Non-destructive testing method for a mechanical part

Abstract

The present invention relates to a computer-implemented, non-destructive testing method for an object (1), comprising a tomographic volume reconstruction consisting of the following steps: a) obtaining (E1) a first set of 2D tomographic projections of the object (1) for a first position; b) obtaining (E2) a second set of projections of the object (1) for a second, different position; c) reconstruction (E4) from the projection sets of the first and second volumes; d) processing (E5) of the volumes to segment a contour of the first and a contour of the second volume; e) processing (E6) of registering the first volume with respect to the second volume by contour registration; f) merging (E7) the registered first volume with the second volume to obtain a single volume; g) processing (E8) of the volume of the object (1) to detect any defects. Figure for the abstract: Fig. 1

Description

Title of the invention: Non-destructive testing method for a mechanical part technical field

[0001] The present invention relates generally to the field of non-destructive testing and in particular to the testing by X-ray tomography of the internal structure of a mechanical part. STATE OF THE ART

[0002] X-ray tomography is used for the non-destructive testing of many objects, for example, industrial mechanical parts. However, due to local information loss in the thickest areas of the parts, which are difficult for X-rays to penetrate, this technique cannot reconstruct the entire three-dimensional volume, particularly of a large or complex-shaped metallic or composite mechanical part. Furthermore, other artifacts are caused by phenomena such as beam hardening or geometric blurring.

[0003] The reconstructed volume therefore frequently contains areas with degraded geometry or material integrity. In some extreme cases, blind spots, incorrectly interpreted as empty, may be obtained. Analyzing a part from an imperfect reconstruction is therefore incomplete and cannot meet the need for high-quality overall inspection.

[0004] To solve these problems, software solutions are known that rely on algorithmic corrections to limit the harmful phenomena that cause artifacts or to assist the operator in optimally positioning the part before taking measurements. Hardware solutions also exist, such as the Beam Hole Array or the Beam Stop Array, which limit the influence of scattered radiation in order to provide a more coherent signal.

[0005] However, while these solutions make it possible to correct certain harmful phenomena such as scattering, beam hardening or metallic artifact, they have no effect on the lack of local information, thus preventing a correct reconstruction of the three-dimensional volume. Description of the invention

[0006] One object of the present application is to remedy the aforementioned drawbacks, by proposing a computer-implemented non-destructive testing method comprising a reconstruction of a tomographic volume of an object in accordance with the attached claims.

[0007] To this end, according to a first aspect of the invention, a non-destructive testing method for an object implemented by computer is proposed, comprising a reconstruction of a tomographic volume including the following steps:

[0008] a) obtaining a first set of 2D tomographic projections of the object for a first position of the object;

[0009] b) obtaining a second set of 2D tomographic projections of the object for a second position of the object different from the first position;

[0010] c) reconstruction from the first and second sets of 2D tomographic projections of a first volume of the object and a second volume of the object;

[0011] d) processing of the first and second volume to segment a first contour of the first volume of the object and a second contour of the second volume of the object;

[0012] e) registration process of the first volume of the object with respect to the second volume of the object by registration of the first contour with respect to the second contour so as to obtain a first registered volume of the object;

[0013] f) merging the first rejected volume of the object with the second volume of the object so as to obtain a volume of the object;

[0014] the method further comprising:

[0015] g) processing of the volume of the object in order to detect at least one possible defect.

[0016] The method according to the invention is advantageously complemented by the following features, taken alone or in any technically possible combination thereof: - a step of normalizing the grey levels of the pixels of the first and second sets of 2D tomographic projections of the object, step c) being implemented on the first and second sets of normalized 2D tomographic projections of the object, step f) using a second volume of the normalized object; - the first position of the object and the second position of the object are defined with respect to calibration markers visible on each 2D tomographic projection and on each reconstructed volume, the registration processing in step e) consisting of registering the calibration markers visible in each volume; - the merging in step f) consists of averaging the grey levels of each volume; - step f) of fusion is carried out for a fraction of the first registered volume with a corresponding fraction of the second volume of the object, so as to obtain a volume of the object; - the processing of the volume of the object in order to detect at least one possible defect includes a comparative metrological analysis; - the processing of the volume of the object in order to detect at least one possible defect includes a characterization of the materials of the object. DESCRIPTION OF THE FIGURES

[0017] Other features, objectives and advantages of the invention will become apparent from the following description, which is purely illustrative and not limiting, and which should be read in conjunction with the accompanying drawings on which:

[0018] Fig. 1 illustrates an object to be reconstructed by tomography according to a first position for the implementation of a process according to the invention;

[0019] Figure 2 illustrates an object to be reconstructed by tomography according to a second position for the implementation of a process according to the invention;

[0020] Figure 3 is a diagram showing the steps of a volume reconstruction process of an object by a tomographic scanner.

[0021] Throughout the figures, similar elements bear identical references. DETAILED DESCRIPTION OF THE INVENTION

[0022] The invention proposes a method, implemented by a computer, of non-destructive testing comprising a reconstruction of a tomographic volume of an object 1 from at least two volumes of the object 1 whose information losses are not located in the same places.

[0023] First embodiment

[0024] To this end, the reconstruction of a tomographic volume according to a first embodiment includes a step of obtaining a first set of 2D tomographic projections of the object 1 for a first position of the object 1 (step 11). The first set of 2D tomographic projections of the object 1 is, for example, obtained from a cone beam volume CT scan or from any other tomography or computed tomography technique. As illustrated in [Fig. 1], such a scan is performed by an X-ray imaging system 2 which includes an X-ray emitter 21 and a detector 22, the object 1 being positioned between the emitter 21 and the detector 22.Preferably, in order to improve the quality of 2D tomographic projections of object 1 by reducing noise and improving contours, particularly for a metallic object 1, the X-ray imaging system 2 includes a beam-hole array (BHA) 4 or a beam-stop array (BSA) 5 plate disposed between the source 21 and object 1.

[0025] Object 1 is also scanned a second time, preferably by the same tomographic scanner 2 and with adapted parameters, in a position different from the position of the first scan, as illustrated [Fig. 2]. The positions are defined according to the geometry of the part by following the usual methods in tomography, in particular by tilting the flat or cylindrical surfaces in the three spatial directions. to limit the shape effects and by limiting the thickness of material to be traversed to avoid information loss.

[0026] This second scan allows obtaining a second set of 2D tomographic projections of object 1 (step E2).

[0027] From the first and second sets of 2D tomographic projections, a reconstruction step makes it possible to obtain a first and a second volume of object 1 (step E4). These volumes of object 1 are made up of voxels, each voxel comprising position coordinates and a grey level.

[0028] Advantageously, it is possible to perform the reconstruction (step E4) with fixed gray levels, that is, to define, during a normalization step, lower and upper thresholds around the peaks of the constituent elements of the sets of 2D tomographic projections of the object 1 to be analyzed, the remaining gray levels being able to be saturated (step E3). An upper (respectively lower) threshold can thus be defined based on the maximum (respectively minimum) value of the constituent elements of the sets of 2D tomographic projections of the object 1. These constituent elements are, for example, the pixels of the sets of 2D projections, each pixel comprising a gray level.

[0029] This normalization step subsequently allows the maximum and minimum grey levels of each element constituting the volumes of object 1 to be superimposed as closely as possible.

[0030] Next, a processing of the first and second volumes is implemented to segment a first contour of the first volume of object 1 and a second contour of the second volume of object 1 (step E5). In this respect, and preferably, it is possible to generate a mesh of the surfaces constituting the contours of the volumes of object 1 and to assign a specific value to the mesh points belonging to the contours, so as to perform a segmentation between the air surrounding the volumes of object 1 and the material constituting object 1.

[0031] Next, a registration process is performed on the first volume relative to the second volume of object 1, by registering the first contour of the first volume of object 1 with respect to the second contour of the second volume of object 1, in order to obtain a registered first volume of object 1 (step E6). The registration process provides a geometric transformation function between the two volumes. Each point of one volume of object 1 is associated with a point of another volume of object 1.

[0032] Finally, in a merging step (step E7), the first recalibrated volume of object 1 is merged with the second volume of object 1 to obtain an improved volume of object 1, in which information loss is reduced or eliminated. The merging of the volumes consists of performing a union operation on the points of each volume, during which the gray level of each point or each The voxel of a pair of points or merged voxels is averaged to obtain the gray level of the point or voxel of the enhanced object 1 volume. Alternatively, the gray levels can be blended, summed, or only the highest gray level among the pair of points can be selected.

[0033] Second embodiment

[0034] In a second embodiment, the first and second sets of 2D tomographic projections of the object 1 also include 2D tomographic projections of calibration markers 3. As illustrated in [Fig. 1] and 2, the calibration markers 3 are placed near the object 1, between the source 21 and the detector 22 of the X-ray imaging system 2. In particular, the calibration markers 3 are placed relative to the object 1 in the same positions for each scan by the X-ray imaging system 2.

[0035] The method according to this second embodiment is very similar to the method according to the first embodiment, that is to say, a first and at least a second set of 2D tomographic projections of object 1 (steps E1 and E2), each set of 2D tomographic projections of object 1 comprising projections of the calibration markers 3. Similarly, each volume of object 1 reconstructed during the process comprises reconstructed volumes of the calibration markers 3 (step E4). In other words, the calibration markers 3 are visible on each 2D tomographic projection and on each volume of object 1, the associated data comprising information relating to the calibration markers 3 and the volumes of object 1. These volumes of object 1 and the calibration markers 3 are typically made up of voxels, each voxel comprising position coordinates and a gray level.

[0036] According to this embodiment, it is also possible to normalize the grey levels of the constituent elements of the sets of 2D tomographic projections of the object 1 and the calibration markers 3, that is to say, to define during a normalization step (step E3) lower and upper thresholds around the peaks of the constituent elements of the sets of 2D tomographic projections of the object 1 and the calibration markers 3, in a manner analogous to the normalization described in the previous embodiment.

[0037] Next, a processing of the first and second volumes is implemented to segment the first contours of the first volume of object 1 and the calibration markers 3, and a second contour of the second volume of object 1 and the calibration markers 3. In this respect, and preferably, it is possible to generate a mesh of the surfaces constituting the contours of the volumes of object 1 and the calibration markers 3, and to assign a specific value to the mesh points belonging to the contours, so as to perform a segmentation between the air surrounding the volumes of object 1 and the calibration markers 3 and the material constituting the object. 1 and calibration markers 3.

[0038] Next, a registration process is performed in which the first position of object 1 and the second position of object 1 are defined with respect to the calibration markers 3 visible on each 2D tomographic projection and on each volume of the reconstructed object 1. Thus, the calibration markers 3 visible in each volume of object 1 are registered with each other (step E6).

[0039] A merging step (step E7), similar to the first embodiment, consists of merging the first recalibrated volume of object 1 with the second volume of object 1 so as to obtain an improved volume of object 1, in which information loss is reduced or eliminated. The merging of the volumes can be performed with the volumes of the calibration markers 3 or by removing them.

[0040] For both the first and second embodiments of the invention, the merging step can be performed partially on a user-selected fraction of a reconstructed volume of object 1, thereby reducing the computation time required for the merging step by decreasing the number of points to be processed. The fraction of the registered volume of object 1 is then merged with a corresponding fraction of the second volume of object 1. By "corresponding," it is understood that each point of the fraction of the first registered volume of object 1 belonging to the fraction is associated by the geometric transformation function of the registration with a point of the second volume of object 1. Thus, each fraction corresponds to the same area on the real object 1. Alternatively, the selected fraction can be chosen automatically, based on a certain predetermined grayscale range.For example, the range could be between a minimum grey level value close to that assigned to air and a maximum grey level value corresponding to matter without loss of information.

[0041] Of course, the steps (steps E1 to E7) for reconstructing a tomographic volume of an object 1 can be performed sequentially. An improved volume resulting from a first iteration of the steps can be processed during a second iteration with a second volume of object 1. This second volume of object 1 may or may not be derived from a third iteration carried out successively or in parallel with the first iteration, using volumes of object 1 that have not undergone reconstruction according to the invention. In this way, it is possible to merge volumes of object 1 acquired in different positions, so as to eliminate all areas of information loss.

[0042] The steps for reconstructing a tomographic volume of an object 1 described above, regardless of the embodiment, can be integrated into a non-destructive testing method for an object 1, comprising a step of obtaining a volume of the object 1 by means of the steps for reconstructing a tomographic volume of the object. 1 (steps E1 to E7) and a processing step (step E8) of the part's volume to detect at least one potential defect, such as a dimensional and / or material condition defect. A dimensional defect is defined as a discrepancy between the volume's dimensions and the specified dimensions exceeding a predetermined threshold. The processing step (step E8) consists of segmenting the air and the material by defining segmentation rules specific to the material, geometry, and conformity requirements of object 1. The segmentation selects the voxels considered to belong to the air or the material constituting object 1. Metrological analyses are then performed according to methods known from the prior art. Similarly, a material condition defect is a discrepancy observed between the mechanical properties or the structure of the object 1's volume and the conformity requirements.The processing step (step E8) then consists of comparing (or segmenting) the gray levels of the volume of object 1 to a threshold in order to characterize the materials constituting object 1, the threshold being determined in a manner adapted to the structure and quality of the volume of the improved object 1. Compared to the prior art, the non-destructive testing method according to the invention makes it possible to ensure superior product quality by performing the tests on more realistic tomographic volumes because they do not involve any loss of local information.

[0043] Such a non-destructive testing method for an object 1 is particularly advantageous in any type of industry requiring precise control of the dimensions and material integrity of any type of object 1, including large, metallic objects. For example, the method is particularly well-suited for the lossless reconstruction of the volume of a turbomachine component.

Claims

Claims

1. A computer-implemented method for non-destructive testing of an object (1), comprising a reconstruction of a tomographic volume comprising the following steps: a) obtaining (El) a first set of 2D tomographic projections of the object (1) for a first position of the object (1); b) obtaining (E2) a second set of 2D tomographic projections of the object (1) for a second position of the object (1) different from the first position; c) reconstruction (E4) from the first and second sets of 2D tomographic projections of a first volume of the object (1) and a second volume of the object (1); d) processing (E5) the first and second volumes to segment a first contour of the first volume of the object (1) and a second contour of the second volume of the object (1);e) processing (E6) of resetting the first volume of the object (1) relative to the second volume of the object (1) by resetting the first contour relative to the second contour so as to obtain a first reset volume of the object (1); f) merging (E7) of the first reset volume of the object (1) with the second volume of the object (1) so as to obtain a volume of the object (1); the method further comprising: g) processing (E8) of the volume of the object (1) with a view to detecting at least one possible defect therein.;

2. Method according to claim 1, comprising a step (E3) of normalizing the gray levels of the pixels of the first and second sets of 2D tomographic projections of the object (1), step c) being implemented on the first and second sets of 2D tomographic projections of the object (1) normalized, step f) using a second volume of the object (1) normalized.

3. Method according to one of claims 1 or 2, in which the first position of the object (1) and the second position of the object (1) are defined relative to calibration markers (3) visible on each 2D tomographic projection and on each reconstructed volume, the registration processing in step e) consisting of registering the calibration markers (3) visible in each volume.

4. Method according to one of claims 1 to 3, in which the fusion (E7) in step f) consists of averaging the gray levels of each volume.

5. Method according to one of claims 1 to 4, in which the merging step f) (E7) is carried out for a fraction of the first volume realigned with a corresponding fraction of the second volume of the object (1), so as to obtain a volume of the object (1).

6. Method according to one of claims 1 to 5, the processing (E8) of the volume of the object (1) with a view to detecting at least one possible defect comprising a comparative metrological analysis.

7. Method according to one of claims 1 to 6, the processing (E8) of the volume of the object (1) with a view to detecting at least one possible defect comprising a characterization of the materials of the object (1).

Images