Method for non-destructive testing of a part

Abstract

One aspect of the invention relates to a method for non-destructive testing of a part (10), comprising determining a depth of a defect (20) internal to the part (10) from a plurality of 2D images of the part (10).

Description

DESCRIPTION TITLE: Non-destructive testing method for a part TECHNICAL FIELD OF THE INVENTION

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

[0002] In particular, the invention relates to a method for non-destructive testing of a mechanical part or a material sample comprising an internal defect, the method comprising a determination of the depth of the internal defect in the mechanical part. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0003] Volumetric Non-Destructive Testing (NDT) techniques allow for the detection, location, and measurement of internal defects in mechanical parts without damaging them. The internal defect can typically be a cavity or the inclusion of foreign material within the part. It can also be a crack, a deformation, or any anomaly indicating a discrepancy between the inspected part and a reference part.

[0004] Among volumetric NDT techniques, radiographic (RT) techniques are widely used in industry to inspect the internal condition of mechanical parts. These techniques involve obtaining an image of the material density of the part by passing X-ray or gamma radiation from a source. The image is obtained using a detector, which can be, for example, a photostimulable screen or an array of digital detectors. The absence of material (in the case of a cavity) or the presence of a material other than that of the part induces a difference in local absorption, which translates into a difference in light intensity in the generated image (for example, a difference in gray level).

[0005] When the generated image is a two-dimensional (2D) image, the depth of the defect relative to a surface of the part cannot be directly determined from the image. To determine the defect depth without reconstructing a 3D volume, the parallax method is commonly used. This method is based on moving the source or the part, and possibly the detector, along an axis orthogonal to the direction of X-ray propagation. The drawback of this method is the increased size of the radiographic system along this axis and the need for one or more motorized translation stages on which the X-ray source, sample, or detector is placed, with the associated constraints of cost, accuracy, and stability.

[0006] There are also techniques for obtaining three-dimensional (3D) images of the part. For example, X-ray tomography allows for a three-dimensional map of the part from multiple acquisitions performed from different viewing angles. In cone-beam tomography, the acquisition is generally performed by rotating the object 360° while it is placed on a rotating platform (conventional tomography). Alternatively, the source and detector can be placed on two independent robot heads, allowing them to be positioned at different locations around the object. Tomography techniques can determine the location of defects within the part but have drawbacks, including the complexity and cost of the mechanical system, the time required for inspection and data processing (during 3D volume reconstruction), and the overall size of the system.In addition, when the part exhibits significant dimensional variability along certain directions (for example, in the case of an elongated part), X-rays are strongly attenuated according to the exposure angles associated with these directions, which prevents a correct reconstruction of the 3D volume, and therefore the detection of defects - as well as the determination of the depth of the defect - in certain areas of the part.

[0007] Tomosynthesis, or laminography, is another 3D imaging technique in which the source and detector follow linear trajectories (rather than circular ones, as in conventional tomography) within the reference frame of the part being inspected. This technique is well-suited for inspecting parts with an elongated shape that cannot be fully rotated due to space constraints or excessive X-ray attenuation in certain directions. However, the spatial resolution obtained is anisotropic and degraded in the direction orthogonal to the detector plane. Again, the consequence is that the 3D volume reconstruction is not sufficiently reliable in certain areas of the part, preventing the detection of defects in these areas—and, consequently, the determination of the depth of these defects.This technique also has other disadvantages, such as the cost of the mechanical system, the time required for control and data processing (during the reconstruction of the 3D volume), as well as the overall size of the system.

[0008] Therefore, there is a need for a method of assessing the depth of a defect (cavity or inclusion) that does not present the aforementioned disadvantages. SUMMARY OF THE INVENTION

[0009] The invention provides a solution to the problems mentioned above by determining the depth of a defect within a part from a plurality of 2D images associated with different magnifications. A reference part can advantageously be used to estimate the magnification associated with each image in the plurality of 2D images.

[0010] It is noted that the invention can be implemented for any type of 2D image. In particular, the invention can be implemented for any image obtained using an imaging system comprising a radiation emission source and a detector (such as an X-ray system or a transmission imaging system).

[0011] One aspect of the invention relates to a computer-implemented method for non-destructive testing of a part, comprising determining the depth of an internal defect in the part, wherein a reference object is positioned opposite the part at a fixed distance from the part, the reference object being associated with a characteristic dimension of predefined value, wherein an assembly comprising the part and the reference object is called a measurement assembly, the determination of the defect depth comprising:

[0012] - receive a model illustrating a link between:

[0013] a dimension representing a projection of the defect onto a 2D image; and

[0014] a magnification associated with the 2D image;

[0015] in which the model depends on a set of parameters;

[0016] - receive a plurality of 2D images of the measurement set, each 2D image being associated with a respective magnification value from among a plurality of pairwise distinct magnification values;

[0017] Estimate, for each 2D image among the plurality of 2D images, a respective magnification value for said 2D image based on:

[0018] the predefined value of the characteristic dimension associated with the reference object; and

[0019] a value of the characteristic dimension associated with the reference object on the 2D image;

[0020] - determine, for each 2D image among the plurality of 2D images, a respective value of a dimension representative of a projection of the defect onto said 2D image;

[0021] - to estimate, using a regression performed on the estimated magnification values ​​and the determined values ​​of the representative dimension, a set of values ​​for the set of parameters; and

[0022] - determine, from the received model and the estimated set of parameters, the depth of the defect relative to the reference surface of the part.

[0023] For the sake of simplicity, the term "part" is used to refer to any sample of material (in particular a sample of material in the process of being manufactured). For example, and without limitation, a "part" could be a mechanical part, such as an aeronautical component.

[0024] By "internal part defect", we mean a defect located inside the part, that is to say located below the surface of the part.

[0025] The term "defect depth" refers to the distance from the defect (e.g., its center) to a predefined reference plane. The reference plane can be, for example, a plane connected to a surface of the part, specifically a plane along which a surface of the part extends. Typically, the reference plane is orthogonal to an axis (SD) between a radiation source and a detector in an imaging system. This axis (SD) corresponds, in particular, to the principal direction of radiation beam emission from the imaging system's source.

[0026] The term "reference object" here refers to an object external to the part, possessing at least one characteristic dimension. This reference object is used within the scope of the invention to estimate a magnification associated with a received image. The term "characteristic dimension" refers to a quantity associated with a dimension of the reference part, for example, a length. For instance, for a rectangular reference object, the characteristic dimension could be the length of one of its sides or its diagonal. For a circular reference object, the characteristic dimension could be the length of its radius or its diameter. The characteristic dimension could also be an area of ​​the reference object, or the square root of an area of ​​the reference object.

[0027] This characteristic dimension associated with the reference object has a predefined (known) value, which is the value associated with the "true" object (as opposed to the projection of the object onto the image).

[0028] On the contrary, the "value of the characteristic dimension associated with the reference object on the 2D image" corresponds to the value measured or read on the image.

[0029] By "positioned opposite the part," it is understood that the reference object is positioned on one side or the other of the part relative to the direction of radiation beam emission from the imaging system's source. In other words, the reference object can be located between the source and the part or between the part and the detector. The reference object is thus positioned opposite the part, so that both the part and the reference object are imaged by the imaging system.

[0030] By "positioned at a fixed distance from the part," it is understood that the reference object is positioned relative to the part such that the distance between the part and the reference object remains constant, even when the part is moved. Thus, the part and the reference object define a set, referred to here as the "measurement set," which can be moved "as a single unit." In other words, when the part is moved, the reference object is moved in the same way.

[0031] According to the present invention, the measuring assembly is moved by translation along the principal emission axis of the imaging system (for example, a radiographic system), so as to obtain several images associated with different magnifications. The depth of the defect is determined from this plurality of images.

[0032] The term "projection of the defect onto a 2D image" refers to the representation of the defect on that 2D image. The term "representative dimension" refers to a quantity representing the area occupied by the projection of the defect onto the 2D image. The representative dimension could be, for example, the area of ​​the projection of the defect, i.e., the area of ​​the representation of the defect on the image.

[0033] By "magnification," we mean a quantity relative to the position of an object to be imaged (here, the measurement set) with respect to the source and detector of the imaging system. In practical terms, two images associated with different magnifications represent the same object at different sizes (the closer the object is to the source, the larger it appears in the image).

[0034] It is understood that the model here is a theoretical model, for example established from geometric laws, and that it models the link between two theoretical quantities associated with the same "theoretical" image (i.e. it is any 2D image in general, not a specific image among those used to determine the depth of the defect).

[0035] By "set of parameters", we mean a set of quantities on which the model depends.

[0036] According to the present invention, a plurality of 2D images are obtained (for example, acquired or received from an external device) to determine the depth of the defect. This plurality of images comprises at least two images. Each image is associated with a respective magnification, and the magnifications associated with the different images are all distinct. This plurality of images is used to obtain pairs of values ​​for the variables of the received model, which allows the model parameters to be estimated by a regression method. The depth of the defect can then be determined from these parameters.

[0037] It is therefore possible, according to the above procedure, to determine the depth of a defect from at least two 2D images.

[0038] The method can therefore be implemented using images obtained via a very standard 2D imaging system, which does not need to be equipped with complex means for translating or rotating the source / detector / object elements. It should be noted that the relative position of the part with respect to the source and detector does not need to be known to implement the above method. Furthermore, the method can be implemented for any type of part and does not suffer from the aforementioned drawbacks for parts with significant length in one or more directions.

[0039] In one or more embodiments, the process may further include:

[0040] - determine, from the received model and the estimated set of parameters, a value for the dimension representative of the internal defect in the part.

[0041] In one or more embodiments, the defect may be a cavity, an inclusion, or chemical segregation.

[0042] The term "cavity" refers to an absence of material in the part being inspected, such as a blade, an air hole, or a crack. An "inclusion" refers to an object made of a material other than that of the part and located within the part being inspected. "Chemical segregation" refers to a defect characterized by a heterogeneity of chemical composition between different points within the part, due to a local variation in the concentrations of the various chemical elements composing the part's material (for example, in a solid alloy part), thus resulting in a variation in density.

[0043] It is understood that the word "defect" is used here simply to refer to any cavity or inclusion in the part being inspected, but that this term is not necessarily associated with an "undesirable" event. Indeed, inclusions or cavities may be intentionally present in the part during manufacturing, and a non-destructive testing process can be implemented, for example, to ensure that these features are correctly positioned within the part.

[0044] In one or more embodiments, for each 2D image among the plurality of 2D images, the respective magnification value for said 2D image can be estimated as a ratio between the predefined value of the characteristic dimension associated with the reference object and the value of the characteristic dimension associated with the reference object on the 2D image.

[0045] In one or more embodiments, the representative dimension of the projection of the defect onto a 2D image may correspond to a projection surface of the defect onto said 2D image.

[0046] In one or more embodiments, the determination, for each 2D image among the plurality of 2D images, of the respective value of the projection surface of the defect on said 2D image may include:

[0047] - obtain a segmentation mask of the defect on the 2D image;

[0048] - determine the value of the projection area on the 2D image from the segmentation mask obtained.

[0049] In one or more embodiments, the model can be a linear equation Y = aX + b, with Y = 1 / / S and X = 1 / y, where S is the value of the projection area of ​​the defect on a 2D image, y is the magnification value associated with the 2D image, and where a and b are two real numbers forming the parameter set.

[0050] In one or more embodiments, the 2D images are obtained via an imaging system comprising a radiation emission source and a detector, the measurement assembly being positioned between the source and the detector, in which the depth x of the defect relative to the reference surface of the part is determined such that:

[0051] x = -^ x SD

[0052] where SD corresponds to a distance between the source and the detector.

[0053] For example, the imaging system could be an X-ray system or a transmission imaging system.

[0054] Another aspect of the invention relates to a device for determining the depth of an internal defect in a part, wherein a reference object is positioned opposite the part at a fixed distance from the part, the reference object being associated with a characteristic dimension of predefined value, wherein an assembly comprising the part and the reference object is called a measuring assembly, the device comprising an input interface for:

[0055] - receive a model illustrating a link between:

[0056] a dimension representing a projection of the defect onto a 2D image; and

[0057] a magnification associated with the 2D image;

[0058] in which the model depends on a set of parameters;

[0059] - receive a plurality of 2D images of the measurement set, each 2D image being associated with a respective magnification value from among a plurality of pairwise distinct magnification values;

[0060] the device further includes a circuit configured for:

[0061] - Estimate, for each 2D image among the plurality of 2D images, a respective magnification value for said 2D image based on:

[0062] the predefined value of the characteristic dimension associated with the reference object; and

[0063] a value of the characteristic dimension associated with the reference object on the 2D image;

[0064] - determine, for each 2D image among the plurality of 2D images, a respective value of a dimension representative of a projection of the defect onto said 2D image;

[0065] - to estimate, using a regression performed on the estimated magnification values ​​and the determined values ​​of the representative dimension, a set of values ​​for the set of parameters; and

[0066] - determine, from the received model and the estimated set of parameters, the depth of the defect relative to the reference surface of the part.

[0067] It is understood that this device can be integrated into an imaging system, or can be external to an imaging system.

[0068] Another aspect of the invention relates to an imaging system comprising the above device.

[0069] For example, the imaging system could be an X-ray system comprising a photon emission source.

[0070] For example, the imaging system includes an X-ray source.

[0071] For example, the imaging system includes a gamma radiation source.

[0072] For example, the imaging system includes a neutron radiation source.

[0073] In one embodiment, the radiation source is a point source.

[0074] In one embodiment, the radiation beam from the source is collimated.

[0075] Alternatively, the imaging system can be a transmission imaging system.

[0076] A computer program, implementing all or part of the process described above, installed on pre-existing equipment, is in itself advantageous.

[0077] 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.

[0078] This program can use any programming language (e.g., an object-oriented language or other), and be in the form of interpretable source code, partially compiled code, or fully compiled code.

[0079] Figure 2, described in detail below, can form the flowchart of the general algorithm of such a computer program.

[0080] 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

[0081] 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.

[0082] Figure 1a represents a radiography system for obtaining images of a mechanical part according to an embodiment of the invention;

[0083] Figure 1b represents a radiography system for obtaining images of a mechanical part according to another embodiment of the invention.

[0084] Figure 2 represents an example of a method for determining the depth of an internal defect in a mechanical part, in one or more embodiments of the invention.

[0085] Figure 3 illustrates images associated with different magnifications, which can be used to implement the process in Figure 2.

[0086] Figure 4 represents an example of a device for determining the depth of an internal defect in a mechanical part, in one or more embodiments of the invention. DETAILED DESCRIPTION

[0087] Figures 1a and 1b illustrate a radiography system for obtaining images of a mechanical part according to different embodiments of the invention. It is understood that the systems shown in Figures 1a and 1b are purely illustrative and in no way limit the invention. In particular, it is understood that the invention can be applied to images obtained by imaging systems other than a radiography system.

[0088] The radiography system in Figure 1a includes a source 1 for emitting electromagnetic radiation, for example, X-rays or gamma rays. In the case of X-rays, the source 1 can be, for example, a conventional X-ray tube or a linear accelerator (Linac). The radiography system in Figure 1a also includes a detector 2. The detector 2 can be analog or digital (for example, a DDA flat panel detector, for "Digital Detector Array", or a reusable memory photostimulable display). The distance SD between the source 1 and the detector 2 is assumed to be fixed. For example, the source 1 and the detector 2 can be located in fixed positions throughout the process of determining the depth of a defect according to the invention.

[0089] Source 1 emits conical beams 15 of X-rays or gamma rays, along a principal direction SD orthogonal to detector 2. In Figure 1a, S denotes the position of the source 1 emitting electromagnetic radiation, and D represents the center of the emission circle resulting from the projection of the conical beam 15 onto detector 2. The line (SD) corresponds to the principal direction of emission of the conical beams 15 of electromagnetic rays from the radiography system.

[0090] A mechanical part 10 to be inspected is placed between the source 1 and the detector 2. This mechanical part 10 can, for example, be placed on a rail so that it can be moved by translation along the axis (SD). As detailed below, moving the mechanical part 10 by translation along the axis (SD) allows the magnification of the radiograph of the part 10 to be varied.

[0091] The mechanical part 10 includes an internal defect 20. By "internal defect," we mean a defect located inside the mechanical part and typically not visible to the naked eye (unlike some surface defects, for example). The internal defect 20 may be a cavity (i.e., an absence of material, for example, a hole or an air gap in the part) or an inclusion (i.e., a foreign body, which may typically be in a material other than that of the part).

[0092] It is noted that the presence of a cavity or inclusion in a mechanical part can be associated with an undesirable event (such as the formation of an unwanted air gap during the manufacturing process), but this is not necessarily the case. Indeed, a cavity or inclusion can also be intentionally inserted into the part during the manufacturing process (for example, a cavity can control wear or deformation of a part's surface, and an inclusion can serve to reinforce the part).

[0093] Thus, the term "defect" is used here to refer in a general way to a cavity or inclusion inside a mechanical part, but does not necessarily refer to an "undesired" or "abnormal" event.

[0094] The internal defect 20 is located at a depth d relative to a datum plane 35, the datum plane being orthogonal to the principal direction (SD) of emission from the source. The term "datum plane" here refers to the fact that this surface is used as a reference to determine the depth at which the internal defect 20 is located. The datum plane 35 is a plane of known position relative to the source 1 (or equivalently, the detector 2). The datum plane can, for example, be a plane connected to one of the surfaces of the mechanical part 10.

[0095] For example, if the mechanical part 10 comprises two opposing surfaces 10a, 10b in planes orthogonal to the principal direction (SD) of emission from the source, one of which 10b is planar (and therefore extends substantially along a plane), the reference plane 35 can be chosen as the plane along which this planar surface 10b extends. This facilitates the localization of the defect 20 within the part 10.

[0096] The depth d can be defined, for example, as the distance between the center (or any characteristic point) of the internal defect 20 and the reference surface 35.

[0097] In the context of the present invention, a reference object 30 is fixed to the part 10. This reference object 30 can be positioned opposite the downstream surface 10b along the principal oriented direction (SD) of radiation emission, at a predetermined distance from this surface 10b. Alternatively, the reference object 30 can be positioned opposite the upstream surface 10a along the principal emission direction. The predetermined distance can be zero, as in the case of Figure 1a. In this case, the reference object is positioned on the surface 10b of the part 10 (i.e., it is in contact with the surface 10b of the part 10).

[0098] In other embodiments, the predetermined distance of the reference object 30 from the surface 10b may be non-zero, as in the example shown in Figure 1b.

[0099] It is noted that the predetermined distance of the datum object 30 from the surface 10b is fixed. Thus, when the mechanical part 10 is moved in a translational motion along the axis SD, the datum object 30 is moved in the same translational motion. If the predetermined distance is non-zero, this simultaneous movement of the part 10 and the datum object 30 can be achieved, for example, by fixing the part 10 and the datum object 30 on the same rail capable of moving in a translational motion along the axis (SD). If the predetermined distance is non-zero, it is sufficient to fix (for example, using temporary fixings) the datum object 30 to the part 10 in such a way as to ensure simultaneous movement of the part 10 and the datum object 30.

[0100] In the example of Figure 1a, the reference object 30 is positioned on the surface 10b of the part 10, the surface 10b extending along the reference plane 35. Thus in this example, the face of the reference object in contact with the part 10 corresponds to the reference plane 35 used to define the depth d of the defect 20.

[0101] It is noted that in the examples in Figures 1a and 1b, the reference object 30 is positioned so as to be "centered" with respect to the conical beam 15. This is not mandatory.

[0102] The reference object 30 is an object of known dimensions, for example, a rectangular or circular plate, but this is not mandatory. The reference object 30 can have any shape, but it is preferable that the reference object 30 have two parallel, opposite flat faces, the two parallel faces being orthogonal to the principal direction (SD) of radiation emission when the reference object 30 is in place in the radiography system. Having two parallel faces ensures a constant thickness of the surface of the reference object 30 through which the radiation emitted by the source passes, which translates in the resulting image into a constant variation in light intensity across the entire surface.

[0103] In some embodiments, the reference object 30 may be made of a different material than the material of the part 10. In particular, the reference object 30 may be made of a material having a higher attenuation coefficient than the material of the part 10. In these embodiments, the reference object is more easily visualized in an image (here, radiographic) of the assembly including the part 10 and the reference object 30. For example, the reference object 30 may be made of tungsten.

[0104] To maximize the accuracy of determining the magnification of the reference object, the reference object can preferably be a relatively thin object (such as a plate, particularly a thin plate) extending substantially in the plane orthogonal to the source-detector axis. In some non-limiting embodiments of the invention, the reference object is a plate of dense material, such as tungsten, and has a thickness on the order of a hundred micrometers to a few millimeters—depending on the attenuation properties of the part and the energy of the radiation beam.

[0105] In the following, the assembly formed by part 10 and reference object 30 is called the "measurement set". Since part 10 and reference object 30 are moved simultaneously, the overall position of the measurement set between source 1 and detector 2 can be represented by a point O on axis (SD), with point O being associated with the measurement set. For example, point O could be, as in the example in Figure 1a, the point between contact surface 35 and axis (SD), or, as in the example in Figure 1b, a midpoint on axis (SD) between part 10 and object 30. Other positions of the origin O are possible.

[0106] According to the invention, the measuring assembly (i.e., part 10 and reference object 30) is moved by translation along the SD axis to obtain several images associated with several magnifications. The magnification is defined, conventionally, by the ratio: SD r ~ SÔ

[0107] where S is the position of the source, O represents a position of the measurement set and S is the position of the detector, the three points S, O and D being located on the same axis.

[0108] In general, the further the measurement set is from source 1 (and therefore closer to detector 2), the smaller its image projected onto detector 2.

[0109] In other words, according to the invention, the measuring system is moved along the SD axis to obtain several images of the measurement set, each image being associated with a respective position on the SD axis (i.e., a respective position of the origin O, and therefore a respective magnification value). Within the scope of the invention, at least two images are thus acquired.

[0110] Each image represents at least a portion of the reference part 30 and the entire defect 20. More precisely, each image represents a portion of the reference object used to measure a characteristic dimension of the reference object. In other words, said portion of the reference part 30 and the defect 20 must be within the conical beam 15, and their projection onto the detector 2 must be contained within the detector's surface.

[0111] To ensure that this condition is met, the measurement assembly can be moved by translation along a direction orthogonal to the SD axis (i.e., along the y-axis of Figures 1a and 1b). For this purpose, the imaging system can be equipped, for example, with a y-axis elevation mechanism, such as a platform on which the measurement assembly rests. This platform can be raised or lowered along the y-axis. This allows, in particular, the repositioning of the defect 20 within the conical beam 15 when it is no longer in the field of view.

[0011] These images can be used to determine the depth of the defect 20 in the part 10 relative to a reference plane 35, as described below with reference to the process shown in Figure 2.

[0113] Figure 2 represents an example of a method for determining the depth of an internal defect in a mechanical part, in one or more embodiments of the invention.

[0114] In step 210, an equation modeling the link between a dimension representing a projection of the defect onto a 2D image and a magnification associated with the 2D image.

[0115] The term "representative dimension" refers to any quantity measuring the extent of the defect's projection onto the image. The representative dimension could be, for example, an area or a length.

[0116] By "projection of the defect onto the image", we mean the representation of the defect that can be seen on the image.

[0117] It is understood here that the equation models the relationship between the representative dimension of the defect projection and the magnification of the same "theoretical" 2D image. This equation is typically determined using geometric laws.

[0118] The magnification value corresponds to the y value defined above for the image considered.

[0119] In some embodiments, the representative dimension corresponds to the projection surface of the defect, that is to say the surface (i.e., the area) of the shape associated with the projected image of the defect on the complete 2D image.

[0120] In these embodiments, the model received at step 210 can be written as: 1 1 1 d — = —= x - — (1) JS JS^ Y SD x 7^

[0121] where S represents the projection surface of the defect onto the image, S o represents the actual surface of the defect in the plane orthogonal to SD (i.e. the surface of a section of the defect in a plane orthogonal to SD), y represents a value of the magnification associated with the image (and therefore also corresponds to the magnification of the reference object), d represents the depth of the defect relative to the reference plane 35 and SD represents the (assumed fixed) distance between the source 1 and the detector 2.

[0122] It is noted that the example equation (1) above is advantageous because it provides a linear relationship between a variable Y depending on the surface S of projection of the defect onto the image and a variable X depending on the magnification y of the image: Y = aX + b

[0123] with Y = 1 / s and X = 1 / y, and a and b two real numbers which are here the parameters of the linear model.

[0124] Note that: 1 d a = — — and b = - — . TJSQ SD x TJSQ

[0125] It is understood that other relationships could be used. In particular, the equation may not be linear. It may also involve parameters other than those in the equation above. Furthermore, other representative dimensions besides the surface area may be used. For example, it is possible to model the shape of the defect as an ellipse, whose representative dimension may be the largest diameter.

[0126] During step 215, an image of the measurement set for a given magnification value is received.

[0127] In practice, the magnification value is not known. It can be estimated during step 220 using:

[0128] - a (known) value of a characteristic dimension of the reference object 30;

[0129] - a value of the same characteristic dimension measured on the projection of the reference object 30 onto the image considered (received at step 215).

[0130] The characteristic dimension can be, for example, a length or a diameter. For instance, when the reference object 30 is a rectangular plate, the characteristic dimension can be the length of one side (the shorter or longer side) or the diagonal of the rectangle. When the reference object 30 is a circular plate, the characteristic dimension can be the diameter or the radius of the circle. Other characteristic dimensions can be used, for example, the square root of the area of ​​the reference object (this quantity has the advantage of being able to characterize a reference object of any geometric shape). It is only important that the "actual" value (i.e., the value of the reference part 30) of this characteristic dimension be known and that the value of this same characteristic dimension can be measured on the image received in step 215.

[0131] For example, the value of the characteristic dimension measured on the projection of the reference object 30 can be obtained by multiplying the number of pixels of the characteristic dimension measured on the image by the (assumed known) size of the pixels.

[0132] The "actual" value of the characteristic dimension is assumed to be known and can typically be measured before implementation of the process in Figure 2 (and even before installation of the system in Figure 1a or 1b, i.e. before connecting the reference object 30 to the part 10 to be checked), for example using a caliper or a three-dimensional measuring machine, depending on the desired accuracy.

[0133] The magnification can be estimated by calculating the ratio between the known value L th of the characteristic dimension of the reference object 30 and the value L meas of the characteristic dimension measured on the image received in step 215: ^meas

[0134] where i is an index used to designate the image under consideration (i being a natural number greater than or equal to 1). y f This corresponds to the estimated magnification value for the i-th image, and Z / meas corresponds to the value of the characteristic dimension measured on the i-th image. It is understood that the value L th does not depend on the image, since it is defined from the reference part 30 as such (and not from an image).

[0135] During step 225, the representative dimension of the projection of the defect onto the image received in step 215 is determined.

[0136] When the representative dimension corresponds to the projection area S of the defect, the image can be segmented to obtain a defect segmentation mask, and the number of pixels inside the segmentation mask can be evaluated. Segmentation can be performed using any known segmentation method, for example, from global thresholding on the image or histogram, such as using an Otsu method, or from a watershed method, a k-means algorithm, etc. The projection area S of the defect can be evaluated by multiplying the number of pixels inside the segmentation mask by the area of ​​one pixel.

[0137] It is noted that steps 220 and 225 can be implemented in any order, or in parallel.

[0138] The sequence of steps 215-220-225 is implemented at least twice, successively or in parallel. Indeed, the method according to the invention requires at least two images of the measurement assembly associated with two different magnification values ​​to determine the depth of the defect 20 in the part 10 to be inspected.

[0139] The number of images received is denoted here as N, where N is a natural number greater than or equal to 2. Thus, the sequence of steps 215-220-225 is implemented N times, for N distinct magnification values, that is, for N positions of the measuring assembly on the SD axis. Therefore, to obtain the N images, the measuring assembly can be moved by translation along the SD axis, either manually or automatically, for example, using a motorized translation stage.

[0140] Following the N implementations of steps 215-220-225, N pairs of values ​​{y S , with i = 1, Or :

[0141] - If corresponds to the value of the dimension representing the projection of the defect onto the i-th image; and

[0142] - Yi corresponds to the magnification value associated with the i-th image.

[0143] From this data, it is possible to use regression to evaluate the parameters of the model obtained in step 210. When the model is linear, as in equation (1) above, the regression is a linear regression. If the model is not linear, it is possible to use other types of regression.

[0144] In the case of model (1) above, a linear regression thus allows us to evaluate the parameters a and b of the model: Y = aX + b

[0145] with Y = l / [S and X = 1 / y. In this example, linear regression is implemented from the pairs of values ​​{x^y for i = 1, with :

[0146] From the values ​​obtained for the parameters of the received model in step 210 using the regression implemented in step 235, it is possible to determine (step 240) the depth d of the defect 20 in the part 10, and, in some embodiments, the representative (actual) dimension S Q of the defect (for example, a cross-sectional area of ​​the defect).

[0147] For example, in the case of a linear regression applied to model (1) above, the depth d of the defect 20 can be determined from the parameters a and b by the following formula: bx SD d = - . has

[0148] The (actual) surface S Q The defect can be determined according to the formula: 1 S o = -, Cl

[0149] It is noted that the above procedure can be applied to a plurality of defects. The images used to determine the depth of two different defects may be the same, may be partially common, or may be completely different (this depends on whether the two defects appear simultaneously in an image or not).

[0150] The process in Figure 2 can be implemented independently for the different defects, successively or in parallel. Some steps can be shared for the different defects (for example, the segmentation of the defects and the measurement of their projection area).

[0151] Figure 3 represents an example of a plurality of images used to implement the process of Figure 2.

[0152] Figure 3 represents in particular 5 images 301, 302, 303, 304, 305, associated with 5 distinct magnifications: Yi = 1-08, y2= 1-22, y3= 1.49, y4= 1.83 and y5= 2.07.

[0153] Images 301, 302, 303 and 304 show two defects 20a and 20b. It can be observed that defect 20a is not visible in image 305.

[0154] Thus, the process in Figure 2 can be implemented from images 301, 302, 303 and 304 for defect 20a, and from images 301, 302, 303, 304 and 305 for defect 20b.

[0155] The above process is typically implemented by a 400 computer such as the one shown in Figure 4.

[0156] Thus, an object according to the invention relates to a device for determining the depth of an internal defect in a mechanical part comprising a computer 400, including a memory 401 for storing instructions enabling the implementation of the method for determining the depth of a defect in a mechanical part, predefined values ​​relating to the dimensions of the reference part and the pixels of the image, as well as the distance between the source and the detector, and temporary data for carrying out different steps of the method described above.

[0157] The 400 computer also includes a 402 circuit. This circuit can be, for example, a processor capable of interpreting instructions in the form of a computer program, or a programmable electronic chip such as an FPGA chip (for "Field-Programmable Gate Array").

[0158] The computer 400 includes an input interface 403 for receiving the model and images, and an output interface 404 for providing the value of the defect depth in the part thus determined. Finally, the computer may include, to allow easy interaction with a user, a screen 405 and a keyboard 406. Of course, the keyboard is optional, particularly in the case of a computer in the form of a tablet, for example.

[0159] It is noted that the above-determination device may or may not be integrated into an imaging system such as that shown in Figures 1a-1b. If the device is not integrated into an imaging system, images may be acquired via an external imaging system and then sent or transferred to the device.

[0160] The functional diagram shown in Figure 2 is a typical example of a program in which certain instructions can be executed using the described device. As such, Figure 2 can be considered a flowchart of the general algorithm of a computer program as defined in the invention.

Claims

DEMANDS

1. A computer-implemented method for non-destructive testing of a part (10), comprising determining the depth of a defect (20) internal to the part (10), wherein a datum object (30) is positioned opposite the part (10) at a fixed distance from the part (10), the datum object (30) being associated with a characteristic dimension of predefined value, wherein an assembly comprising the part (10) and the datum object (30) is called a measurement assembly, the determination of the depth of the defect (20) comprising: - receive (210) a model modeling a link between: o a dimension representing a projection of the defect (20) onto a 2D image; and o a magnification associated with the 2D image; in which the model depends on a set of parameters; - receive (215) a plurality of 2D images of the measurement set, each 2D image being associated with a respective magnification value from among a plurality of pairwise distinct magnification values; - estimate (220), for each 2D image among the plurality of 2D images, a respective magnification value for said 2D image from: o the predefined value of the characteristic dimension associated with the reference object (30); and o a value of the characteristic dimension associated with the reference object (30) on the 2D image; - determine (225), for each 2D image among the plurality of 2D images, a respective value of a dimension representative of a projection of the defect (20) onto said 2D image; - estimate (235), using a regression carried out from the estimated magnification values ​​and the determined representative dimension values, a set of values ​​for the parameter set; and determine (240), from the received model and the estimated parameter set, the depth of the defect (20) relative to a reference surface (10b) of the part (10).

2. A method according to claim 1, further comprising: - determine, from the received model and the estimated set of parameters, a value of the dimension representative of the defect (20) internal to the part (10).

3. A method according to any one of the preceding claims, wherein, for each 2D image among the plurality of 2D images, the respective magnification value (y) for said 2D image is estimated as a ratio between the predefined value of the characteristic dimension associated with the reference object (30) and the value of the characteristic dimension associated with the reference object (30) on the 2D image.

4. A method according to any one of the preceding claims, wherein the representative dimension of the projection of the defect (20) onto a 2D image corresponds to a projection surface of the defect (20) onto said 2D image.

5. A method according to the preceding claim, wherein the determination, for each 2D image among the plurality of 2D images, of the respective value of the projection surface of the defect (20) onto said 2D image comprises: - obtain a segmentation mask of the defect on the 2D image; - determine the value of the projection area on the 2D image from the segmentation mask obtained.

6. A method according to claim 4 or 5, wherein the model is a linear equation Y = aX + b, with Y = 1 / / S and X = 1 / y, where S is the value of the projection area of ​​the defect (20) onto a 2D image, y is the magnification value associated with the 2D image, and where a and b are two real numbers forming the parameter set.

7. A method according to the preceding claim, wherein the 2D images are obtained via an imaging system comprising a radiation emission source (1) and a detector (2), the measuring assembly being positioned between the source (1) and the detector (2), wherein the depth d of the defect relative to the reference surface (10b) of the part (10) is determined such that: b d = — x SD has where SD corresponds to a distance between the source (1) and the detector (2).

8. Device for determining the depth of a defect (20) internal to the part (10), wherein a datum object (30) is positioned opposite the part (10) at a fixed distance from the part (10), the datum object (30) being associated with a characteristic dimension of predefined value, wherein an assembly comprising the part (10) and the datum object (30) is called a measuring assembly, the device comprising an input interface (403) for: - receive a model illustrating a link between: o a dimension representing a projection of the defect (20) onto a 2D image; and o a magnification associated with the 2D image; in which the model depends on a set of parameters; - to receive a plurality of 2D images of the measurement set, each 2D image being associated with a respective magnification value from among a plurality of pairwise distinct magnification values; the device further comprising a circuit (402) configured to: - Estimate, for each 2D image among the plurality of 2D images, a respective magnification value for said 2D image based on: o the predefined value of the characteristic dimension associated with the reference object (30); and o a value of the characteristic dimension associated with the reference object (30) on the 2D image; - determine, for each 2D image among the plurality of 2D images, a respective value of a dimension representative of a projection of the defect (20) onto said 2D image; - estimate, using a regression performed on the estimated magnification values ​​and the determined values ​​of the representative dimension, a set of values ​​for the set of parameters; and - determine, from the received model and the estimated set of parameters, the depth of the defect (20) relative to a reference surface (10b) of the part (10).

9. Imaging system comprising a device according to claim 8.

10. Product computer program comprising instructions to implement the method according to any one of claims 1 to 7 when this program is executed by a processor.

Images