 1 / 100 of the fundamental frequency of the first periodic content.

[0076] In one embodiment, step 140 for determining the second filtered signal may also include a substep 140-3 for determining the first periodic content, if this first periodic content is not known. This substep 140-3 is implemented before step 140-4 for removing the first periodic content.

[0077] In one embodiment, the principal periodic components of the structure of part 10 of the second periodic content are determined in the first filtered signal, after removal of the first periodic content, by identifying spectral signatures in this first filtered signal by comparison with a plurality of reference signatures. The spectral signatures are then removed from the first filtered signal until a condition relating to the power of the principal periodic components of the structure of the part is satisfied.

[0078] Reference signatures are spectral signatures linked to particular periodic structures of reference parts (i.e., a set of harmonics that are excited by the periodic structure of these reference parts). The reference signatures are predetermined, that is, they are determined via any technique known prior to the implementation of method 100.

[0079] In particular, the determination of the second periodic content in substep 140-5 can be implemented by determining one or more spectral signatures for each acquisition position of the plurality of acquisition positions. Each spectral signature is determined by identifying, for each acquisition position, in the first filtered signal (after removal of the first periodic content), the fundamental frequency corresponding to each harmonic of said first filtered signal (after removal of the first periodic content), and then by comparing this fundamental frequency and its associated harmonic(s) in this first filtered signal with the reference signatures. A fundamental frequency and its associated frequency(ies) are thus considered to form a spectral signature when the comparison with the reference signatures reveals a strong similarity with one of the reference signatures.

[0080] The comparison is, for example, carried out by a calculated difference between the value of the frequency and / or amplitude of the fundamental frequency and its associated harmonic(s), and each reference signature of the plurality of signatures of A spectral signature is thus identified when the deviation is less than or equal to a first reference deviation. The identified spectral signature includes the fundamental frequency and its associated harmonic(s).

[0081] The first reference deviation is predetermined, that is to say determined prior to the implementation of method 100. The first reference deviation is, for example, expressed as a percentage relative to the reference signature considered, and is typically less than or equal to 10%, or less than or equal to 5%, or even less than or equal to 2%, typically equal to 2% or even to 1%.

[0082] The removal of the second periodic content can thus be implemented, in substep 140-6, by removing, for each acquisition position of the plurality of acquisition positions, the spectral signatures identified in substep 140-5 until the condition relating to the power of the principal periodic components of the part structure (i.e., the power of the identified spectral signatures) is satisfied. For example, the condition is that the sum of the powers of the unremoved spectral signatures be less than or equal to a first predefined power, typically proportional to the total power of the first filtered signal (after removal of the first periodic content).The first predefined power is typically between 1% and 20% of the total power of the first filtered signal (after removal of the first periodic content), for example equal to 15%, or equal to 10%, or even equal to 5%, or even equal to 2% or 1% of the total power of the first filtered signal (after removal of the first periodic content).

[0083] The first predefined power is determined, that is to say determined prior to the implementation of method 100.

[0084] The second periodic content thus corresponds to the set of identified spectral signatures to be removed to satisfy the condition relating to the power of the main periodic components of the structure of the part.

[0085] In one embodiment, the removal of the second periodic content in substep 140-6 is performed by subtracting, according to a second bandwidth, the frequencies corresponding to said second periodic content (i.e., corresponding to the spectral signatures identified and to be removed to satisfy the condition relating to the power of the principal periodic components of the part structure). The second bandwidth is proportional to the fundamental frequency of the second periodic content. For example, the second bandwidth is, for each identified spectral signature to be removed, between 1 / 10,000 and 1 / 100 of the fundamental frequency of said identified spectral signature to be removed, or even between 5 / 10,000 and 5 / 1,000 of the fundamental frequency of said identified spectral signature to be removed.Typically, the second bandwidth is equal to 1 / 1000 of the fundamental frequency of the identified spectral signature to be removed.

[0086] In one embodiment, method 100 also includes a step 160 of denoising the second filtered signal, before the implementation of the indication detection.

[0087] This denoising step aims to remove measurement noise and / or speckle noise that may be present in the second filtered signal. The denoising of the second filtered signal is, for example, performed in the time domain.

[0088] To this end, the denoising step 160 may include a substep 160-1 for determining and removing measurement noise. This measurement noise is generated in particular by the device used to move the sensor 20 along the trajectory 30.

[0089] Measurement noise can be determined and removed by any known technique.

[0090] Step 160 may also, or alternatively, include a substep 160-2 for determining and removing speckle noise. This noise corresponds to granular noise generated by interference from the coherent wavefronts of the echo from the bottom of the part. The bottom of the part is the surface of the part that is opposite the sensor 20 along the thickness of the part. This surface is also called the output surface, as opposed to the input surface, which is the surface of the part that is directly above the sensor 20 during acquisition.

[0091] The speck noise can be determined and removed by any known technique.

[0092] In one embodiment, the measurement noise can be determined, in substep 160-1, in a manner similar to the second periodic content in substep 140-5. That is, this measurement noise can be determined by identifying a correspondence between the fundamental frequency of each harmonic of the second filtered signal and reference measurement noise signatures, each of the reference measurement noise signatures corresponding to a known measurement noise source. Thus, when a calculated deviation between the fundamental frequency and its associated harmonic(s) is less than a second reference deviation, the calculated fundamental frequency and its associated harmonic(s) are considered to be a measurement noise source.The fundamental frequencies and their associated harmonic(s), identified as measurement noise sources, are then subtracted from the second filtered signal until a condition is met regarding the sum of the powers of the frequencies corresponding to these measurement noise sources. For example, the condition is that the sum of the powers of the fundamental frequencies and their associated harmonic(s) identified as measurement noise sources that have not been subtracted is less than or equal to a second predefined power, typically proportional to the total power of the second filtered signal. This second predefined power is typically between 1% and 20% of the total power. total power of the second filtered signal, for example equal to 15%, or equal to 10%, or even equal to 5%, or even equal to 2% or 1% of the total power of the second filtered signal.

[0093] The second predefined power is determined, i.e., determined prior to the implementation of method 100.

[0094] The second reference deviation is predetermined, that is to say determined prior to the implementation of method 100. The second reference deviation is, for example, expressed as a percentage relative to the reference measurement noise signature considered, and is less than or equal to 10%, or less than or equal to 5%, or even less than or equal to 2%, typically equal to 2% or even 1%.

[0095] The measurement noise thus corresponds to the set of fundamental frequencies and their associated harmonic(s) identified as measurement noise sources to be removed to satisfy the condition relating to the power of these measurement noise sources.

[0096] In one embodiment, at step 170, an indication corresponding to a fault is detected when the amplitude of the second filtered signal that corresponds to the indication in question is greater than or equal to a predefined amplitude.

[0097] The predefined amplitude is a predetermined amplitude value, that is to say defined before the implementation of step 170, by the operator or an application, whether manually or automatically, for example based on business knowledge.

[0098] When the amplitude of the indication is greater than or equal to the predefined amplitude, then the indication corresponds to a defect in the part, otherwise the indication does not correspond to a defect.

[0099] In one embodiment, the indication corresponding to a fault is, in addition, detected as a function of an echo generated by said fault, in particular relative to a propagation time of said echo.

[0100] Indeed, knowing the propagation time of the echo of the supposed defect, it is possible to determine a spatial extent of the supposed defect for this given propagation time, and therefore to discriminate whether the echo is indeed that of an indication corresponding to a defect, or not.

[0101] In other words, the spatial extent corresponds to the set of samples in the second filtered signal that are associated with the same intermediate echo, that is to say that these samples are associated with the same propagation time (or propagation times close to each other).

[0102] The term "intermediate echo" means an echo in the second filtered signal which does not correspond to the echo from the bottom of the room 10 or to the echo from the input surface but which corresponds to a supposed defect.

[0103] These samples therefore originate from a hypothetical defect located at a given depth, or extending continuously into the depths. The spatial extent of the hypothetical defect is thus characterized by a continuous distribution of its associated propagation time. Therefore, there is no discontinuity in the spatial extent, from the point of view of the propagation time of the associated intermediate echo.

[0104] Thus, when the spatial extent (typically given in units of length, area, and / or volume) is significant, the corresponding echo in the second filtered signal does indeed indicate a defect. Conversely, when this spatial extent is not significant, the corresponding echo in the second filtered signal does not indicate a defect. In particular, the spatial extent of the alleged defect can be compared to a reference, for example, a reference spatial extent (typically given in units of length, area, and / or volume, depending on the dimension that characterizes the spatial extent of the echo of the alleged defect). When the spatial extent corresponding to the echo of the alleged defect is greater than or equal to the reference spatial extent, the indication is determined to correspond to a defect in the part; otherwise, the indication is determined not to correspond to a defect.

[0105] To this end, step 170 of detecting an indication corresponding to a defect may include a substep 170-1 of determining the spatial extent of the echo of the supposed defect. As described above, the spatial extent corresponds to the set of signal samples that correspond to said echo and that form a propagation time continuum. The spatial extent is then characterized by a length (for example, its greatest length), an area (for example, a cross-section at a given propagation time of the spatial extent or a projection of said spatial extent onto a plane), and / or its volume.

[0106] Step 170 of detecting an indication corresponding to a defect may then include a substep 170-2 of identifying the defect indication. Identification is performed by comparing the spatial extent with the reference spatial extent (i.e., by comparing the characteristic dimension of the spatial extent with the corresponding characteristic dimension of the reference spatial extent). Thus, when the characteristic dimension of the spatial extent is greater than or equal to the corresponding characteristic dimension of the reference spatial extent, the indication is determined to correspond to a defect in the part. Otherwise, when the characteristic dimension of the spatial extent is less than the corresponding characteristic dimension of the reference spatial extent, the indication is determined not to correspond to a defect.

[0107] An illustration of this embodiment is shown in [Fig. 4]. In [Fig. 4], (a) a portion of the second filtered signal is shown, in this case in the form of a two-dimensional representation of the sum, according to the thickness of the part, of the amplitudes of the second filtered signal. In order to improve the contrast in the image, a thresholding was applied to this two-dimensional representation and the echoes of the supposed defects are displayed in white, as shown in [Fig.4] (b).

[0108] In [Fig. 4] (c), a color scale represents, for each assumed defect echo (i.e., in white in [Fig. 4] (b)), the propagation time corresponding to each sample of said echo of the assumed defect. This color scale allows us to assess the spatial extent of each assumed defect and thus evaluate whether it is an indication of a defect or not.

[0109] On this [Fig.4] (c) three indications corresponding to two defects D are thus identified.

[0110] In one embodiment, method 100 also includes a step 180 for determining the criticality level of the detected indication. The criticality level can be determined by any known technique. For example, the criticality level can be determined by comparing the detected indication with a reference indication or by comparing the surface area of ​​the detected indication to a predefined threshold.

[0111] The reference indication is, for example, a predetermined indication, which corresponds to a known defect in a part that has already been inspected or simulated. The comparison then allows for the evaluation of a distance between the detected indication and the predetermined indication. When this distance is less than a predefined distance, the detected indication is assigned a criticality level of "critical". Conversely, when the distance is greater than or equal to the predefined distance, the detected indication is assigned a criticality level of "non-critical".

[0112] The predefined distance is, for example, determined by the operator or an application, manually or automatically, for example based on the operator's business knowledge.

[0113] The distance can be any known form of distance, for example a maximum, average or minimum value of a difference, a maximum, average or minimum root mean square error, etc.

[0114] The predefined threshold is, for example, determined by the operator or an application, manually or automatically, for example based on the operator's business knowledge. Typically, when the area of ​​the detected indication is greater than or equal to the predefined threshold, then the detected indication is assigned a criticality level of "critical". Conversely, when the area of ​​the detected indication is less than the predefined threshold, then the detected indication is assigned a criticality level of "non-critical".

[0115] In one embodiment, the degree of criticality, at step 180, can alternatively, or in addition, be determined according to the position in the part of the indication detected. Indeed, the same defect may not have the same criticality depending on whether it is more or less deep in the part, near a more or less fragile section of the part, near an edge or a connection element, etc. Since the position in part 10 of each echo corresponding to a defect indication is known, it is possible to evaluate, in relation to the specifications relating to said part 10, the degree of criticality of the detected indication.

[0116] In one embodiment, method 100 includes step 110 of obtaining the acquired signal. The signal is, for example, received directly from the sensor 20 or from a device used to acquire the signal.

[0117] The received signal can be one-dimensional, then called a 1D signal or A-scan. This A-scan commonly corresponds to the chronological concatenation (i.e., the joining together) of each plurality of samples, each being acquired for one of the acquisition positions of the sensor 20 during its movement along the trajectory 30.

[0118] In this case, the method 100 also includes a step 115 for forming a two- or three-dimensional signal, called a 2D or 3D signal, or B-scan. This B-scan commonly corresponds to the juxtaposition of each plurality of samples such that two plurality of samples acquired at contiguous acquisition positions are side by side in the B-scan. In fact, each plurality of samples is positioned in the B-scan so as to correspond to the acquisition position of the sensor 20 with respect to the part 10. Thus, depending on the path taken by the sensor 20, the B-scan can be two-dimensional or three-dimensional.

[0119] The received signal may, alternatively, be the B-scan. In this case, it is not necessary to implement step 115 in order to form this B-scan.

[0120] In one embodiment, the regression is implemented on the B-scan, i.e., on the signal in B-scan format. This signal comprises a first direction and a second direction. The second direction is orthogonal to the first direction. The first direction corresponds, for example, to the repetition direction DI of the periodic pattern along which the trajectory is formed, as illustrated in [Fig. 2]. The second direction is also orthogonal to the direction of the part thickness.

[0121] The regression may be the same in the first direction and in the second direction, or it may be different. For example, the regression may be carried out by a first regression function in the first direction and by a second regression function in the second direction.

[0122] In one exemplary embodiment, the first regression function is a periodic function, such as a cosine or sine function, or a sum of cosine and / or sine functions, and the second regression function is a lower-order function or equal to 5 (for example, a polynomial of order less than or equal to 5), typically of order equal to 3 or 4 (such as a polynomial of order 3 or 4).

[0123] In one embodiment, step 140 for determining the second filtered signal may also include a substep 140-1 for constructing a first vector. This substep 140-1 is implemented, where applicable, before substep 140-3 for determining the first periodic content, when the latter is implemented, and before substep 140-4 for removing the first periodic content. This first vector comprises a concatenation of the time samples of the first filtered signal, that is, the time samples of the spatial distribution after removal of the background echo signal, ordered in chronological order of acquisition. The aim here is therefore to form a new 1D signal from the first filtered signal. The advantage of having a long signal duration is to perform a Fourier Transform applied to said first vector, thereby improving the resolution of said spectrum.

[0124] Step 140 for determining the second filtered signal may also include a substep 140-2 for determining a spectrum of the first vector, implemented after substep 140-1 for constructing the first vector. This spectrum is obtained by applying a Fourier Transform to the first vector, that is, to the time samples contained in the first vector.

[0125] Step 140, for determining the second filtered signal, may then include substep 140-3, for determining the first periodic content, when necessary. The first periodic content is determined in the first vector.

[0126] Step 140 of determining the second filtered signal then includes substep 140-4 of removing the first periodic content. The first periodic content is then removed from the first vector. This removal makes it possible to form a second vector corresponding to the first vector from which the first periodic content has been removed.

[0127] Step 140 of determining the second filtered signal then includes substep 140-5 of determining the second periodic content. This second periodic content is determined in the second vector.

[0128] Step 140 of determining the second filtered signal then includes substep 140-6 of removing the second periodic content. The second periodic content is then removed from the second vector. This removal allows the formation of the second filtered signal, corresponding to the second vector from which the second periodic content has been removed.

[0129] In one embodiment, method 100 also includes a step 150 of reconstructing a spatial distribution of the second filtered signal, called filtered distribution, before applying the possible denoising of step 160 and before performing the detection of the indication relating to a defect in the part, in step 170.

[0130] To this end, step 150 includes a substep 150-1 of applying an Inverse Fourier Transform to the second filtered signal, obtained after the removal of the second periodic content in step 140-6.

[0131] Step 150 also includes a substep 150-2 for assigning each time sample, that is, the signal amplitude value associated with that sample, of the second filtered signal to a position in the room. This assignment is implemented analogously to the construction of the spatial distribution in step 115 and allows the filtered distribution in question to be formed. The filtered distribution therefore corresponds to the same section of the room as the spatial distribution.

[0132] In such an embodiment, the denoising of step 160 is applied to the filtered distribution, i.e. after the assignment of the time samples of the second filtered signal has been carried out.

[0133] Similarly, in such an embodiment, the detection of the indication corresponding to a fault is applied to the filtered distribution, where appropriate after its denoising via the denoising step 160.

[0134] In one embodiment, method 100 also includes a step 125 for detecting a geometry defect. The detection is performed using the backscatter echo signal, in particular based on the propagation time of the backscatter echo, and makes it possible to determine a defect in the thickness of the part. Such detection makes it possible, for example, to rebalance the mass of the part, such as by removing or adding mass, in order, in particular, to compensate for geometry anomalies. The rebalancing can advantageously also be performed based on one or more defects, for example, delamination, detected during step 170.

[0135] By way of example, as illustrated in [Fig. 3], plotting the regression function allows us to estimate the variations in the thickness of the part, in this example a tube. [Fig. 3] (a) shows a flattened view of the variation in the propagation time of the back echo of the part in microseconds. [Fig. 3] (b) shows the average angular profile of the tube thickness. [Fig. 3] (c) shows the average longitudinal profile of the tube. [Fig. 3] (d) is an angular view in polar representation of the angular profile of [Fig. 3] (c). It can clearly be observed that the thickness of the part is not constant in either its angular or longitudinal profile.

[0136] Analysis of the regression function makes it possible to detect and, if necessary, characterize a defect in the geometry of the part. This is achieved in particular by comparing the regression function to the reference geometry of the part. The reference geometry is, for example, the theoretical geometry of the part as designed using a computer-aided design tool, or the geometry measured during a previous inspection of the part, for example, immediately after its manufacture.

[0137] In particular, by comparing the part thickness estimated via the backscatter echo signal with the reference geometry, it is possible to determine a deviation between the regression and the reference geometry. When the deviation is greater than a predefined error, then the part geometry is said to be "defective." Conversely, when this deviation is less than the predefined error, then the part geometry is said to be "non-defective."

[0138] The predefined error is predetermined by the operator or an application, manually or automatically, for example on the basis of the operator's business knowledge.

[0139] The deviation can be any known form of deviation, for example a maximum, mean or minimum value of a difference, a maximum, mean or minimum squared deviation, etc.

[0140] In one embodiment, method 100 also includes a step 190 for characterizing the periodic structure of the part. Indeed, the second periodic content differs depending, in particular, on the orientation of the fibers and / or grains of the periodic structure of the part. By comparing this second periodic content to a reference periodic content, it is possible to characterize the orientation of the fibers and / or grains of the material of the part.

[0141] The reference periodic content is, for example, predetermined numerically, via a suitable simulation tool, and / or via prior experimentation on a part whose periodic structure is known, allowing this reference periodic content to be determined. The periodic content is thus associated with the periodic structure of the part used to determine it, otherwise called the reference structure.

[0142] In particular, it is possible to calculate a difference between the second periodic content and the reference periodic content. When this difference is less than or equal to a predetermined value, the second periodic content corresponds to the reference periodic content and the structure of the part corresponds to the reference structure. Conversely, when this difference is greater than or equal to the predetermined value, the second periodic content does not correspond to the reference periodic content and the structure of the part does not correspond to the reference structure.

[0143] The predetermined value is, for example, by the operator or an application, whether manually or automatically, for example based on business knowledge.

[0144] The difference can be any known form of difference, for example a maximum, mean or minimum value of a difference, a maximum, mean or minimum root mean square deviation, etc.

[0145] Another aspect of the invention relates to a device configured to implement Method 100. In particular, the device may include volatile or non-volatile memory and a processor. The memory may include instructions which, when executed by the processor, cause the processor to implement Method 100.

[0146] The device may also include a receiving module adapted to receive the signal acquired by the sensor. The connection module is, for example, a connection module for a wired or wireless network, a connection interface, for example, for connecting a storage medium to the device, the storage module containing the signal in question in its memory, or any other module enabling the acquisition of said signal. In the case of a connection module adapted to connect the device to a network, the signal is obtained via this network, for example, after transmission by a computer on which the signal is stored, for example, the acquisition system used for its acquisition.

[0147] It is possible to obtain the signal directly from the sensor, in which case the device can be included in the acquisition system, which is used to maneuver the ultrasonic sensor 20 and to acquire the signal.

Claims

Demands

1. A computer-implemented method (100) for inspecting a part (10) from a signal acquired via an ultrasonic sensor, the signal corresponding to at least one ultrasonic wave propagating in the part (10), the ultrasonic sensor (20) undergoing periodic displacement during signal acquisition, the part (10) comprising a periodic structure, the method (100) comprising: - Determine (120) a signal of an echo from a background of the room (10) by a regression of a propagation time corresponding to the background echo in the signal; - Determine (130) a first filtered signal of the signal, by removing the background echo signal from the signal; - Determine (140) a second filtered signal from the signal by removing a first periodic content and a second periodic content, the first periodic content corresponding to a frequency content generated by the periodic displacement of the ultrasonic sensor (20) during the acquisition of the signal, the second periodic content corresponding to a frequency content generated by the periodic structure of the part (10); - Detect (170), in the second filtered signal, an indication corresponding to a defect in the part (10).

2. Method (100) according to the preceding claim, wherein the signal corresponds to a juxtaposition of several pluralities of time samples, each plurality of time samples being acquired at an acquisition position among a plurality of acquisition positions of the ultrasonic sensor (20), the signal comprising a first direction and a second direction orthogonal to the first direction, and wherein the background echo signal is obtained by: - Regression of the signal by a first regression function along the first direction, the first function being periodic; - Regression of the signal by a second regression function along the second direction.

3. Method (100) according to any one of the preceding claims, wherein the second filtered signal is determined by: - ​​Construction (140-1) of a first vector comprising a concatenation in chronological order of acquisition of the time samples included in the first filtered signal; - Determination (140-2) of a spectrum of the first vector; - Removal (140-4) of the first periodic content of the first vector to form a second vector; - Determination (140-5) of the second periodic content; - Removal (140-6) of the second periodic content of the first filtered signal.

4. Method (100) according to any one of the preceding claims, further comprising the removal (160-1) of measurement noise in the second filtered signal, the measurement noise being generated by a device adapted to perform the periodic displacement of the ultrasonic sensor (20).

5. Method (100) according to any one of the preceding claims, further comprising the removal (160-2) of a speck noise in the second filtered signal, the speck noise being generated by the bottom of the part (10).

6. Method (100) according to any one of the preceding claims, wherein the indication is detected when an amplitude of the second filtered signal corresponding to said indication is greater than or equal to a predefined amplitude.

7. Method (100) according to any one of the preceding claims, wherein the indication corresponding to a defect in the part (10) is detected relative to a propagation time of an echo generated by said defect.

8. Method (100) according to any one of the preceding claims, wherein a degree of criticality is determined (180) for the detected indication, the degree of criticality being determined by comparison of the detected indication with a reference indication and / or according to the position of the detected indication in the part (10).

9. Computer program product comprising instructions which, when the program is executed on a computer, cause that

10. ci to implement the steps of method (100) according to one of the preceding claims. Computer-readable recording medium comprising instructions which, when executed by a computer, cause the computer to carry out the steps of method (100) according to any one of claims 1 to 8.

Images