METHOD FOR DETERMINING DAMAGE TO A PART BY IMAGE STEREO-CORRELATION AND CLASSIFICATION IN SUPERVISED MODE

Abstract

Method for determining the damage to a part, said method comprising the steps: a) determining (101) the displacements of said part in the directions x, y, z over a given area, by stereo-correlation of images, b) calculating (102) the deformations in said directions, with the data from step a), c) carrying out (103) a classification of the data obtained at the end of step b), in supervised mode, on at least two classes, d) analyzing (104) the data thus classified in step c) to determine whether the part is damaged. Figure for abstract: Figure 3

Description

Title of the invention: METHOD FOR DETERMINING DAMAGE TO A PART BY STEREO-CORRELATION OF IMAGES AND CLASSIFICATION IN SUPERVISED MODE Technical field of the invention

[0001] The present invention relates to a method for determining the damage to a part subjected to mechanical stress. The method involves determining displacements of the part, for example by stereo-correlation measurements of digital images, and classification. The invention also relates to a device for implementing the method. Technological background

[0002] Stereo correlation of digital images is a known technique which allows measurement of displacement fields of a part.

[0003] It is a non-destructive surface control technique.

[0004] To do this, a black and white speckle is applied to the surface of the part whose displacement field under the effect of mechanical stress is to be determined, in order to provide the surface with a sufficiently random texture. This allows for tracking using cameras. These cameras are generally equipped with CCD sensors (from the English charge coupled device) to perform dynamic tracking with an acquisition frequency depending on the phenomenon studied. This can be from a few images per second to several thousand images per second for cases with rapid or cyclic dynamics.

[0005] From the displacement field, it is then possible to obtain the deformations. This technique consists in fact of matching two images of a surface of the part observed in two distinct states of deformation, a so-called reference state and a so-called deformed state. Each pixel of the first image is compared to the corresponding pixel on the second image which makes it possible to obtain the associated deformations.

[0006] The digital image stereo correlation technique can be used during a static type mechanical test.

[0007] A static test can be a test to characterize the properties of materials, a verification of the quality of the material, or a test to validate the mechanical strength of mechanical parts. During these static mechanical tests, it is necessary to determine the conditions under which damage occurs.

[0008] “Damage” means any form of degradation of the properties of the materials which may be linked to the appearance of microcracks, delamination at the different scales of the material or in the case of a part made of composite material of decohesions of fibers with respect to the matrix (resin) or even a breakage of fibers, for example at the level of a fiber, a strand, a weaving plane, several weaving planes or the entire material. In the case of a metallic material, it may be microcracks between the grain boundaries or cracks within a single crystal for example.

[0009] Conventionally, the deformations obtained by stereo-correlation of digital images are then analyzed in the direction of mechanical stress in order to deduce limit values ​​of deformations on the surface of the part. However, the results obtained today are subject to interpretation because no indicator makes it possible to differentiate the end of a so-called "healthy" zone of the part from the transition to a "damaged" state.

[0010] The digital image stereo-correlation technique can also be used during a dynamic type mechanical test.

[0011] A dynamic test is typically a fatigue or impact test. During these dynamic tests, it is also necessary to determine the conditions under which damage occurs. For example, during a fatigue test, digital image stereo-correlation is used to estimate the Wohler limit. The approach then consists of estimating the area of ​​the fatigue hysteresis during a dynamic cycle.

[0012] An objective of the invention is to propose a method allowing improved detection of damage to a part subjected to mechanical stress. Summary of the invention

[0013] A method is thus proposed for determining the damage to a part subjected to mechanical stress, comprising the following steps:

[0014] a) determining the displacements of said part according to the different directions x, y, z of space over a determined area thereof, by stereo-correlation of images,

[0015] b) calculating the deformations of said part according to the different directions of space x, y, z over said determined area, from the displacements determined in step a),

[0016] c) performing a classification of the data obtained at the end of step b), in supervised mode, on at least two classes,

[0017] d) analyzing the data thus classified in step c) to determine whether the part is damaged.

[0018] Thus, thanks to the method according to the invention, damage resulting from mechanical stress on a part can be studied on the basis of a criterion defined over the entire area considered, in this case a criterion relating to a supervised classification.

[0019] The method according to the invention may comprise one or more of the steps below, taken in isolation from one another or in combination with one another: - repeating steps a), b) and c) at least a plurality of times to define a dynamic impact test; - step c) is carried out on at least three classes; - step c) is carried out on at least four classes; - step c) is carried out by k-neighbor analysis or by trees of classification; - the part is a composite part comprising a resin in which fibers are embedded; - the part is a turbomachine part, in particular a turbomachine casing and preferably a turbomachine fan casing.

[0020] The invention also relates to a device for implementing a method according to the invention, comprising:

[0021] - means for determining movements of the part by stereo-correlation images on said determined area, and

[0022] - means for implementing by computer at least step c).

[0023] The device according to the invention may provide that the means for implementing by computer at least step c), also implement step d).

[0024] The invention also relates to an aircraft comprising at least one device according to the invention. Brief description of the figures

[0025] The invention will be better understood with the aid of the following description, given solely by way of example and with reference to the appended drawings in which:

[0026] [Fig.l] is a schematic view of a test bench comprising a mechanical part, cameras and a device for implementing the method for determining damage to a part according to the invention;

[0027] [Fig.2] represents more precisely the surface of the mechanical part, in particular a mechanical part made of woven composite material;

[0028] [Fig. 3] is a schematic view representing the different stages of the method according to the invention;

[0029] [Fig.4] represents the results of a two-class supervised classification, the first class (0 or the darkest) corresponding to the unsolicited zone mechanically and the second class (1 or the lightest) corresponding to a mechanically stressed area;

[0030] [Fig.5] represents the results of a supervised three-class classification, this classification also making it possible to identify more precisely the type of mechanical stress, namely a zone stressed in tension (dark gray zone) in the center of the figure and zones stressed in compression above and above the zone in tension (darker zones), the rest being not mechanically stressed (lighter zone);

[0031] [Fig.6] represents the results of a four-class supervised classification, the addition of an additional class allowing, in addition to the unstressed zones (the darkest), the zones in tension (in the center, dark gray) and the zones in compression (to the right and left of the zone in tension, the lightest), the damaged zones (in the center of the zone in tension). Detailed description of the invention

[0032] [Fig.l] represents a test bench D of a mechanical part 10 comprising cameras C for the acquisition of digital images by stereo-correlation (allows the measurement of displacements), and a device 20 provided with computer means for implementing the method for determining damage to the mechanical part as described below.

[0033] A speckle M is, in addition, affixed to the surface of the mechanical part 10 to facilitate tracking by the cameras C.

[0034] The C cameras for image acquisition are CCD sensor cameras with an acquisition frequency of a few images per second. The acquisition area is illuminated by light-emitting diode lamps. Furthermore, the illumination can be carried out continuously or in synchronization with the C cameras.

[0035] In the following, the embodiments are more particularly concerned with parts made of composite materials, and in particular with 3D woven composite parts. 3D woven composite parts are typically used to produce certain turbomachine parts, in particular turbomachine casings and more particularly turbomachine fan casings.

[0036] However, these parts can also be parts made of 2D woven composite materials, composite parts with unidirectional fibers or even metal parts.

[0037] [Fig. 2] shows a part 10 made of 3D woven composite material. The composite material comprises a fabric formed of weft fibers 11 and warp fibers 12 woven together on a weaving plane P, or several weaving planes superimposed along a thickness direction. The weft fibers 11 and warp fibers 12 may for example be carbon fibers, glass fibers or Kevlar® fibers, i.e. poly(p-phenyleneterephthalamide). Other types of fibers may be considered. Alternatively, the weft 11 and warp 12 fibers may be a combination of at least two fibers chosen from carbon fibers, glass fibers, Kevlar® fibers or others. These fibers are embedded in a matrix 13 which comprises one or more polymers capable of stiffening so as to form the final mechanical part 10. The matrix may for example be chosen from a thermosetting resin or a thermoplastic resin.

[0038] More complex woven structures can also be used in these tests. For example, the warp fibers 12 are no longer limited to a weaving plane P but cross the planes of the material.

[0039] In practice, the part 10 may be a mechanical part of an aircraft. In particular, this part may be a turbomachine casing and for example a casing of a turbomachine fan.

[0040] During a mechanical test, the mechanical part 10 is subjected to stresses which can cause deformations of the part 10, or even damage.

[0041] The method according to the invention therefore aims to determine the damage to a part subjected to mechanical stress, said method comprising the following steps:

[0042] a) determining the movements of said part according to the different directions x, y, z of space over a determined area thereof, by stereo-correlation of images,

[0043] b) implementing by computer a classification of the data obtained at the end of step a), in supervised mode, on at least two classes,

[0044] c) analyzing the data thus classified in step b) to determine whether the part is damaged.

[0045] The different stages of this method 100 are shown diagrammatically in [Fig.3].

[0046] Step a), referenced 101 in [Fig. 3], makes it possible to obtain the movements by stereo-correlation of digital images. This technique is well known to those skilled in the art.

[0047] On the type of part likely to be considered in the tests (plate), we can take into consideration Ux (the displacement along the x axis) and Uy (the displacement along the y axis). This is only an example, but we will keep it to illustrate the different calculations likely to be carried out.

[0048] Step b) implemented by computer, referenced 102 in [Fig. 3], aims to calculate the deformations from the displacements determined in step a). This is obtained, in general, with the following formula:

[0049] [Math. 1] _ j. \(Eqtl) eij - 2 x ( du . + dU; /

[0050] where Ey are the deformations (or more precisely the coefficients of the deformation tensor), with i, j integers such that l <i<3etl<j<3 selon les trois directions (orthogonales) x, y, z de l’espace, et U; ou Uj sont les champs de déplacements associés qui ont été déterminés à l’étape a).

[0051] According to our example, on the type of part likely to be considered in the tests (plate), the strain tensor may be limited to 3 components, namely: exx (strain along the x axis), eyy (strain along the y axis) and yxy (shear defined along the z axis), or even to only 2 components: exx (strain along the x axis), eyy (strain along the y axis) sometimes.

[0052] These calculations can be carried out by the software provided with the means for implementing the test or can be completely carried out (recalculation) by software capable of also implementing all the steps of the method according to the invention.

[0053] Step c) implemented by computer, referenced 103 in [Fig.3], consists of carrying out a classification of the data obtained at the end of step b), in supervised mode, on at least two classes.

[0054] The supervised mode allows the classes to be defined as desired depending on the information sought. In unsupervised mode, it can indeed be difficult to provide, during step d), a correct mechanical analysis of the tested part.

[0055] For example, two classes can be defined, namely a first “undamaged” class and a second “damaged” class. An “undamaged” class is a class in which the data can be analyzed, for example by the equation (Eqtl) in the case where the tested part is a plate. All non-analyzable data are then classified in the “damaged” class. We can refer to [Fig.4].

[0056] Alternatively, step c) implemented by computer may consist of carrying out a classification of the data obtained at the end of step b), in supervised mode, on at least three classes.

[0057] In particular, three classes can be defined, for example a first “unstressed” class (which means that the part does not undergo any deformation), a second “stressed” class (which means that the part undergoes deformation, its nature, traction, compression, shear not being distinguished however) and a third “damaged” class. This amounts to separating the “undamaged” class into an “unstressed” subclass and a “stressed” subclass. We can refer to [Fig.5].

[0058] According to another variant, step c) implemented by computer may consist of carrying out a classification of the data obtained at the end of step b), in supervised mode, on at least four classes.

[0059] In particular, four classes can be defined, for example a first class “unstressed”, a second class “stressed in compression”, a third class “stressed in tension” and a fourth class “damaged”. Here, this makes it possible to obtain additional information on the nature of the stress: compression or tension. Reference can be made to [Fig.6].

[0060] According to yet another variant, step c) implemented by computer may consist of carrying out a classification of the data obtained at the end of step b), in supervised mode, on at least five classes.

[0061] In particular, five classes can be defined, for example a first class “unstressed”, a second class “stressed in compression”, a third class “stressed in tension”, a fourth class “stressed in shear” and a fifth class “damaged”. Here, this makes it possible to obtain additional information relating to the nature of the stress: shear.

[0062] Thus, if three classes are defined, it will be possible to alternatively provide a first “not stressed in tension” (which covers cases where the part does not undergo any deformation but also cases where the part undergoes deformation, for example compression or shear), a second “stressed in tension” class and a third “damaged” class.

[0063] Step c) can be carried out by k-neighbor analysis or by classification trees.

[0064] It is then a question, during step d), of analyzing the data available at the end of the classification to conclude on possible damage to the part.

[0065] Step d) can in particular be implemented by computer.

[0066] Furthermore, a dynamic impact test can be carried out on the part to be tested. For this purpose, steps a), b) and c) are repeated at least a plurality of times. The analysis can then be carried out, during step d), of the temporal evolution of the different zones of the part as a function of the evolution of their classification.

[0067] We can then compare the last classification image obtained with the previous one and make a comparison between the two. Such a comparison then makes it possible to see if there is a discontinuity in the data (damage).

[0068] The invention also proposes a device D for implementing the method according to the invention, comprising:

[0069] - means C for determining movements of the part by stereo-correlation images on said determined area, and

[0070] - means 20 for implementing by computer at least step c).

[0071] Advantageously, said means 20 also make it possible to implement step d) by computer.

[0072] The invention also proposes an aircraft comprising at least one device D for implementing the method 100, as described previously. The device D is then mounted on at least one part to be examined in the aircraft. In this way, the material health and / or damage of the part(s) to be examined can be examined directly during a flight of the aircraft in order to detect, from the first moment of their possible appearance, deformations, cracks and / or damage and to take appropriate decisions.

Claims

Claims

1. Method (100) for determining the damage to a part subjected to mechanical stress, said method comprising the following steps: a) determining (101) the displacements of said part according to the different directions x, y, z of space over a determined area thereof, by stereo-correlation of images, b) calculating (102) the deformations of said part according to the different directions of space x, y, z over said determined area, from the displacements determined in step a), c) implementing by computer (103) a classification of the data obtained at the end of step b), in supervised mode, on at least two classes, d) analyzing (104) the data thus classified in step c) to determine whether the part is damaged.

2. The method (100) of claim 1, wherein steps a), b) and c) are repeated at least a plurality of times to define a dynamic impact test.

3. Method according to one of the preceding claims, in which step c) is carried out on at least three classes.

4. Method according to one of the preceding claims, in which step c) is carried out on at least four classes.

5. Method according to one of the preceding claims, in which step c) is carried out by a k-neighbor analysis or by classification trees.

6. A method (100) according to any preceding claim, wherein the part (10) is a composite part comprising a resin in which fibers are embedded.

7. Method (100) according to any one of the preceding claims, wherein the part (10) is a turbomachine part, in particular a turbomachine casing and preferably a turbomachine fan casing.

8. Device (D) for implementing a method (100) according to one of the preceding claims, comprising: - means (C) for determining movements of the part by stereo-correlation of images on said determined zone, and 10 - means (20) for implementing by computer at least step c).

9. Device according to the preceding claim, in which the means (20) for implementing by computer at least step c) also implement step d).

10. Aircraft comprising at least one device (D) according to claim 8 or 9.

Images