Tomographic analysis method

Abstract

The invention relates to a method for the tomographic analysis of a composite part (20) comprising fibers (21), the method comprising the following steps: - acquiring (ACQ) at least one tomographic image of the part (20) by means of a tomography device (1); - dividing (DIV) the tomographic image into elementary sub-portions; - measuring (MES) variations in at least one first quantity (G), characteristic of the greyscale in the sub-portion; - obtaining (OBT) variations in a comparison quantity (C) in at least one reference part; and - comparing (COMP) the variations in the first quantity (G) with the variations in the comparison quantity (C). The method comprises a step of: - registering (REC) the comparison quantity (C) in order to compensate for an offset (u) between the first quantity (G) and the comparison quantity (C).

Description

[0001] DESCRIPTION

[0002] TITLE: Tomographic Analysis Procedure

[0003] Technical field of the invention

[0004] The invention relates to a method for tomographic analysis of a mechanical part, particularly for the purpose of performing non-destructive testing on the part. The part may, for example, be part of an aircraft, and more specifically of a turbomachine, such as a section of a fan casing, a fan blade, or part of a fixed blade structure.

[0005] Prior art

[0006] It is known to inspect structural mechanical parts using tomographic methods, for example radiographic or acoustic, in order to detect the presence of possible defects on the surface and inside the part.

[0007] These methods are reliable, non-invasive and allow for inspection of the inside of the parts, which makes it possible to quickly determine the condition of the part and decide whether to potentially put it back into operation or replace it.

[0008] A tomographic measurement consists of scanning an observed object, here a mechanical part, using a beam of waves, and measuring the transmitted beam in all directions in order to reconstruct a three-dimensional image of the object.

[0009] Figure 1 illustrates a tomography device 1 during the inspection of a mechanical part 20.

[0010] The tomography device 1 includes at least one emitting device 3, configured to emit an incident beam 5 of wave pulses towards the part 20, for example radio frequency waves, X-rays or acoustic waves, and at least one receiver 7 capable of capturing a transmitted beam 9 of waves, arranged on either side of the part 20.

[0011] The part 20 is generally mounted on a support 11 that rotates around an axis A, so that it can be observed from all directions during the tomographic measurement. A reverse mounting is also possible, where the part is fixed and the transmitting / receiving device pair rotates around it.

[0012] Part 20 observed by tomography may be made of a composite material, as shown in detail in Figure 2.

[0013] Such a composite material may exhibit a so-called two-dimensional weave, which includes weft fibers generally oriented along a weft direction X and warp fibers generally oriented along a warp direction Y, woven together to form a weave plane P. The composite material may include several such weave planes P, or plies, superimposed along a thickness direction Z to form a fibrous preform, whose fibers are not intertwined from one weave plane to another.

[0014] Alternatively, the weaving of the 21 fibers can be three-dimensional, with 21 fibers extending along the three directions of space X, Y, Z, so as to form a thick three-dimensional fibrous preform.

[0015] The 21 fibers can be carbon fibers, glass fibers, or a mixture of both.

[0016] The fibers 21 of the material are embedded in a matrix 23 comprising, for example, one or more polymers and / or resins. To achieve this, the fibrous preform is generally placed in a mold into which the matrix is ​​injected in fluid form; the fluid then solidifies to form the final part 20. Alternatively, sheets, or plies, of woven fibers are coated with the fluid intended to form the matrix and draped in one or more layers onto a forming support before solidification.

[0017] During the tomographic analysis of such a composite part, the amplitudes of the waves transmitted under the different angles of observation are translated into grey levels in a processing device 13, and a digital analysis makes it possible to reconstruct a three-dimensional grey level image of the volume of the part 20. The processing device 13 includes at least one computer.

[0018] The resulting gray levels are assigned to the voxels forming the three-dimensional image, and the gray level of each voxel depends on the degree of wave absorption by the material at that voxel. This degree of absorption depends on the type of waves used and is strongly correlated with the density of the material within the voxel.

[0019] This three-dimensional image can then be subdivided into two-dimensional images obtained by cutting the image along parallel cutting planes, depending on the type of analysis planned.

[0020] Thus, tomography makes it possible to distinguish the different components, or phases, of the composite part, in particular by distinguishing the fibers of one or more types, the matrix, and possibly defects such as hollows formed for example by air bubbles, or others.

[0021] Tomography analysis provides a visual representation of both the exterior and interior of the part. This type of inspection offers the major advantage of allowing visualization through the thickness of the material and the part being studied, while remaining non-destructive and reliable.

[0022] A human operator can then assess the quality of the part by searching the three-dimensional image for the presence or absence of interlacing anomalies, or weave indications, and evaluating the material's conformity to applicable standards. For example, in the case of a woven composite part, weave anisotropy, fiber continuity, and matrix continuity are relevant parameters visually assessed by the operator. However, the analysis of a part such as a blower blade by a qualified inspector is time-consuming. This operation is manual and tedious, with much of the time spent scrolling through the tomography images. The operation also generates significant eye strain, which can reduce the inspector's vigilance and the quality of their analysis after a certain period.

[0023] It is therefore desirable to automate some of the processing steps for the obtained tomography images, particularly to enable the detection and isolation of regions likely to contain an anomaly. This allows the operator to focus on the most crucial areas of the part without having to examine the entire piece. An example of such processing, implementing a heuristic algorithm, is described in patent application FR 3141245 A1.

[0024] However, this type of treatment can still be improved.

[0025] Indeed, this type of analysis is based on comparing the obtained image with a two-dimensional or three-dimensional image from a sound sample—that is, one without weaving defects—measured beforehand and used as a reference. To do this, representative quantities of the gray-level distributions in the image, or metrics, are calculated for regions of the image comprising one or more pixels or voxels, as appropriate. These metrics are recorded and compared to values ​​obtained from sound reference samples. These metrics generally exhibit a significantly periodic variation across the entire piece, due to the periodic nature of the weave.However, it has been observed that during the manufacturing process, particularly during the placement of the fibrous preform in a mold and the injection of the matrix precursor, preform misalignment can occur. This is a small-amplitude displacement of the fibers that varies from one part to another. This misalignment, schematically represented in Figure 3, can have an amplitude of up to a few millimeters along each of the fiber alignment directions, but it remains impossible to predict.

[0026] This misalignment makes it difficult to compare the measured pieces with the reference pieces, once combined with the random nature of the weaving defects sought, which limits the performance of the heuristic detection algorithm by introducing false positives and negatives.

[0027] Presentation of the invention

[0028] The invention aims to remedy these drawbacks by proposing an improvement in the comparison of measured and reference images in such a method of analyzing tomographic images (in particular images of aircraft mechanical parts) based on a heuristic algorithm.

[0029] To this end, the invention relates to a method for tomographic analysis of a composite part comprising fibers, the method comprising the following computer-implemented steps: obtaining at least one tomographic image of the part acquired by means of a tomography device, said tomographic image comprising voxels exhibiting respective gray levels, dividing the tomographic image into elementary sub-parts, measuring, for each sub-part, the variations of at least a first characteristic quantity of the gray levels over at least one region of the sub-part, obtaining, for each sub-part, the variations of a reference quantity characteristic of the gray levels measured in at least one reference part, over a reference region corresponding to said region of said sub-part, and comparing, for each sub-part, the variations of the first quantity, determined on the part,with variations in the reference quantity, obtained on at least one reference part, and determination of conformity of the subpart from this comparison, characterized in that the method comprises, prior to the comparison step, a registration step, comprising, for each subpart, the calculation of at least one spatial offset, along at least one direction, between variations in the first quantity and variations in the reference quantity, and obtaining variations in the first registered quantity to compensate for the spatial offset, said first registered quantity being used in the comparison step, wherein the first quantity is a grey level value, the variations of which are measured along a measurement path extending into the subpart,the measurement path being preferably substantially straight and extending preferably substantially perpendicularly to a local fiber extent direction in the part.

[0030] Such a process makes it possible to compensate for the effects of cropping on the processing and comparison of tomographic images and thus reduces the number of false positives and false negatives, which improves the quality of the automatic evaluation of the part and the detection of sub-parts at risk.

[0031] The first quantity, measured along a measurement path traversing a sub-part, allows for a measurement on a region accurately characterizing a large number of fibers within the sub-part, taking into account intrinsic variations in the material and improving the accuracy of the comparison. The measurement step may include obtaining at least one two-dimensional image for each sub-part, each two-dimensional image comprising a plurality of pixels, each pixel exhibiting a respective gray level, and in which the first quantity is measured on at least one two-dimensional image.

[0032] Such a feature allows for the simulation of an analysis by observation and processes the data in a simple way and makes it observable later by an operator.

[0033] The reference quantity can be an average of the grey level value(s), measured in each reference region along a comparison path extending through the reference region along a path corresponding to that of the measurement path in the sub-part.

[0034] Such a characteristic allows for a good evaluation of periodic phenomena in a piece of woven composite material.

[0035] The process may include determining a comparison quantity, which is a sum of the reference quantity and a constant margin, determined from a standard deviation of the grey levels calculated on the reference parts and / or on the reference region, said margin being multiplied by a predetermined coefficient.

[0036] This feature allows the margin to be adapted to the random variations of each specific part.

[0037] The comparison step may include detecting and counting points for which the first recalibrated quantity is greater than the comparison quantity, and comparing the number of such detected points with a predetermined compliance threshold.

[0038] Such a feature makes it possible to detect large-dimensional regions containing anomalies, without taking into account isolated outliers due to measurement errors.

[0039] The spatial shift calculation step can be performed by maximizing the correlation between the first quantity and the comparison quantity.

[0040] Such a feature allows for a reliable and rapid calculation of the offset to be compensated between the compared quantities.

[0041] The realignment step can be carried out in two directions substantially perpendicular to each other, preferably each perpendicular to one of the directions of extent of the fibers in the part.

[0042] Such a feature makes it possible to compensate for misalignment along the directions of the warp and weft fibers.

[0043] The value of the offset can vary over the reference region, the recalibration step including recalibrating the variations of the first quantity of the local offset values ​​to obtain the variations of the first recalibrated quantity.

[0044] This feature allows spatial registration to be adapted to local variations in the weave. The invention also relates to a computer program product comprising instructions that, when executed on a computer controlling a tomographic analysis device, enable the implementation of the method described above.

[0045] Brief description of the figures

[0046] Figure 1 is a schematic side view of a tomography device during the implementation of a method according to an example of the invention.

[0047] Figure 2 is a schematic detail view of a part made of woven composite material.

[0048] Figure 3 is a schematic representation of one type of fiber unframing during the manufacturing of the part.

[0049] Figure 4 is a schematic representation of the steps of an analysis process according to an example of the invention.

[0050] Figure 5 is a graphical representation of the comparison of the uncropped signals of the part and a reference part.

[0051] Figure 6 is a graphical representation of the comparison of the cropped signals of the part and the reference part.

[0052] Detailed description of the invention

[0053] A method according to the invention for the tomographic analysis of a part 20 in order to determine its conformity will now be described with reference to figure 4.

[0054] The described process is executed entirely by a computer connected to the tomographic measurement system and allows for a diagnosis to be obtained quickly and without requiring the intervention of a qualified inspector.

[0055] Part 20 is made, for example, of composite material.

[0056] Part 20 here is part of a blade, for example from a blower blade.

[0057] The composite material comprises fibers 21, which extend for example in two directions of space perpendicular to each other to form weaving planes P, embedded in a matrix 23. The average gap between two neighboring fibers 21 of the same direction is on the order of 2 mm.

[0058] For example, part 20 is a part made of three-dimensional woven composite material, as shown in Figure 2.

[0059] The fibers 21 of the part then extend along three directions in space, along weaving planes perpendicular to the X and Y directions, tangent to the surface of the blade, with for example the X direction oriented along the upstream-downstream direction of the blade (in operating conditions), the Y direction oriented along the elongation direction of the blade, from the foot to the tip of the blade, and fibers extending along the Z direction, which is the thickness direction of the part along which the weaving planes P are organized.

[0060] The thickness of such a part 20, measured along the Z direction, is for example between 5 and 25 mm, notably close to 10 mm.

[0061] The 21 fibers can be of the same or different materials (glass or carbon for example).

[0062] Matrix 23 comprises at least one organic polymer and / or at least one resin.

[0063] The different types of fibers and the matrix are referred to as the phases of the composite material composing part 20.

[0064] As previously stated, during the manufacture of part 20, the fibers 21 may undergo a deframe, as shown in Figure 3, which is an overall displacement of the fibers 21 from their initial position in the mold.

[0065] This misalignment has little impact on the properties of part 20, but disrupts the tomographic analysis by shifting the fibers 21 relative to their position in reference parts used to compare part 20.

[0066] In the example shown in Figure 3, the fibers 21 are offset by a factor u along the X direction in the view on the right, relative to their initial position shown on the left. The process includes, for example, a QA acquisition step of at least one three-dimensional image of the part 20 using the tomography device 1, as shown in Figure 1. Alternatively, the three-dimensional image may have been acquired prior to the implementation of process 20 and retrieved, for example, from memory.

[0067] An incident beam 5 of wave pulses is emitted by the emitting device 3, towards the part 20. The incident beam 5 is, for example, a beam of radio frequency waves.

[0068] The waves pass through part 20, and a transmitted beam 9 from part 20 is captured by the receiver 7. The intensity distribution of the transmitted beam 9 obtained is converted into a two-dimensional greyscale image of part 20 by the processing device 13.

[0069] Part 20 is rotated by means of support 11, and two-dimensional images of part 20 are acquired in all directions.

[0070] A three-dimensional image of part 20 in greyscale, or tomographic image, is then reconstructed by image processing using the processing device 13.

[0071] The three-dimensional image comprises a plurality of voxels, each presenting a respective grey level representative of the attenuation coefficient of the material's phase and its corresponding density in part 20.

[0072] The tomographic analysis process then includes a step of dividing the three-dimensional tomographic image into elementary sub-parts. These sub-parts have dimensions, for example, along three spatial directions X, Y, Z, calculated from the material characteristics of part 20. The sub-parts are, for example, parallelepiped in shape.

[0073] The dimensions of the sub-parts are chosen for example as in patent application FR 3141245 A1.

[0074] The process then includes a MES measurement step, in each sub-part, of the variations of at least one first quantity, which is characteristic of the grey levels on at least one region of the sub-part.

[0075] According to one embodiment, the MES measurement step includes obtaining at least one two-dimensional image for each sub-part, each image being obtained for example by a planar section of the three-dimensional sub-part.

[0076] These images can be cross-sections of the three-dimensional sub-part in planes perpendicular to the X, Y, and Z directions of space, or in oblique planes (i.e., not orthogonal to the X, Y, and Z directions).

[0077] The number of such images obtained for each sub-part is predetermined empirically, according to the number of sections relevant for the conformity analysis of the sub-part.

[0078] Each image is thus two-dimensional and comprises a plurality of pixels, each pixel exhibiting a respective level of grey.

[0079] The process may advantageously include an image processing step to remove spurious grey levels that could distort subsequent digital processing as described in patent application FR 3141245 A1.

[0080] For example, pixels in the image whose grey level is outside a predetermined median range are replaced by pixels whose grey level is within that median range.

[0081] The median range extends, for example, between 10% and 90% of the full grayscale of the acquisition device. Pixels with a grayscale level between 0 and 10% and between 90% and 100% of the maximum of the scale (0 generally corresponding to black and 100% generally corresponding to white) are therefore considered noise and replaced.

[0082] These values ​​of 10% and 90% are given as an indication and may vary depending on the case, depending on the nature of the material and the sensitivity of the sensors of the acquisition device 1, for example.

[0083] The replacement gray level value is advantageously determined based on the gray level distribution of image 32, for example, chosen to be equal to the median gray level of the image pixels. More advantageously, the gray level of each replaced pixel in image 32 is determined randomly based on the gray level distribution of the pixels in image 32 located in the median range. Thus, the replaced pixels create white noise in the image and do not alter the subsequent processing steps.

[0084] The first quantity is then measured on at least one two-dimensional image, in particular on the set of two-dimensional images obtained for this sub-part.

[0085] Alternatively, the first quantity can be measured on the voxels of the subpart, without resorting to a two-dimensional image.

[0086] The first quantity is, for example, a numerical value associated with the grey level, and the variations of this grey level along a measurement path extending into the sub-part are measured to form a variation curve, as shown in Figure 5. The measurement path is, for example, substantially rectilinear and extends over the two-dimensional image, preferably perpendicular to one of the X, Y directions of fiber extent 21.

[0087] Other relevant parameters can be calculated, in addition to or instead of the variations in grey level, such as a standard deviation of the grey levels on the image, an autocorrelation score and / or an anisotropy score, or an image obtained by applying a filter to the two-dimensional image, for example a Gabor filter.

[0088] The variations of the first quantity can be determined along two or three directions perpendicular to each other, for example the X, Y and Z directions, by repeating the measurement on two or three different measurement paths, oriented along these directions.

[0089] In the case of a measurement on two-dimensional images, the variations of the first quantity along the Z direction perpendicular to the planar images is obtained by scrolling through several images obtained in the sub-part and aligned along the Z direction.

[0090] Thus, one or more evolution curves of the first quantity in the sub-part, along one or more directions, are obtained. An example of such a curve is shown in Figure 5 (spatial variations of the first quantity, curve G), and exhibits a spatial periodicity due to the presence of regularly spaced fibers in the material.

[0091] The process then includes an OBT step, for each subpart, of the variations of a reference quantity measured in at least one reference part, on a reference region corresponding to said region of said subpart.

[0092] Each reference piece is of the same type as the analyzed piece 20 and has previously been imaged as described above and certified as sound, i.e. not containing any significant anomalies, in order to serve as a basis for comparison.

[0093] The reference quantity is intended to be compared to the first quantity, which is characteristic of the grey levels for the acquired two-dimensional image, and is measured on a measurement path identical to that of the first quantity, in order to compare the first quantity and the reference quantity at the same points of the parts.

[0094] The spatial variations of the reference quantity are determined in the same way as those of the first quantity.

[0095] Variations in the reference quantity on the reference part(s) are, for example, pre-recorded for each of the sub-parts measured in the process, so as to implement the same comparison for each part analyzed.

[0096] The reference value is, for example, the average of the grey levels calculated on the reference pieces.

[0097] To account for random variations between weaves, a comparison size can be determined from the reference size, for example by adding a margin to the values ​​of the reference size, which makes it possible to determine a tolerable deviation for the first size with the reference size.

[0098] The constant margin is determined, for example, from the standard deviation of grey levels calculated on the different reference pieces, and / or the standard deviation of grey levels calculated on the reference region.

[0099] This standard deviation is, for example, multiplied by a predetermined coefficient empirically to calculate the fixed margin added at every point of the grey level variations to obtain the comparison magnitude.

[0100] The curves representing the variations of the reference quantity, i.e. the average grey levels alone (curve M), and of the variations of the comparison quantity, i.e. the grey levels added to the constant margin (curve C) are shown in Figure 5.

[0101] An abnormal region A is observed, in which the first quantity is greater than the comparison quantity, and which should therefore correspond to a defect in the part. However, this detection is a false positive due to the framing, which introduces a phase difference between curves G and C, and which the process aims to correct.

[0102] The process includes for this purpose a REC recalibration step, comprising, for each sub-part, the calculation of at least one spatial offset u of the variations of the first quantity with respect to their reference position, and the recalibration of the curve of variations of the first quantity according to the value of offset u obtained, in order to compensate for the misalignment.

[0103] This calculation of the offset u can be implemented by maximizing a correlation product between the two curves, in this example between the curves of the variations of the first quantity G and the variations of the reference quantity M.

[0104] We therefore wish to identify the lag value u that maximizes the correlation r between G and M, which is defined as:

[0105] Where * represents the convolution product.

[0106] Its analytical solution is obtained using the following formula: u = arg maxF -1 {F(G(x))' x F(M(x))}

[0107] Where F represents the Fourier transform, F -1 its inverse and ' the conjugate.

[0108] The curve of variations of the first quantity, P, is then recalibrated, according to the offset value u obtained, in order to compensate for the framing.

[0109] Figure 6 represents the curves G, G', M and C, where G' is the curve of the variations of the first quantity, recalibrated on the reference quantity M by the value of the offset u.

[0110] In one embodiment, the recalibration performed on the variations of the first quantity (G) is not constant over the entire measured range. The compensated offset is then a function u(x), which allows for better compensation of a misalignment that does not correspond to a rigid overall movement.

[0111] Such a lag u(x) can be estimated to be constant per segment, for example. The segments can, for instance, have a length that is a multiple of the observed period of the variations of the first quantity or the reference quantity. In particular, the segments can have a length equal to said period, with each maximum of the variations of the first quantity G then being individually recalibrated.

[0112] Alternatively, the lag u(x) can be calculated as a continuous function that maximizes the convolution product described above.

[0113] The process then includes a COMP comparison step, for each sub-part, of the variations of the first recalibrated quantity G', with the variations of the comparison quantity C, and of determining the conformity of the sub-part from the comparison.

[0114] In the example in Figure 6, the comparison involves detecting and counting the points for which the first quantity G' is greater than the comparison quantity C, and then comparing the number of such detected points to a predetermined threshold. The sub-part is deemed non-compliant if the number of detected points exceeds this threshold.

[0115] Figure 6 shows that the sub-part is detected as compliant once the signals have been recalibrated.

[0116] If necessary, an operator can then re-examine the part. In this case, the conformities of each sub-part and the curves showing the evolution of the characteristic quantities allow for the rapid location of the relevant regions of part 20 to be inspected, and facilitate this additional observation.

Claims

DEMANDS 1. Tomographic analysis method for a composite part (20) comprising fibers (21), the method comprising the following computer-implemented steps: obtaining (ACQ) at least one tomographic image of the part (20) acquired by means of a tomography device (1), said tomographic image comprising voxels having respective grey levels, dividing (DIV) the tomographic image into elementary sub-parts, measuring (MES), for each sub-part, the variations of at least a first quantity (G) characteristic of the grey levels on at least one region of the sub-part, obtaining (OBT), for each sub-part, the variations of a reference quantity (M) characteristic of the grey levels measured in at least one reference part, on a reference region corresponding to said region of said sub-part, and comparing (COMP), for each sub-part, the variations of the first quantity (G), determined on the part (20),with the variations of the reference quantity (M), obtained on at least one reference part, and determination of the conformity of the subpart from this comparison, characterized in that the method comprises, prior to the comparison step (COMP), a recalibration step (REC), comprising, for each subpart, the calculation of at least one spatial offset (u), along at least one direction, between variations of the first quantity (G) and the variations of the reference quantity (M), and obtaining variations of the first quantity (G') recalibrated to compensate for the spatial offset (u), said first quantity (G') recalibrated being used in the comparison step (COMP), in which the first quantity (G) is a grey level value, the variations of which are measured along a measurement path extending in the subpart,the measurement path being preferably substantially straight and extending preferably substantially perpendicularly to a direction (X, Y) of local extent of the fibers (21) in the part (20).

2. A method according to claim 1, wherein the measurement step (MES) comprises obtaining at least one two-dimensional image for each sub-part, each two-dimensional image comprising a plurality of pixels, each pixel having a respective gray level, and in which the first quantity (G) is measured on at least one two-dimensional image.

3. Method according to claim 1 or 2, wherein the reference quantity (M) is an average of the grey level value(s), measured in each reference region along a comparison path extending in the reference region along a path corresponding to that of the measurement path in the sub-part.

4. Method according to the preceding claim, comprising determining a comparison quantity (C), which is a sum of the reference quantity (M) and a constant margin, determined from a standard deviation of the grey levels calculated on the reference parts and / or on the reference region, said margin being multiplied by a predetermined coefficient.

5. Method according to the preceding claim, wherein the comparison step (COMP) comprises the detection and counting of points for which the first recalibrated quantity (G') is greater than the comparison quantity (C), and the comparison of the number of such detected points with a predetermined conformity threshold.

6. A method according to any one of the preceding claims, wherein the step of calculating the spatial offset (u) (CALC) is carried out by maximizing the correlation between the first quantity (G) and the comparison quantity (C).

7. A method according to any one of the preceding claims, wherein the registration step (REC) is carried out along two directions (X, Y) substantially perpendicular to each other, preferably each perpendicular to one of the directions (X, Y) of extent of the fibers (21) in the part (20).

8. A method according to any one of the preceding claims, wherein the value of the offset (u) varies over the reference region, the recalibration step (REC) comprising recalibrating the variations of the first quantity (G) of the local values ​​of the offset (u) to obtain the recalibrated variations of the first quantity (G).

9. Product computer program comprising instructions enabling, when executed on a computer controlling a tomographic analysis device (1), the implementation of the method according to one of the preceding claims.

Images