Method for detecting ultrasonically defects of air-pocket type in a part

Abstract

One aspect of the invention relates to a method for detecting ultrasonically defects of air-pocket type in a part to be inspected, comprising: - receiving an ultrasonic map of the part to be inspected; - receiving a thickness map of at least one segment of the part to be inspected; - determining, based on the thickness map, a simulated ultrasonic map of a part free of defects of air-pocket type having the same composition and the same thickness map as the part to be inspected; - determining, based on the ultrasonic map of the part to be inspected and on the simulated ultrasonic map, a detection map of the part to be inspected; - determining, based on the detection map, whether or not a defect of air-pocket type is present in the part to be inspected.

Description

DESCRIPTION TITLE: Ultrasonic method for detecting air zone defects in a room TECHNICAL FIELD OF THE INVENTION

[0001] The technical field of the invention is that of the non-destructive testing of a part by ultrasound, and in particular by analysis of the attenuation of ultrasonic waves in the part to be tested.

[0002] The invention thus relates to an ultrasonic method for detecting defects using thickness data from at least a part of the part. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0003] In industry, some parts are manufactured from digital models. However, during the manufacturing process, there may be differences in structure or composition between the digital model (which represents the "ideal" part one wishes to produce) and the part actually manufactured. For example, for parts made of woven composite material, there may be a difference between the weave of the manufactured part and the theoretical weave supplied to the loom. These differences are not necessarily problematic and do not always need to be detected.

[0004] The manufactured part may also have defects such as cracks, fissures, delamination, or delamination (in the case of bonded materials), which result in the presence of "air pockets" within the part. Such defects alter the mechanical properties of the part and weaken it, and must therefore be detected.

[0005] To detect "air zone" defects, ultrasonic testing methods are commonly used. In ultrasonic testing, ultrasonic waves are emitted by a transducer placed on the surface of the part being inspected and propagate through it. When an ultrasonic wave encounters an interface separating two areas with different acoustic impedances within the part (particularly when the ultrasonic wave encounters a defect), it is detected. inside the room), part of the ultrasonic wave is reflected and the unreflected part of the wave is attenuated.

[0006] Ultrasonic testing methods are classically divided into two categories: reflection methods and transmission methods (also called attenuation methods). In reflection methods, the same transducer both emits ultrasonic waves and receives reflected waves (i.e., "echoes" or "return signals"). From the intensities and arrival times of these return signals, information about the presence and location of defects in the part can then be extracted. In transmission methods, a receiver, separate from the transmitter and placed on a different surface of the part (for example, a surface opposite the one on which the transducer emitting the ultrasonic waves is located), receives the attenuated ultrasonic waves.Defect detection is performed based on the quantity of waves that have reached the second surface after passing through the part (only once, therefore, as opposed to twice - there and back - for reflection methods).

[0007] Because air zone defects strongly attenuate ultrasonic waves, transmission methods are well suited for detecting such defects.

[0008] For example, such methods are known to be used for bonded parts, such as aeronautical components comprising a composite material section to which a metal reinforcement is bonded. An example of such a part is the LEAP engine's composite fan blade, to which a titanium reinforcement is bonded. In the bonded areas, there may be anomalies such as porosity or lack of adhesive, which are air gap defects. Ultrasonic testing is typically performed on the manufactured part to verify the absence of such defects.

[0009] The 3D weave of the composite material strongly disperses ultrasonic waves. Furthermore, the resin within the composite material significantly attenuates ultrasonic waves. For these reasons, a reflection method is very difficult to implement, and an attenuation method is preferable.

[0010] The result of the ultrasonic testing can be represented using a C-type visualization, known as a "C-scan," which is a map linking the amplitude of the The signal is transmitted through the part to an inspection position for the surface of the controlled area. The C-scan therefore provides a plan and top view of the part.

[0011] Defects are not always easily visible in the C-scan of the manufactured part, and C-scan processing is generally implemented to highlight them. Specifically, a reference part can be used. This reference part is of the same type as the part being inspected (for example, a composite blade with a bonded titanium leading edge and predefined dimensions) and is known to be free of air gap defects. A C-scan map of this reference part is then generated and used as the reference map for the part.

[0012] For each part to be inspected, a C-scan map is determined and subtracted from the reference map. The result of this subtraction is a "processed" map on which the defects appear, along with other image patterns. The image is then thresholded to remove patterns other than the defects, so that only the defects remain visible in the thresholded image. The threshold used for image thresholding is a predefined threshold, which can optionally be readjusted as described below.

[0013] Defect detection according to the method described above is shown in Figures 1a–1e. In particular, Figures 1a–1e depict a part 101 to be inspected. In this example, the part 101 shown in Figure 1a is a fan blade made of composite material with a titanium leading edge 101b bonded to a body 101a made of composite material. Figure 1b shows a C-scan map 102 of the part 101. This map is typically obtained using an attenuation-based ultrasonic testing method. The C-scan map 102 comprises several zones 102a, 102b, 102c, and 102d, which correspond to different ultrasonic attenuation values.

[0014] Figure 1c represents a C-scan 103 map of a reference part. The reference part corresponds to a part of the same type as the part to be inspected 101 (for example, manufactured from the same digital model, and therefore theoretically possessing the same dimensions, structure, and components), but which is known not to have defects of the type that we are trying to detect in the part to be inspected 101. In other words, the part of The reference part is an "ideal" and defect-free component. In the example shown in Figures 1a-1e, the reference part is an assembled blade consisting of a perfectly sound, pre-machined composite body and a bonded titanium leading edge with a geometry as close as possible to the nominal geometry. The bonding process was previously controlled and inspected to ensure it was free of defects.

[0015] Similar to the C-scan 102 map of the part to be inspected, the C-scan 103 map of the reference part has several zones 103a, 103b, 103c, 103d which correspond to different ultrasonic attenuation values.

[0016] Figure 1d shows a processed map 104 obtained by subtracting the C-scan map 103 of the reference part from the C-scan map 102 of the part to be inspected. As shown in Figure 1d, map 104 exhibits an overall narrower range of variation in ultrasonic attenuation values ​​than C-scan maps 102 and 103. However, the processed map 104 includes areas 104a, 104b, and 104c, referred to as "ultrasonic indications," which correspond to peak attenuation areas and may represent air-type defects. To identify such defects, a threshold can then be applied to the processed map 104 to obtain the thresholded map 105 shown in Figure 1e. In other words, a threshold is applied to the attenuation values ​​of the processed map 104. In particular, this threshold can be defined as a percentage of the maximum amplitude of the received signal (i.e.Amplitudes below this percentage of the maximum amplitude are "cut off." For example, the threshold can be a predefined value, corresponding, for instance, to a signal attenuation of -6 dB. In other words, differences between the C-scan map 102 of the part to be inspected and the C-scan map 103 of the reference part are considered air zone defects (specifically, bonding defects) when they correspond to an attenuation greater than -6 dB. As described below, this threshold can be determined in a detailed step. In the example of the thresholded map 105, a defect 105a appears.

[0017] Subtracting a C-scan 103 map of a reference part from the C-scan 102 map of the part to be checked therefore makes it possible to overcome the complex geometry and the difference in thickness of the part, and thus a better detectability of defects such as air zones.

[0018] The threshold to be applied to the processed mapping 104 can be determined or adjusted in a further step, using a so-called "defect part." This part is of the same type as the part to be inspected, but it contains "controlled" defects, meaning that the location, dimensions, and type of defect are known for each defect in the defect part. The defect part is typically obtained from a reference part (i.e., without defects) into which elements have been inserted that behave similarly to air-zone defects. For example, Teflon elements can be inserted, knowing that a Teflon element behaves in a way that is generally similar to that of an air bubble with respect to ultrasonic signals (the signals are attenuated in approximately the same way when passing through a Teflon element and when passing through an air bubble or an air gap).This defect map allows you to define the threshold to be applied, for example in decibels (dB), according to tolerances relative to the actual size of the defects. In other words, since the positions of the defects in the defect map are known, it is possible to adjust the detection threshold to obtain the best compromise between "false positives" (detections on the threshold map that do not correspond to a defect that you are trying to control) and "false negatives" (no element appears on the threshold map even though there is a defect).

[0019] A flowchart of the method for detecting defects in a part according to the prior art described above is shown in Figure 2.

[0020] In step 210, a C-scan map of a defect-free reference part (C-scan map 103 in Figures 1a-1e), also called a "gain correction vane," is obtained. In step 220, a C-scan map of the part to be inspected (C-scan map 102 in Figures 1a-1e) is obtained. The C-scan map of the (defect-free) reference part is subtracted (step 230) from the C-scan map of the part to be inspected to obtain a processed map (map 104 in Figures 1a-1e).

[0021] In step 240, a C-scan map of the defective part (the defects being known, as detailed above) is obtained, and used to determine an optimal threshold (step 250) for detecting the defects, as described above.

[0022] In step 260, the threshold determined in step 250 can be applied to the processed map obtained in step 230. The defect(s) can then be identified in the thresholded map in step 270. Indeed, the deviations between the part to be checked and the defect-free reference part, which are "ultrasonic indications", should theoretically only be bonding anomalies, i.e. air zone type defects.

[0023] The C-scan mapping of the reference part, the C-scan mapping of the part to be inspected, and the C-scan mapping of the defective part are typically obtained by a transmission ultrasonic testing method.

[0024] It is noted that steps 210, 220 and 240 can be implemented in parallel or in any order (although generally, steps 210 and 240 on the reference part and the defective part are implemented before carrying out the actual inspection of a part, therefore before step 220).

[0025] In the prior art process described above, the role of the defect-free reference part is to obtain the most homogeneous processed C-scan map possible, which is free from the attenuation variabilities induced by the material(s) used and the thicknesses of the part (in particular the composite material and / or the thicknesses of the different materials such as titanium in the previous example).

[0026] However, the defects we seek to detect are not the only ones that generate ultrasonic indications, i.e., areas of ultrasonic attenuation peaks. Other factors can generate ultrasonic indications, notably the variability in the structure or thickness of the part (in the case of a composite blade with a bonded titanium leading edge, for example, the variability in the leading edge geometry and / or the variability in the woven weave generate such ultrasonic indications). These unanticipated variations in the implementation of the inspection can arise, for example, from the diversity of suppliers for the same product, changes in geometry, the manufacturing process, and / or the broadening of the geometric acceptance criteria for the part (or intermediate parts, i.e., the leading edge and the pre-machined body in the previous example).These variabilities give rise to false positives (false defects) detected on the mapping after thresholding.

[0027] One solution would be to adapt the threshold to the different parts of the component being inspected, but such a solution is difficult to implement because it requires first determining the areas where attenuation variability is acceptable and establishing the associated acceptability threshold. Furthermore, this solution carries the risk of increasing the number of false negatives, that is, the number of undetected defects (particularly because they are located in a part of the component where greater attenuation variability was deemed acceptable).

[0028] There is therefore a need to improve the detection of air zone defects in parts and to eliminate irrelevant ultrasonic indications (i.e., areas that appear on the thresholded map as defects when they are not). SUMMARY OF THE INVENTION

[0029] The invention offers a solution to the problems mentioned above by using thickness data from the manufactured part to be inspected to construct a simulated ultrasonic map of a virtual part, and using this simulated ultrasonic map as a reference map (instead of the ultrasonic map of a defect-free reference part, as in the prior art method described above). The use of such a simulated ultrasonic map reduces irrelevant ultrasonic readings, particularly those resulting from variations in the geometry of the manufactured part compared to the digital model on which it is based.

[0030] One aspect of the invention relates to a computer-implemented method for detecting air zone defects in a part to be inspected using ultrasound. The method comprises:

[0031] - receive an ultrasonic map of the part to be inspected;

[0032] - receive a thickness map of at least a portion of the part to be checked;

[0033] - determine, from the thickness map of at least one portion of the part to be checked, a simulated ultrasonic map of at least one portion of a part without air zone type defects having the same composition and the same thickness map as at least one portion of the part to be checked;

[0034] - determine, from the ultrasonic mapping of the part to be inspected and the simulated ultrasonic mapping, a detection map of the part to be inspected;

[0035] - determine, by applying a predefined threshold to the detection map of the part to be checked, the presence or absence of an air zone type defect in the part to be checked.

[0036] An "air zone defect" is defined as a defect in the room whose behavior with respect to ultrasound (specifically, the ultrasonic attenuation coefficient) is similar to that of an air zone present in the room. When ultrasonic waves encounter such a defect, they are typically attenuated more than when they pass through the rest of the room. Examples of such a defect include an air bubble or a crack.

[0037] The term "part to be inspected" refers to the mechanical part in which the possible presence of a defect is sought. It should be noted that the procedure can be applied to a complete part or to a portion of a part (rather than a complete part), both referred to as "part" or "part to be inspected" hereafter, for the sake of simplicity.

[0038] An "ultrasonic map" is understood to be a graphical representation of the ultrasonic behavior at different points within a part. For example, an ultrasonic map can be a 2D image projecting the part onto a predefined plane, representing the amplitudes of ultrasonic signals at various points on the projection after they have passed through the part. Specifically, each point on the map can be associated with data relating to the ultrasonic behavior of the part along an axis passing through that point in a direction orthogonal to the projection plane. For example, this data could be the amplitude value of the signal exiting the part, which reflects the attenuation experienced by the ultrasonic wave as it passes through the part. Since air pockets attenuate the ultrasonic signals passing through them, this attenuation is visible in the ultrasonic map of the part.

[0039] The "ultrasonic mapping of the part to be inspected" is therefore a mapping obtained on the complete part (or portion of part), after its manufacture.

[0040] A "thickness map" is understood to be a graphical representation of the part's thickness along a given direction. Typically, a thickness map represents a projection of the part onto the same plane as the ultrasonic map, where each point on the projection is associated with a value representing the part's thickness at that point, along the direction orthogonal to the plane in question. For example, the thickness map can be a 2D image comprising a plurality of pixels, each pixel having a value calculated based on the part's thickness at a point corresponding to that pixel.

[0041] Depending on the method, the thickness map obtained is either a thickness map of the entire part being inspected or a thickness map of only a portion of the part. For example, when the part being inspected is an engine blade comprising a composite body and a titanium leading edge bonded together, the thickness map obtained may be a map of the leading edge only. Conversely, the ultrasonic map obtained is an ultrasonic map of the entire part.

[0042] By "simulated ultrasonic mapping" of the part or portion of the part, we mean a mapping that "simulates" an ultrasonic mapping of the part or portion of the part, but which was not established (at least in part) using an ultrasonic method applied to the part or portion of the part being inspected. More precisely, the simulated ultrasonic mapping is determined here from the thickness mapping. In other words, the thickness mapping is "transformed" into an equivalent ultrasonic mapping, which would theoretically be obtained if an ultrasonic method were applied to the part or portion of the part to study its ultrasonic behavior and deduce an ultrasonic map, assuming that this part or portion of the part does not contain any defects.

[0043] Simulated ultrasonic mapping corresponds to the ultrasonic mapping of a "virtual" part (or a portion of a virtual part) that would be similar to the part (or portion of a part) under consideration, and which would not have any air gap defects. The underlying idea is that, in the absence of defects, the attenuation of ultrasonic waves depends on an average attenuation coefficient of the medium through which they pass and the length traveled by the ultrasonic signals (and therefore the thickness of the part). Using attenuation data for parts By considering the same medium (which may be heterogeneous, for example, several media successively traversed) and the actual thickness data of the part (i.e., the true thickness of the manufactured part, incorporating its possible variations relative to the digital model according to which it was manufactured), it is possible to deduce a simulated ultrasonic map of a part with the same thickness profile, but which would have no defects. The simulated ultrasonic map thus provides a reference map of the ultrasonic behavior in the part in the absence of defects.

[0044] In particular, the thickness mapping may include a plurality of pixels, each pixel of the plurality of pixels being associated with a respective value, and the determination of the simulated ultrasonic mapping may include a multiplication of the value of each pixel of the plurality of pixels by a respective coefficient.

[0045] The term "composition" refers to a set of parameters defining the structure and the material(s) constituting the part (or portion of a part). For example, for a part made of woven composite material, the composition includes the weave type, the type of warp and weft yarns, the warp / weft ratio, etc. Therefore, a "defect-free part with the same composition and thickness mapping" as the part to be inspected is understood to be a virtual part that is "identical" in terms of composition and dimensions to the part to be inspected, but which is assumed to be free of air zone defects.

[0046] The term "detection map" refers to a representation of a projection of the part on which any potential defects are determined. In this process, this detection map is obtained from the (actual) ultrasonic map of the manufactured part and the simulated ultrasonic map (which serves as a reference).

[0047] The above method advantageously utilizes the actual thickness data of the part being inspected to construct a (simulated) ultrasonic map that serves as a reference for determining the presence or absence of air zone defects. The method is thus more precise than the traditional prior art method described above, in which the reference map is derived from a "Ideal" (physical) part, which may have differences in thickness compared to the part being checked.

[0048] In one or more embodiments, the detection mapping of the part to be inspected may comprise a plurality of pixels, each pixel among the plurality of pixels being associated with a respective value. Determining the presence or absence of an air zone defect in the part to be inspected may include:

[0049] - determine if there is a group of neighboring pixels in the detection map of the room to be inspected for which the values ​​associated with the pixels of said group are below or above the predefined threshold:

[0050] if there is a group of neighboring pixels in the detection map of the part to be checked for which the values ​​associated with the pixels of said group are below or above the predefined threshold, detect an air zone type defect;

[0051] if there is no group of neighboring pixels in the detection map of the part to be checked for which the values ​​associated with the pixels of said group are less than or greater than the predefined threshold, detect an absence of air zone type defects.

[0052] According to these embodiments, the detection map is an image comprising a plurality of pixels, each pixel being associated with a respective value. The value associated with a pixel can be, for example, a gray level or a light intensity value. It is understood that the invention is not limited to a single value per pixel; for example, each pixel can be associated with a triplet of values ​​representing the intensity values ​​on the pixel's three red, green, and blue channels.

[0053] By "neighboring pixels" we mean pixels connected to each other by a connection relationship, for example, a 4-connectivity relationship. By "group of neighboring pixels" we mean that each pixel in the group is a neighbor of at least one other pixel in the group. The group of pixels may consist of a single pixel or more than two pixels. In some embodiments, a condition may be added regarding the number of pixels forming the group (thus, a set of neighboring pixels is considered a "group" in the above sense if it comprises a number of pixels greater than or equal to a predefined lower limit).

[0054] Depending on the embodiment, the pixel groups may include pixels whose values ​​are above the predefined threshold or pixels whose values ​​are below the predefined threshold. These embodiments are equivalent.

[0055] When the pixel values ​​of the detection map are above (or below) the predefined threshold, it means that the signals were more strongly attenuated when passing through the part, which may indicate the presence of a defect.

[0056] In some embodiments, the thickness mapping can be a thickness map of the entire part. The thickness map of the part to be inspected can comprise a plurality of pixels, each pixel of the plurality of pixels being associated with a respective value, and the simulated ultrasonic mapping can be determined by changing the value of at least one pixel among the plurality of pixels.

[0057] For example, the simulated ultrasonic mapping can be determined by multiplying the value of each pixel in the plurality of pixels of the thickness map of the part to be inspected by a respective coefficient. This coefficient can be the same for all pixels or a coefficient specific to each pixel.

[0058] It is understood that the invention is not limited to this embodiment. For example, the simulated thickness map can be obtained automatically from the thickness map of the part, using a machine learning algorithm, such as a generative adversarial network (or "GAN").

[0059] According to these embodiments, the thickness mapping is therefore performed on the entire part (and not on a portion of the part). The thickness map is "transformed" into a simulated ultrasonic map, which is therefore the ultrasonic map that a part similar to the part being inspected would have in the absence of defects.

[0060] In some embodiments, the process may further include:

[0061] - to realign the thickness mapping of the part to be checked with the ultrasonic mapping of the part to be checked;

[0062] and the transformation can be applied to the thickness mapping of the reworked part.

[0063] By "registration," it is understood that pixels from the two maps corresponding to the same points on the room's projection are aligned. This allows the respective information from the maps to be combined.

[0064] According to these embodiments, the registration is applied before transforming the thickness map to obtain the simulated ultrasonic map.

[0065] Alternatively, the process may also include:

[0066] - to calibrate the simulated ultrasonic mapping of the room with the ultrasonic mapping of the room;

[0067] and the detection map can be determined from the ultrasonic mapping of the part and the simulated and recalibrated ultrasonic mapping of the part.

[0068] According to these embodiments, the registration is applied after the transformation, on the simulated ultrasonic mapping.

[0069] In alternative embodiments, the part to be inspected comprises a first part and a second part bonded together, the second part being manufactured according to a predefined digital model. The thickness mapping can be a thickness map of the first part of the part to be inspected before assembly with the second part, and the ultrasonic mapping of the part to be inspected can be an ultrasonic map of the assembled part. The process may further include:

[0070] - receive a reference ultrasonic map of the second part of the part to be inspected, the reference ultrasonic map of the second part of the part to be inspected corresponding to an ultrasonic map of a second reference part manufactured from the predefined digital model and not including any air zone type defects;

[0071] and the simulated ultrasonic mapping can be determined from the thickness mapping of the first part of the part to be inspected and the reference ultrasonic mapping of the second part of the part to be inspected.

[0072] According to these embodiments, the part to be inspected comprises at least two parts bonded together. The bonding area may typically contain defects such as air pockets, for example, due to insufficient adhesive or an air bubble in the adhesive. Ultrasonic mapping is always performed on the complete (and assembled) part. In contrast, thickness mapping only concerns the first part of the part, before it is bonded to the second part. In these embodiments, a so-called "reference" ultrasonic map of the second part is also obtained.This reference ultrasonic mapping of the second part of the part is typically an ultrasonic mapping performed on a second part of the part "similar" to the second part of the part being inspected but which is known to be free of defects (as was the case in the prior art method on the whole part; conversely here it only concerns the second part of the part).

[0073] Such embodiments are particularly advantageous when the part includes a portion whose thickness varies very little, and a portion whose thickness varies more.

[0074] In these embodiments, the thickness mapping of the first part of the component to be inspected may comprise a plurality of pixels, each pixel of the plurality of pixels being associated with a respective value. The method may further comprise:

[0075] - modify the value of at least one pixel among the plurality of pixels of the thickness map of the first part of the part to be checked to obtain a so-called transformed thickness map of the first part of the part to be checked;

[0076] in which the simulated ultrasonic mapping can be determined from the reference ultrasonic mapping of the second part of the part to be inspected and the transformed thickness mapping of the first part of the part to be inspected.

[0077] For example, this modification can be made by multiplying the value of each pixel by the plurality of pixels in the thickness map of the first The part of the piece to be controlled by a respective coefficient. This coefficient can be the same coefficient for all pixels or a coefficient specific to each pixel.

[0078] It is understood that the invention is not limited to this embodiment. For example, the transformed thickness map can be obtained automatically from the thickness map of the first part of the part to be controlled, using a machine learning algorithm, such as a generative adversarial network (or "GAN").

[0079] The thickness map of the first part of the part is thus transformed into an equivalent ultrasonic map in the absence of defects, and the simulated ultrasonic map is determined by "grouping" the reference ultrasonic map of the second part and the equivalent ultrasonic map obtained for the first part.

[0080] In some embodiments, the process may further include:

[0081] - realign the thickness mapping of the first part of the part and the ultrasonic mapping of the second part of the part with the ultrasonic mapping of the part;

[0082] in which the transformation can be applied to the thickness mapping of the reworked part; and

[0083] in which the simulated ultrasonic mapping of the part can be obtained by summing the ultrasonic mapping of the second part of the re-marked part and the mapping obtained by applying the transformation to the thickness mapping of the re-marked part.

[0084] Alternatively, the process may also include:

[0085] - to realign the mapping obtained by applying the transformation to the thickness mapping of the realigned part and the ultrasonic mapping of the second part of the part with the ultrasonic mapping of the part;

[0086] in which the simulated ultrasonic mapping of the part can be obtained by summing the ultrasonic mapping of the second part of the re-marked part and the mapping obtained by applying the transformation to the thickness mapping of the re-marked part.

[0087] In one or more embodiments, the transformation can be a multiplication by a predetermined coefficient.

[0088] In particular, the process may include a preliminary calculation of the predetermined coefficient. The preliminary calculation of the coefficient may include:

[0089] - for each part in a set of parts without air zone defects having the same composition as the part to be inspected:

[0090] receive a thickness map of said part, the thickness map comprising a plurality of pixels, each pixel being associated with a respective value;

[0091] to receive an ultrasonic map of said room comprising a plurality of pixels, each pixel being associated with a respective value;

[0092] determine a relationship between pixel values ​​of the thickness map with corresponding pixel values ​​of the ultrasonic map;

[0093] - calculate the coefficient from the determined ratios.

[0094] In some embodiments, the detection map is obtained by subtracting the ultrasonic map of the part to be inspected from the simulated ultrasonic map.

[0095] For example, the part could be an aeronautical component. In particular, in the case where the part comprises two parts, the part could be an engine blade, in which the first part of the part is a metal leading edge, and in which the second part of the part is a body made of woven composite material.

[0096] In some embodiments, the air zone type defect may be the presence of a crack, fissure, delamination or delamination.

[0097] In some embodiments, maps are type C visualizations.

[0098] Another aspect of the invention relates to an ultrasonic device for detecting air zone defects in a room to be inspected. The device may include:

[0099] - an input interface configured for:

[0100] receive an ultrasonic map of the room to be inspected;

[0101] receive a thickness map of at least a portion of the part to be checked;

[0102] - a circuit configured for:

[0103] determine, from the thickness mapping of at least one portion of the part to be checked, a simulated ultrasonic mapping of at least one portion of a part without air zone type defects having the same composition and the same thickness mapping as the at least one portion of the part to be checked;

[0104] to determine, from the ultrasonic mapping of the part to be inspected and the simulated ultrasonic mapping, a detection map of the part to be inspected;

[0105] determine, by applying a predefined threshold to the detection map of the part to be inspected, the presence or absence of an air zone type defect in the part to be inspected.

[0106] A computer program, implementing all or part of the process described above, installed on pre-existing equipment, is in itself advantageous.

[0107] Thus, the present invention also relates to a computer program comprising instructions for the implementation of certain steps of the process described above, when this program is executed by a processor.

[0108] This program can use any programming language (for example, an object-oriented language or other), and be in the form of interpretable source code, partially compiled code, or fully compiled code.

[0109] Figure 3, described in detail below, can form the flowchart of the general algorithm of such a computer program.

[0110] 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

[0111] Other features and advantages of the invention will become apparent from the description, which can be read in conjunction with the figures. These figures are provided for illustrative purposes only and are not intended to limit the scope of the invention.

[0112] Figures 1a to 1e illustrate some steps of a prior art method for detecting air zone defects.

[0113] Figure 2 represents a flowchart of a prior art method for detecting air zone defects.

[0114] Figure 3 represents a flowchart of a method for detecting air zone defects according to an embodiment of the invention.

[0115] Figure 4 represents a flowchart of a method for detecting air zone defects according to another embodiment of the invention.

[0116] Figures 5a, 5b and 5c illustrate a step in determining a simulated ultrasonic map of the part according to an embodiment of the invention.

[0117] Figure 6 represents a flowchart of a method for detecting air zone defects incorporating a determination of the threshold to be applied according to embodiments of the invention.

[0118] Figure 7 represents an example of an air zone type defect detection device according to embodiments of the invention. DETAILED DESCRIPTION

[0119] Figure 3 represents a flowchart of a method for detecting air zone defects according to an embodiment of the invention.

[0120] According to this embodiment, the part is considered as a whole (whereas in the embodiment described with reference to Figure 4, it is considered as an assembly of two parts glued together).

[0121] In a step 220 similar to that described with reference to Figure 2, an ultrasonic map, for example a C-scan map, of the room is received. The ultrasonic map of the room is typically an image comprising a plurality of pixels, in which each pixel is associated with a light intensity value that represents an amplitude of the received signal (or equivalently, an attenuation value of the ultrasonic signal as it propagates through the room, also called the ultrasonic attenuation value) in the region of the room corresponding to that pixel. In other words, each pixel reflects how the emitted ultrasonic signal is attenuated as it propagates through the room. Thus, pixels corresponding to areas of the room containing air pockets are generally associated with light intensity values ​​indicating greater ultrasonic signal attenuation.

[0122] In step 310, a thickness map of the part is obtained. The part thickness map is typically an image comprising a plurality of pixels, in which each pixel is associated with a light intensity value that represents a part thickness along a given direction. For example, if the part is in an (X, Y, Z) coordinate system, a part map along the Z-axis represents the part thickness along the Z-axis, shown in the (X, Y) plane.

[0123] Thickness mapping can be obtained, for example, using a state-of-the-art dimensional control method. For instance, thickness mapping can be established from measurement data obtained using a coordinate measuring machine (CMM). Such a machine, using probes and / or optical sensors moved by a measuring arm, determines the dimensions, shape, and position of the object being measured within a given coordinate system, and thus the object's thickness in several directions. The measured data can then be processed, for example, by metrology software that allows for object modeling and the detection of dimensional deviations from the digital model used to manufacture the part.

[0124] The ultrasonic mapping of the received part in step 220 and the thickness mapping of the received part in step 310 are typically images of the same dimensions (i.e., having the same number of pixels along both image axes). It should be noted that steps 220 and 310 can be implemented in any order, or in parallel.

[0125] In one or more embodiments, it is possible to implement a 320 registration process to align the ultrasonic mapping of the part with the thickness mapping of the part. Indeed, depending on how the two maps were obtained, they may not perfectly match. For example, a pixel corresponding to a point on the part may not to have the same coordinates in both images. Typically, there may be a rotation and / or translation of one of the images relative to the other. In order to match the two maps and thus combine their respective information, a transformation can therefore be applied to one of the maps to "realign" it with the other map.

[0126] Such image registration techniques are known to those skilled in the art. In particular, it is possible to use a registration technique based on reference points of the part, whose positions are known. The pixels corresponding to these reference points can be identified on each of the maps, and the transformation to be applied to perform the registration can be determined from these pixels. For example, if a point M of the part is associated with pixel p with coordinates (x, y) in the ultrasonic map and with pixel p' with coordinates (x', y') in the thickness map, the transformation T to be applied is such that T(x', y') = (x, y) (or such that T(x, y) = (x', y')). As mentioned above, these techniques are known and are not further detailed here. An example of such a registration is illustrated in Figures 5a, 5b, and 5c, in a particular embodiment of the invention.

[0127] In one or more embodiments, a transformation can be applied to the thickness map of the part during step 330 to modify the intensity values ​​of the thickness map pixels in order to obtain a so-called "simulated" ultrasonic map of a part similar to the part under consideration (i.e., having the same composition, structure, and dimensions) that would not exhibit any defects. At the end of step 330, a simulated ultrasonic map of the part is thus obtained. The word "simulated" here refers to the fact that the map obtained "simulates" an ultrasonic map, even though it is obtained (at least partially) from a thickness map. It is important to note that the word "simulated" here does not refer to a map that would be entirely derived from a computer simulation.

[0128] For example, in step 330, it is possible to multiply the light intensity values ​​of each pixel of the thickness map by the same predefined coefficient to obtain the simulated ultrasonic map. This coefficient can be determined upstream of the detection method according to the invention, for example, from a part having several known thicknesses, for which An ultrasonic mapping is performed. For example, a stepped shim of varying thicknesses can be used (e.g., 5 steps with thicknesses of 1 mm for the first step, 2 mm for the second, and so on, up to 5 mm for the fifth step; the number of steps and their respective thicknesses are provided as an example only and are not exhaustive), with the thickness of each step being known. The ultrasonic mapping allows for the determination of ultrasonic attenuation values ​​for each step (e.g., -3 dB for the first step, -4 dB for the second step, and so on, up to -7 dB for the fifth step). A correlation between the thickness values ​​and the corresponding attenuation values ​​is then established.For example, a relationship of the type: Att = (Ep x n) + z is established, where Att represents the ultrasonic attenuation, Ep represents the thickness, and n and z are two parameters determined from the experimental values ​​of ultrasonic attenuation and thickness obtained on the stepped block. Models of the relationship between ultrasonic attenuation and thickness other than a linear model like the one above are, of course, possible.

[0129] Alternatively, it is possible to define different coefficients for different regions of the part's thickness map. For example, these regions could correspond to areas of the part made of different materials (and therefore with different average attenuation coefficients), and the coefficients to be applied to the pixels in each region could be determined as before.

[0130] Other methods for determining the coefficient(s) can be implemented, for example, machine learning methods. For instance, a machine learning model can be trained on a training database comprising pairs of maps, each pair including a thickness map of a part and an ultrasonic map of the same part, to determine, from a thickness map, the coefficient(s) to apply to obtain the "equivalent" ultrasonic map. Such a model could be, for example, a neural network, in particular a GAN (Generative Adversarial Network), but the invention is not limited to this example.

[0131] At the end of step 330, the simulated ultrasonic map is "similar" to an ultrasonic map, in the sense that the light intensity value of a pixel of the simulated ultrasonic map is at least approximately equal to a light intensity value of a pixel corresponding to the same point in the ultrasonic mapping. In other words, the applied transformation allows thickness values ​​to be converted into equivalent ultrasonic attenuation values.

[0132] The transformation applied in step 330 advantageously allows the light intensity values ​​of the ultrasonic mapping and the thickness mapping to be sufficiently comparable, so that the information from these two maps can be combined.

[0133] In the preceding section, step 330 was described in the case where the transformation is applied to the thickness map of the part. It is noted that alternatively, the transformation can be applied to the ultrasonic map of the part to modify the intensity values ​​of the pixels in the ultrasonic map so that they coincide with the intensity values ​​of the pixels in the thickness map of the part, in a manner similar to that described above.

[0134] In step 340, the simulated ultrasonic map determined in step 330 is subtracted from the received ultrasonic map in step 220. Alternatively, it is of course possible to subtract the ultrasonic map from the simulated ultrasonic map. By "subtraction," we mean a pixel-by-pixel subtraction, which means that each pixel of the subtracted image is associated with a value that corresponds to the difference between the value of the corresponding pixel (i.e., the pixel with the same coordinates) in one of the maps and the value of the corresponding pixel in the other map.

[0135] The map obtained at the end of step 340, also called the "detection map," reveals air zone defects. These zones are indeed associated with higher pixel values. This is because, in the absence of a defect, the ultrasonic attenuation values ​​are directly correlated to the thickness of the part. Conversely, if the ultrasonic signal passes through an air zone defect, it is attenuated more. For the pixels corresponding to this defect, a difference therefore appears between the values ​​of the ultrasonic map (which include the defect) and the values ​​of the thickness map (which represent values ​​"equivalent" to the attenuation values ​​of the part, without the defect).

[0136] The detection map obtained in step 340 exhibits fewer false positives (i.e., false ultrasonic readings) than the processed map obtained in step 230 of the prior art method described with reference to Figure 2, because it is much less susceptible to factors related to material variability. This detection map therefore provides significantly more reliable detection than the prior art method described above.

[0137] Similar to the method in Figure 2, a thresholding 260 can be applied to the detection map obtained in step 340. In particular, it is possible to retain only the pixel values ​​above a predetermined threshold. In one or more embodiments, the threshold can be determined as described with reference to Figure 2 (step 250). For example, the threshold could correspond to an attenuation percentage of -6 dB.

[0138] In one embodiment, pixels with values ​​above the threshold can be set to a first reference value, for example, a gray level of 0 (black pixel), and pixels with values ​​below the threshold can be set to a second reference value, for example, a gray level of 255 (white pixel). The thresholded detection map is then a black and white image in which defects appear black and the background white. Alternatively, pixels with values ​​above the threshold can retain their current values, and pixels with values ​​below the threshold can be set to a reference value, for example, a gray level of 255 (white pixel). The detection map is then an image with a white background, and the defects retain their intensity values ​​calculated in step 260. Other embodiments are possible.

[0139] Similar to the method in Figure 2, the presence or absence of air zone defects can be determined in step 270, based on the thresholded detection map obtained in step 260. In one embodiment, zones corresponding to groups of neighboring pixels whose values ​​are all above the predefined threshold used for thresholding 260 are considered defects. "Neighboring pixels" is understood to mean a set of pixels in which each pixel is connected to at least one other pixel in the set by a connection link (e.g., a 4-connectivity link). In some embodiments, it is possible to define a minimum number Nmin of pixels for each group. According to these embodiments, to be detected as corresponding to a defect, the group of neighboring pixels must include at least Nmin neighboring pixels. If there is no group of neighboring pixels whose values ​​exceed the predefined threshold, there is no defect in the inspected part.

[0140] Figure 4 represents a flowchart of a method for detecting air zone defects according to another embodiment of the invention.

[0141] In this embodiment, the part to be inspected comprises two parts (a first part and a second part) intended to be bonded together. For example, the part to be inspected could be an engine blade comprising a body and a leading edge, the body being made of a woven composite material and the leading edge being made of metal, for example, titanium. The body and the leading edge are manufactured separately and are bonded together to form the engine blade. In such parts comprising bonded sections, there is a risk that the bonding may contain air pockets (due, for example, to insufficient or missing adhesive or to porous areas), which weaken the part and must be detected. It is understood that this embodiment can be extended to a part comprising more than two parts intended to be bonded together (at least one part and another part).

[0142] During step 220, an ultrasonic map, for example a C-scan map, of the complete and assembled part (i.e., the parts of the part are already bonded) is received. This step is similar to step 220 described with reference to Figure 2 or Figure 3. In the example of the motor blade above, during step 220, an ultrasonic map of the assembled blade is therefore received.

[0143] During step 302, a reference ultrasonic map, for example a C-scan, of a first part is received. By "reference ultrasonic map" it is understood that the ultrasonic map is not necessarily a map of the first part of the part to be inspected, but can be an ultrasonic map of a first part of the same type as the first part of the part to be inspected.

[0144] In the example of the engine blade, the first part of the component is typically the body made of composite material. The body is conventionally manufactured from a digital reference model specific to a particular type of blade, which is The same applies to all blades manufactured according to this particular blade type. Thus, in step 302, a reference ultrasonic map of a blade body manufactured from the digital reference model associated with the blade to be inspected is received. Generally, the first part of the component is one whose thickness varies relatively little from one part to another (this is the case for parts made of woven composite material).

[0145] It is therefore understood that the same ultrasonic reference map of the first part of the part can be determined upstream of the control process of Figure 4, and that it can be used to control several different parts from the same digital model specific to this type of first part of the part (i.e. that this reference map can be used for steps 302 of several control processes according to Figure 4 applied to several parts to be controlled).

[0146] It is noted that the reference ultrasonic mapping of the first part received in step 302 is an ultrasonic mapping of a first part (e.g., a reference part) before it is bonded to the second part.

[0147] In step 304, a thickness map of the second part of the component to be inspected is received before it is bonded to the first part. In the example above, this is the thickness map of the titanium leading edge of the component to be inspected, before it is bonded to the woven composite blade body. Indeed, in an engine blade comprising a bonded assembly of a woven composite body and a titanium leading edge, it is the titanium leading edge that exhibits the greatest variability in thickness from one component to another. Therefore, it is this component whose exact thickness we want to know for the component in question. Generally, the second part of the component is chosen as the one with the greatest variability in thickness from one component to another (and thus relative to a reference model according to which the second parts of components are manufactured).

[0148] It is therefore noted that step 302 concerns a reference ultrasonic mapping, while step 304 concerns a thickness mapping of the second part of the part under consideration (to be checked).

[0149] Thickness mapping of the second part of the part to be checked is typically obtained by a part thickness measurement technique as described above, with reference to step 310 of Figure 3.

[0150] At step 325, a registration of the different maps can be performed. This registration makes it possible to "coincide" the pixels corresponding to the same points of the part in the different maps (i.e., to put them at the same coordinates, so that all the maps correspond to the same image of the part).

[0151] An example of such a registration 325 is illustrated in Figures 5a, 5b, and 5c. Figure 5a represents an ultrasonic map 510 of the assembled part to be inspected. The crosses on the ultrasonic map 510 represent "reference" points, used to register the images relative to each other. In practice, these reference points are positioned according to physical features of the parts before assembly, visible on the different maps. On the ultrasonic map 510 of the assembled part to be inspected, the two parts 512 and 514 of the part appear; that is, in the example of the blade, the leading edge and the body, respectively.

[0152] Figure 5a also represents the 520 reference ultrasonic mapping of the second part 514 of the part (here, the blade body), as well as the reference points (also represented by crosses).

[0153] The reference points of maps 510 and 520 are matched to determine the transformation used to register the reference ultrasonic map 520 of the second part of the part to portion 514 of the ultrasonic map 510 (portion 512 is ignored for this registration step). In the illustrated case, the transformation is composed of a translation, a rotation, and a scaling. Once the transformation is determined and applied to all pixels of the reference ultrasonic map 520 of the second part of the part, a transformed reference ultrasonic map 530 of the second part of the part is obtained.

[0154] Figure 5b shows the same 510 ultrasonic mapping of the assembled part to be inspected as Figure 5a. Figure 5b also shows the mapping Thickness Tl of 540 of the first part of the piece (here, the leading edge), as well as the reference points (also represented by crosses).

[0155] As before, the reference points of the 510 and 540 maps are aligned to register (step 325) the thickness map 540 of the first part of the component with the ultrasonic map 510 of the assembled component to be inspected. Furthermore, a transformation can be applied to the thickness map 540 of the first part of the component (before or after registration) so that the light intensity levels of the pixels in the thickness map 540 of the first part of the component generally correspond to the light intensity levels of the corresponding pixels (i.e., representing the same point on the component after registration), as detailed previously with reference to step 330 of Figure 3.After the registration and transformation (for example, multiplying the pixel values ​​of the thickness map 540 of the first part of the part by one or more coefficients), we obtain a simulated ultrasonic map 550 of the first part of the part, registered with the ultrasonic map of the complete part 510 (more precisely, with the portion 512 of the map 510).

[0156] Figure 5c represents a simulated ultrasonic map 560 of the part obtained from the transformed reference ultrasonic map 530 of the second part of the part and the simulated ultrasonic map 550 of the first part of the part. For example, this simulated ultrasonic map 560 of the complete part is obtained by summing, pixel by pixel, the maps 530 and 550. It is noted that here, the word "simulated" is used to denote the fact that the map 560 simulates an ultrasonic map even though it is partially derived from thickness data (for the first part of the part, i.e., the titanium leading edge in the previous example).

[0157] At the output of step 330, we therefore obtain a simulated ultrasonic map 560 of the part as represented by element 560 of Figure 5c.

[0158] Steps 340, 260, and 270 in Figure 4 are similar to steps 340, 260, and 270 in Figure 3. Thus, in step 340, a detection map is obtained by subtracting the simulated ultrasonic map determined in step 330 from the ultrasonic map received in step 220 (in one direction or in the other). the other). Then, a threshold 260 can be applied to the detection map obtained in step 340, and the presence or absence of air zone type defects can be determined in step 270, from the thresholded detection map obtained in step 260.

[0159] Figure 6 represents a flowchart of a method for detecting air zone defects incorporating a determination of the threshold to be applied according to embodiments of the invention.

[0160] In step 330 (which replaces step 210 in Figure 2), a simulated ultrasonic map of the part is obtained. Step 330 can, for example, be implemented as described previously with reference to Figures 3 and 4. In step 220, an ultrasonic map of the part to be inspected is obtained. The simulated ultrasonic map of the part can be subtracted (step 340) from the C-scan map of the part to obtain a detection map.

[0161] As described with reference to Figure 2, in step 240, a C-scan map of the defective part (the defects being known, as detailed above) is obtained and used to determine an optimal threshold (step 250) for detecting the defects, as described above. In step 260, the threshold determined in step 250 can be applied to the processed map obtained in step 340. The defect(s) can then be identified in the thresholded map in step 270.

[0162] Steps 330, 220 and 240 can be implemented in parallel or in any order (although generally, steps 330 and 240 on the reference part and the defective part are implemented before carrying out the actual inspection of a part, therefore before step 220).

[0163] The process in Figure 6 is therefore similar to the process in Figure 2, but instead of using an ultrasonic map of a reference part, it uses a simulated ultrasonic map of the part, determined from thickness data of the manufactured part. The variability in thickness due to the manufacturing process (which does not necessarily indicate the presence of a defect) is thus advantageously taken into account and integrated into the detection process. The resulting detection process is therefore more accurate and generates fewer false positives. related to the variability in manufacturing the part that the prior art process describes with reference to Figure 2.

[0164] Figure 7 represents an example of an air zone type defect detection device according to embodiments of the invention.

[0165] In these embodiments, the device includes a computer 700, comprising a memory 701 for storing instructions enabling the implementation of the process, the various maps from which the detection process is implemented, and temporary data for carrying out different steps of the detection process described above.

[0166] The 700 computer also includes a 702 circuit. This circuit can be, for example, a processor capable of interpreting instructions in the form of a computer program, an electronic card whose steps of the process of the invention are described in silicon, or a programmable electronic chip such as an FPGA chip (for "Field-Programmable Gate Array").

[0167] The computer 700 includes an input interface 703 for receiving ultrasonic and / or thickness maps, and an output interface 704 for providing a detection map or one or more pieces of information relating to defect detections (for example, an indication of the absence or presence of defects, and / or, when a defect is detected, an indication of its location). Finally, the computer may include a screen 705 and a keyboard 706 to allow for easy interaction with a user. Of course, the keyboard is optional, particularly in the case of a computer in the form of a tablet, for example.

[0168] Furthermore, the functional diagrams shown in Figures 3 and 4 are typical examples of programs in which certain instructions can be executed using the described device. As such, Figure 3 can be considered the flowchart of the general algorithm of a computer program as defined by the invention.

[0169] Of course, the present invention is not limited to the embodiments described above by way of example. It extends to other variations. For example, the method described above is advantageously applicable in cases of ultrasonic inspection of industrial parts with complex geometries. and variations in thickness that can generate false readings. Such industrial parts may include bonded or welded assemblies for which it is necessary to guarantee the quality of the bonding or welding, and in particular to determine if they have areas of missing material, porosity, or contain foreign matter. Such industrial parts may also include homogeneous parts subjected to material integrity testing, during which volumetric defects such as delamination or inclusions are sought.

[0170] In order to obtain equivalent control results and reliable detection of possible defects over the entire inspected area, it is currently necessary to correct for attenuation variations related to the different thicknesses traversed. This correction can be digital, via a multi-gain control, or physical, through the use of a reference part.

[0171] In these scenarios, the invention makes it possible to reduce acquisition times by performing a single-gain acquisition, but also to overcome the misleading indications caused by differences between production parts and the reference part.

Claims

CLAIMS

1. A computer-implemented method of ultrasonic detection of air zone defects in a part to be tested, the method comprising: - receive an ultrasonic map of the part to be checked; - receive a thickness map of at least a portion of the part to be checked; - determining, from the thickness map of at least one portion of the part to be inspected, a simulated ultrasonic map of at least one portion of a part without air zone type defects having the same composition and the same thickness map as the at least one portion of the part to be inspected; - determine, from the ultrasonic mapping of the part to be checked and the simulated ultrasonic mapping, a detection mapping of the part to be checked; - determine, by applying a predefined threshold to the detection map of the part to be checked, the presence or absence of an air zone type defect in the part to be checked.

2. The method of claim 1, wherein the thickness map is a thickness map of the entire part, wherein the thickness map of the part to be tested comprises a plurality of pixels, each pixel of the plurality of pixels being associated with a respective value, wherein the simulated ultrasonic map is determined by modifying the value of at least one pixel among the plurality of pixels of the thickness map of the part to be tested.

3. The method of claim 1, wherein the part to be inspected comprises a first part and a second part assembled by gluing, the second part being manufactured according to a predefined digital model, wherein the thickness mapping is a thickness mapping of the first part of the part to be inspected before assembly with the second part of the part to be inspected, wherein the ultrasonic mapping of the part to be inspected is an ultrasonic mapping of the assembled part to be inspected, the method further comprising: - receiving a reference ultrasonic map of the second part of the part to be inspected, the reference ultrasonic map of the second part of the part to be inspected corresponding to an ultrasonic map of a second reference part manufactured from the predefined digital model and not including an air zone type defect; wherein the simulated ultrasonic map is determined from the thickness map of the first part of the part to be inspected and the reference ultrasonic map of the second part of the part to be inspected.

4. The method of claim 3, wherein the thickness mapping of the first portion of the part to be inspected comprises a plurality of pixels, each pixel of the plurality of pixels being associated with a respective value, the method further comprising: - modifying the value of at least one pixel among the plurality of pixels of the thickness map of the first part of the part to be checked to obtain a so-called transformed thickness map of the first part of the part to be checked; in which the simulated ultrasonic map is determined from the reference ultrasonic map of the second part of the part to be checked and the transformed thickness map of the first part of the part to be checked.

5. Method according to one of the preceding claims, in which the detection map is obtained by subtracting the ultrasonic map of the part to be checked and the simulated ultrasonic map.

6. Method according to one of the preceding claims, in which the part is an aeronautical part.

7. A method according to the preceding claim in combination with one of claims 3 and 4, wherein the part is an engine blade, wherein the first part of the part is a metal leading edge, and wherein the second part of the part is a body of woven composite material.

8. Method according to one of the preceding claims, in which the air zone type defect is a presence of crack, fissure, delamination or detachment.

9. A device for ultrasonic detection of air zone type defects in a part to be tested, the device comprising: - an input interface configured to: o receive an ultrasonic map of the part to be checked; o receive a thickness map of at least a portion of the part to be checked; - a circuit configured to: o determine, from the thickness map of the at least one portion of the part to be inspected, a simulated ultrasonic map of at least one portion of a part without an air zone type defect having the same composition and the same thickness map as the at least one portion of the part to be inspected; o determine, from the ultrasonic map of the part to be inspected and the simulated ultrasonic map, a detection map of the part to be inspected; o determine, by applying a predefined threshold to the detection map of the part to be inspected, a presence or absence of an air zone type defect in the part to be inspected.

10. Computer program product comprising instructions for implementing the method according to one of claims 1 to 8 when this program is executed by a processor.

Images