Method for determining damage to a part subjected to mechanical stress and device for implementing such a method

Abstract

The invention relates to a method (1) for determining damage to a part subjected to mechanical stress based on input data representative of the deformation fields according to formula (I) that are obtained by stereo-correlation of images of a zone (S1) to be observed at different times k preceding and following the mechanical stress. The input data are then transformed to obtain two-dimensional output data. These data are then put into the form of a series of images Ik representative of the deformation fields that each comprise a useful zone ZUk. An area Ak of each of the useful zones ZUk is then estimated in order to obtain a function f(k) = Ak representative of the evolution of the area as a function of time, and the areas Ak for which f(k) is greater than or equal to a threshold ɸ1 characteristic of damage are defined. The first time ke1 associated with the first damage can also be identified simultaneously.

Description

[0001] DESCRIPTION

[0002] TITLE: METHOD FOR DETERMINING THE DAMAGE TO A PART SUBJECTED TO MECHANICAL STRESS AND DEVICE FOR IMPLEMENTING SUCH A METHOD

[0003] Technical field of the invention

[0004] The invention relates to the technical field of methods for determining damage to a part subjected to mechanical stress, in particular those based on the analysis of digital images obtained by stereo-correlation (S-CIN, Stereo-Digital Image Correlation (Stereo-D IC)).

[0005] Technical background

[0006] There are many known techniques for studying damage in a material. "Damage" refers to any form of degradation of the properties of a material that may be linked to a change in the internal structure, such as the appearance of microcracks, fiber / resin debonding, fiber breakage, or delamination at different scales of the material. These scales may be those of a carbon fiber, a carbon strand (several thousand fibers), a weave plane, several weave planes, or the entire material. In the case of a metallic material, it may be microcracks between grain boundaries or cracks within a single crystal.

[0007] Digital Image Correlation (DIC) is a non-destructive testing technique that uses digital image correlation to measure displacement vectors on the surface of a part. This method is based on the acquisition by a single digital sensor, such as a camera, of an image of a part in its reference state, this image being called the reference image, and an image of this part in its deformed state, this image being called the "deformed state image". It is therefore intrinsically intended for the two-dimensional measurement of surfaces. By applying the principle of conservation of the gray level, it is possible to determine at each point of the analyzed surface the displacement vector which ensures the concordance between a point of the reference image, and this same translated point of the displacement in the image in the deformed state.

[0008] Xiangyun R. et al., Strain distribution and fatigue life estimation for steel plate weld joint low cycle fatigue based on DIC, Optics and Lasers in Engineering, Volume 124, 2020, uses Digital Image Correlation to estimate the Wôhler limit of a metallic material by analyzing the fatigue hysteresis area during a dynamic cycle. Filipe R., et al., Tension-tension fatigue behavior of hybrid glass / carbon and carbon / carbon composites, International Journal of Fatigue, Volume 146, 2021, presents a CIN fatigue study for composite materials.

[0009] Infrared thermography is also widely used for surface analysis. This non-destructive surface testing technique uses an infrared camera, placed remotely in front of the part, to record the thermal scene. It allows the measurement of the temperature dissipated by the structure during a self-heating test. However, the use of an obscuring insulating curtain may be necessary to isolate the thermal scene and avoid external radiation disturbances that can alter the temperature measurement.

[0010] Infrared thermography can also be implemented at high acquisition frequency to identify the onset of damage in a dynamic impact test. This is for example described in the document Carosena M. et al, A quantitative approach to retrieve delamination extension from thermal images recorded during impact tests, NDT & E International, Volume 100, 2018, Pages 142-152. The choice of post-impact images depends on the parameters affecting the thermal effects associated with the impact, such as the material type, geometry and impact energy. This choice can be dictated by empirical considerations or, as is the case in this document, by the measurement of the heat diffusion time in the material.

[0011] However, a common point to all the techniques previously mentioned is that all surface control parameters are limited to the pixel scale (thermal or visual). This significantly limits the detection of the onset of damage and the monitoring of its propagation in the part under study. Moreover, as can be seen with infrared thermography, the same indicator cannot be used to characterize and monitor the propagation of surface damage for both a dynamic fatigue test and a dynamic impact test.

[0012] Stereo image correlation is a non-destructive testing technique that allows the measurement of displacement fields on the surface of the part during a static or dynamic mechanical test. Knowing the displacement fields, it is then possible to determine the deformation fields on the surface of the part. It is therefore increasingly used in experimental mechanics. Traditionally, a stereo image correlation system includes at least two digital sensors (usually cameras) each acquiring images of the part from different viewpoints. This makes it possible to measure surfaces in three-dimensional space, unlike conventional digital image correlation. Each digital sensor acquires an image of the part in its reference state. Then, after the part has been subjected to mechanical stress, each sensor acquires one or more images of this part in its successive deformed states.

[0013] Different approaches have been developed to obtain reliable data on measured displacements. They involve either camera calibration or calibration of the camera model directly on the part, and measurement of the part's shape. In practice, an area located on the surface of this part, called the area of ​​interest, is studied. In local methods, calibration is performed by determining the displacement of a given point in the area of ​​interest between the reference image and the deformed image in a limited set of pixels centered around this point. This limited set of pixels is called a subset, or correlation window. Calibration is performed by finding the position of the image in the deformed image. If there are several deformed images, calibration must therefore in principle be implemented for each of them.However, despite this calibration, it remains difficult to obtain reliable displacement measurements because calibration and shape measurement remain complicated to implement. Currently, some algorithms make it possible to obtain deformation fields by knowing the size of the image only used.

[0014] The parameters traditionally used in the analysis of displacement data are based on the analysis of deformations in the direction of mechanical stress or, for multiaxial stresses, in the stress directions. They allow the deduction of limit values ​​of deformations at the surface of the part. However, for the reasons mentioned above, these results are still subject to interpretation since no current indicator allows a clear differentiation by an objective criterion of the moment at which the materials pass from a "healthy" state to a "damaged" state. It is therefore not possible to arrive at a decision on the conditions of appearance of damage.

[0015] Moreover, even knowing that damage is present, it is sometimes difficult to clearly identify the direction favored by the stress (fiber direction for a composite or for metallic necking). However, the appearance of damage is a crucial event during the design of a structure in dynamic fatigue, particularly with the estimation of the Wôhler limit. Similarly, during a dynamic impact test, knowledge of the instant of appearance of damage on the surface of the part and the associated deformation field provides information necessary for the design of structures for impact. The document Wu Rong et al. “Health monitoring of wing turbine blades in operation using three-dimensional digital image correlation”, Mechanical system and signal processing, vol. 130, May 22, 2019, 470-483 discloses a method for determining the damage of a part subjected to mechanical stress.

[0016] The invention aims to overcome at least some of the aforementioned problems and proposes to use a single indicator to study the propagation of damage in the context of a dynamic fatigue test as well as in that of a dynamic impact. This indicator makes it possible to differentiate by an objective criterion the transition of materials from a "healthy" state to a "damaged" state so that at the end of the method according to the invention, the test data relating to the appearance or not of damage during the test can be directly validated.

[0017] Summary of the invention

[0018] The invention proposes for this purpose a method for determining the damage of a part subjected to mechanical stress, said part being defined in a three-dimensional space having orthogonal axes X, Y, Z of direction vectors X, Y and Z, the method comprising the following steps, implemented by computer, in this order: 10) providing input data representative of the deformation fields along X, Y and Z obtained by stereo-correlation of images of an area of ​​the part to be observed at different times k (k = [0...M], M integer) preceding and following the mechanical stress, 20) transforming the input data so as to obtain two-dimensional output data and representing these output data in a two-dimensional representation frame of direction vectors XI, ŸÎ so as to obtain a series of images L representative of the deformation fields, each of the images l k including a useful area ZUk,

[0019] 30) estimate an area A k of each of the useful zones ZUk in order to obtain a function f(k) = A k representative of the evolution of the area as a function of time,

[0020] 40) determine the area or areas A k for which f(k) is greater than or equal to a threshold 4)! characteristic of damage and determine a first instant k ei for which f (k ei ) is greater than or equal to said first instant k ei being associated with initial damage.

[0021] The method according to the invention performs a transformation of the input data representative of the deformation fields along X, Y and Z obtained by stereo-image correlation (S-CIN), then represents them in the two-dimensional representation frame in order to obtain a series of images l k representative of the deformation fields. It is then possible to estimate the area A kcovered by the distribution of points (pixels) in the useful area ZUk of each of the images thus obtained and, thus, possible to analyze the evolution of this area over time to follow the propagation of the damage from its appearance.

[0022] Estimating area A kof the distribution of points in the useful area makes it possible to differentiate by an objective criterion the passage of materials from a "healthy" state to a "damaged" state. Indeed, the calculation of the area is based on the two-dimensional representation of the data representative of the deformation fields. Thus, rather than fixing a criterion with the rules of the art, it is possible to directly observe with the estimation of the area a significant change in the mechanical behavior of the part. This also makes the method according to the invention a repeatable and suitable method for studying the appearance and propagation of damage in the context of a dynamic fatigue test as well as in that of a dynamic impact.

[0023] According to different characteristics of the invention which may be taken together or separately:

[0024] - the method further comprises, following step 40), the following steps:

[0025] 50) analyze the deformation fields associated with the instants k < k ei preceding the first damage and determine reference parameters representative of the deformation fields before the first damage,

[0026] 60) compare the reference parameters with the deformation fields associated with the instants k > k ei following the first damage in order to determine a physical quantity characteristic of the damage to the part;

[0027] - the method further comprises, following step 50) or step 60), the following step: 70) comparing the reference parameters with pre-existing parameters representative of a material from which the part is made;

[0028] - the mechanical stress is carried out during a dynamic fatigue test, the physical quantity to be determined during step 60) being used to characterize the dynamic fatigue of the material;

[0029] - step 40) includes the following sub-steps:

[0030] 42) determine a maximum f(k) = A k , max of the function f(k) = A xy , k ,

[0031] 44) define the damage threshold cj'i as a function of the maximum f ma x(k) determined in the previous step,

[0032] 46) determine the first instant k ei for which f(k e (i) is greater than or equal to the damage threshold cfi;

[0033] - the “damage threshold” defined in step 44) represents between 20% and 60% of A k m ax - step 60) includes the following substeps:

[0034] 62) analyze the image l kei deformation fields associated with the instant k = k ei of the first damage in order to determine the stress, the constraint, a displacement or a deformation of the material,

[0035] 64) estimate a threshold <j>2 characteristic of an endurance limit from area A kei associated with said first instant k ei ;

[0036] - the mechanical stress is carried out during a dynamic impact test, the physical quantity to be determined during step 60) being used to characterize the dynamic impact on the material;

[0037] - step 40) includes the following sub-steps:

[0038] 42') identify the area A, associated with an instant k = k, (k, < k e ) of an impact,

[0039] 44') set the threshold damage, said threshold damage corresponding to an area A ke i, A kei > A,, whose value differs by at least 25% from A,,

[0040] 46') identify the first instant k ei for which f(k e i) = 4>i !

[0041] - step 60) includes the following sub-steps:

[0042] 62') analyze the image l kei deformation fields associated with the instant k = k ei of the first damage,

[0043] 64') estimate a threshold of “damage characteristic of an energy limit or dynamic load or dynamic deformation of the material;

[0044] - step 20) comprises a first sub-step 22) consisting of carrying out a linear regression of the deformation fields;

[0045] - which step 20) comprises a first sub-step 22) consisting of carrying out a principal component analysis of the deformation fields;

[0046] - the part is a casing of a fan of a turbomachine;

[0047] - the useful area ZU k of each image being the area of ​​the image over which the data represented in the representation frame R1 extends.

[0048] The invention further relates to a device for implementing a method as previously described, comprising means for implementing by computer at least steps 20) to 40).

[0049] Brief description of the figures

[0050] Other objects, characteristics and advantages of the invention will appear more clearly in the description which follows, made with reference to the appended figures, in which: - figure 1a is a schematic view illustrating the different stages of a method according to the invention,

[0051] - figure 1 b is a schematic view illustrating the different stages of a method according to a first embodiment of the invention,

[0052] - figure 1 c is a schematic view illustrating the different stages of a method according to a second embodiment of the invention,

[0053] - figure 2 illustrates a part whose damage is to be studied using the method according to the invention;

[0054] - Figure 3a illustrates examples of possible speckles for performing fatigue and dynamic impact tests by S-CIN on a solid part;

[0055] - Figure 3b illustrates examples of possible speckles for carrying out fatigue and dynamic impact tests by S-CIN on a part with small dimensions;

[0056] - Figures 4a, 4b and 4c illustrate the images l k obtained during the implementation of the method according to the invention, the images l k having been obtained from data representative of the deformation fields at different times during a dynamic fatigue test;

[0057] - Figure 5 illustrates a curve representing the evolution of the function f(k) = A k as a function of time for the same tests as those in Figures 4a, 4b and 4c: the horizontal dotted line indicates the position of the “damage” threshold while the vertical dotted line indicates the time of the start of the fatigue test;

[0058] - Figures 6a, 6b and 6c illustrate the images l k obtained during the implementation of the method according to the invention, the images l k having been obtained from data representative of the deformation fields at different times during a dynamic impact test;

[0059] - Figure 7 illustrates a curve representing the evolution of the function f(k) = A k as a function of time for the same tests as those in figures 6a, 6b, 6c: the horizontal dotted line indicates the position of the damage threshold cfi;

[0060] - Figure 8 illustrates the results of a test carried out on an undamaged part but in the process of deformation for a study by tomography: figure 8, section a) is an image of the part, figures 8b, 8c and 8d are representations of the component, respectively, along an axis (Z), along an axis (Y) and along an axis (X) of the displacement fields of the part;

[0061] - Figure 9 illustrates the results of a test carried out on a damaged part for a tomographic study: Figure 9, section a) illustrates an image of the part, Figures 9, section b, 9, section c and 9, section d are representations of the component, respectively, along an axis (Z), along an axis (Y) and along an axis (X) of the displacement fields of the part.

[0062] Detailed description of the invention

[0063] The invention relates to a method 1 for determining the damage to a part subjected to mechanical stress.

[0064] 1. Definitions - general comments

[0065] Throughout the description of the present application, the term S-CIN designates the technique of Stereo-Correlation of Digital Images. Unless otherwise indicated, the displacement field and deformation field data are all obtained by S-CIN. Although S-CIN can be implemented to carry out measurements in the context of static mechanical tests and dynamic mechanical tests, the method according to the invention is applied to data obtained in the dynamic domain, in particular by dynamic fatigue and dynamic impact tests. We will return to this in the following sections.

[0066] These tests may have the objectives of characterizing the properties of a material or verifying the quality of the material. It is also possible to apply the method according to the invention to data from part resistance validation tests, for example from a part subject to a derogation during its production. In this case, it is possible to extract all or part of the part depending on the need for justification and the area whose acceptability must be validated, then carry out a mechanical test on, respectively, this part or the part of it.

[0067] At this point, it should be emphasized that the tests themselves are not the subject of the invention and only the data relating to the displacement fields and / or the deformation fields obtained from these tests are useful. In other words, the invention is independent of the test phase and everything related thereto. In other words, the invention can be implemented independently of the test phase and the test data collection phase, in particular the data relating to the displacement fields and / or the deformation fields. That being said, the following sections present the experimental conditions under which the tests were carried out in order to facilitate their reproduction by a person skilled in the art.

[0068] In the context of the invention, the S-CIN data are acquired by CCD cameras, namely cameras using photographic sensors based on a charge-coupled device. CCD cameras have an acquisition frequency suitable for fatigue and dynamic impact tests. A CCD camera capable of acquiring a few images per second is suitable for performing dynamic fatigue tests at low frequency but is insufficient for performing dynamic impact tests. In the latter case, a camera capable of acquiring several thousand images per second is preferable. That being said, the method according to the invention, which will be detailed below, aims to enable determination of the damage to a part subjected to mechanical stress in the context of both dynamic fatigue and dynamic impact tests.A CCD camera capable of acquiring several thousand images per second was therefore used, but this is not obligatory, as one can understand.

[0069] The acquisition area was illuminated using light-emitting diode (LED) lamps. The lighting can be performed continuously or in synchronization with the cameras depending on the acquisition device in question. When the lighting is used continuously, this allows the frequency of the LED lamps to be decorrelated from that of the CCD cameras.

[0070] The part on which the test is to be carried out can be any part. This part is not limited by its shape, dimensions, color, etc. This part comprises an external surface S comprising an area to be observed S1. An example of an embodiment is illustrated in Figure 2. A close-up view of damage is visible in the dotted box. The part is defined in a three-dimensional space having orthogonal axes X, Y and Z respectively associated with direction vectors X, Y, and Z. In the following, the X axis is called the longitudinal axis, the Y axis is called the transverse axis and the Z axis is called the orthogonal axis.

[0071] In the context of the invention, the part is preferably made of a woven composite material. For example, the part may be made of a woven continuous carbon fiber composite material. Such a 3D woven composite is for example used in the context of a crankcase fan for an aircraft engine. It comprises a plurality of warp fibers, arranged along a longitudinal axis, which are woven with a weft fiber in one or more weaving planes and embedded in a polymer matrix that has hardened. The weft fibers and the warp fibers may then be made of carbon, glass, or a mixture of carbon and glass. They may also be made of other types of fibers. The invention may be applied to parts made from other types of composite materials.

[0072] Conventionally, S-CIN measurements are implemented by means of a speckle deposited on the surface of the part whose damage propagation is to be studied. Indeed, the measurements of displacement fields and / or deformation fields require that the image acquired by the digital sensor has a random, contrasted texture and dimensions adapted to the part considered. Very few surfaces naturally present the conditions required for the implementation of these measurements. The speckle serves precisely to provide the appropriate texture to the surface of the part. There are different types of speckles. In the context of the invention, a black and white speckle was used to provide the appropriate texture to the part considered.

[0073] Figures 3a and 3b are images showing examples of possible speckles for performing fatigue and dynamic impact tests by S-CIN. In Figure 3a, the speckle was deposited on a massive part, while in Figure 3b, the speckle was deposited on a small part. A "massive part" is defined as a part whose largest dimension is between one meter and several meters. For example, Figure 3a illustrates a part of a composite impact panel. A "small part" is defined as a part whose largest dimension is less than one meter. As an example, Figure 3b illustrates a typical laboratory specimen.

[0074] Prior to the actual test, a calibration step can be implemented. This calibration step can include, on the one hand, the calibration of the CCD camera parameters and, on the other hand, the measurement of the actual shape of the part under consideration. Different approaches exist. For example, the calibration of the CCD camera parameters can be carried out by the gray level minimization approach while the measurement of the part shape can be carried out by applying the gray level conservation principle. The calibration step can be implemented based on these well-known approaches.

[0075] During dynamic impact tests, the part is subjected to a shot. This shot generates a mechanical stress that can cause damage (a definition of the term "damage" is given in the preamble to this description). In the case of dynamic impact with damage, it is possible to visually detect when damage has occurred. The method according to the invention not only makes it possible to monitor the propagation of the damage from its onset, but it also makes it possible to define a threshold from which the material from which the part in question is made is damaged. This threshold is called the "damage" threshold. It should be noted that the method according to the invention can also be applied to the case of dynamic impact without damage.

[0076] In the illustrative examples of the present invention, the dynamic impact tests were implemented on panels or coupons of real parts in order to ensure good mechanical resistance during certification validation and in order to simulate flight conditions such as blade loss, ingestion of birds or debris in flight.

[0077] In dynamic fatigue testing, the part is subjected to a load cyclically in order to estimate the endurance limit of the material from which the part is made. This endurance limit, hereinafter called the threshold <j>2d endurance, can be known from the first cycle of the test, so that it is not mandatory to subject the part to the load for several cycles. In any case, the method according to the invention makes it possible to estimate this endurance limit in a single test instead of a test campaign with several loads as is currently the case in the literature. Like a dynamic impact test, the dynamic fatigue test can result in generating or not damage to the material. Thus, the method according to the invention makes it possible not only to define a threshold <j>2nd endurance of the material, but it also allows the propagation of damage to be monitored as soon as it appears.

[0078] In the illustrative examples of the present invention, the dynamic fatigue tests are carried out in order to characterize the mechanical resistance linked to the living environment of an engine. In the case of a turbomachine part fixed on an aircraft, this concerns, for example, the takeoff, landing, cruise or rapid descent phases.

[0079] 2. Description of the general method of implementation

[0080] With reference to Figure 1a and as already indicated previously, the invention relates to a method 1 for determining the damage of a part subjected to mechanical stress. As also already indicated, this mechanical stress can be carried out within the framework of a fatigue test or a dynamic impact test.

[0081] According to a first step 10) of the method 1 according to the invention, input data representative of the deformation fields along X, Y and Z obtained by stereo-correlation of images of a zone S1 (of the part) to be observed at different times k (k = [0...N], N integer) preceding and following the mechanical stress are provided.

[0082] The displacement fields and / or deformation fields along X, Y and Z obtained by S-CIN are in the three-dimensional domain, and are therefore by nature more precise than point or averaged information such as that obtained with the methods of the prior art. The deformation fields can be obtained in two ways. The deformation fields can be directly measured by a computer program provided for this purpose. In this respect, it is advantageous for the size of the image or facet used for the S-CIN to be entered in the computer program so that it directly calculates the deformation fields. However, this is not mandatory. It is possible to calculate the deformation fields from the displacement fields determined by the computer program.Regardless of how the deformation fields along X, Y and Z are obtained, they constitute the input data for implementing the following steps of method 1 according to the invention.

[0083] In practice, the input data relating to the deformation fields can be provided directly by the S-CIN computer program. If this is not the case, they can be recalculated from the displacement fields necessarily provided with the stereo-correlation data. As a reminder, the deformations are calculated with the following formula:

[0084] This formula is the general case for calculating deformations in the most common cases. If it is necessary to calculate deformations, one approach is to use finite difference schemes. It is advantageous if the schemes used are consistent with the data. If only the deformation data are required, it is possible to limit oneself to a stencil of two or three (number of data points used for the derivation) with centered schemes in the nominal case and decentered upstream or downstream if one is on the edges of the image. These schemes allow the entire image to be preserved and avoid changing the size of the area to be observed.

[0085] In the context of the invention, it is desired not only to be able to identify the appearance of damage to the external surface S of the part but, in addition, to follow the propagation of the damage in the part considered. In this respect, during the test phase, it is important to ensure that the images are acquired at the different times preceding and following the mechanical stress. As we will see below, the input data representative of the deformation fields relating to the times preceding the mechanical stress characterize the zone S1 to be observed while it is healthy. The zone S1 to be observed is said to be “healthy” because the mechanical properties of the material have not yet undergone any alteration due to damage.As for the input data representing the deformation fields relating to the instants following the mechanical stress, they make it possible to characterize the deformations, and any damage that the material has suffered. They are therefore necessary for the analysis of the part in the deformed state.

[0086] However, it should be noted that the input data provided during this first step 10) of the method according to the invention do not necessarily represent all the input data that could be obtained (first configuration) or all the input data that are actually obtained from the S-CIN measurements (second configuration).

[0087] As for the first configuration and as an example, during a dynamic fatigue test, it is possible to limit oneself to the analysis by S-CIN of a series of images acquired over one cycle every 10% of total cycles. In other words, for a test of 100,000 cycles, the analysis can be done at 10,000, 20,000, 30,000 cycles, etc. These values ​​can be modified according to the operator's needs and the computing capacity. In the case of a dynamic impact test, one could in fact limit oneself to one data acquisition before the impact and up to ten data acquisitions after the start of the impact when damage on the external face is visible after the test.

[0088] As for the second configuration, for example, during a dynamic impact test, if no damage is visible on the external face, it may not be necessary to analyze all the data obtained. It may be possible to limit ourselves to those with the highest levels of deformation.

[0089] According to a second step 20) of the method 1 according to the invention, the input data are transformed so as to obtain two-dimensional output data and these output data are represented in a two-dimensional representation frame R1 of direction vectors XI, Ÿï so as to obtain a series of images l k representative of the deformation fields, each of the images k including a useful area ZU k .

[0090] Thus, a transformation of the data relating to the deformation fields can be carried out during a first sub-step 22) of the second step 20). These three-dimensional data provided during the first step 10) are reduced to two-dimensional data. Thus, the first sub-step 22) is nothing more than a reduction of the dimension of the input data. In this respect, different methods of reducing the variables can be implemented. Each of the data according to X, Y and Z of the deformation fields corresponds to a variable and there is a redundancy between all of these variables because these variables are linked to each other (correlated). The aim of this first sub-step 22) is therefore to extract the data whose sum covers all of the information contained in the input data provided during the first step 10).According to a first particular implementation, the first sub-step 22) can be implemented by linear regression applied to the deformation fields. This reduction method is well known and its application to the deformation fields does not pose any particular question. According to a second implementation, the first sub-step 22) can be implemented by the technique of principal component analysis. Principal component analysis is a relatively well-known decomposition method which ensures that each new variable, called principal component, constitutes a vector orthogonal to the system composed of the other principal components. In the context of the invention, the number of principal components is determined by the capacity of the model to represent all of the deformation fields.

[0091] The two variable reduction methods presented above, namely linear regression and principal component analysis, make it possible to obtain output data that objectively represent the input data while allowing a reduction in the number of variables. The first sub-step 22) therefore ultimately aims to obtain the best possible representation of the data relating to the two-dimensional deformation fields.

[0092] In a second sub-step 24) of the second step 20), a representation of the output data is carried out in a two-dimensional frame of reference. A plurality of images l is thus obtained. k characterizing the deformations of the surface at the different instants before and after the mechanical stress. The image Is designates the image associated with the input data acquired at the eighth instant, which means that this instant is not necessarily equal to 8 seconds. This instant can be taken 1.45 milliseconds after the start of the test as one second depending on the camera used and depending on the input data retained by the operator for the purposes of analysis. The second sub-step 24) therefore constitutes a simple representation step. The whole challenge therefore lies in the first sub-step 22) since it is necessary to carry out a qualitative reduction of the input data in order to obtain the best representation of the deformation fields along X, Y and Z.

[0093] Each of the images l k includes a useful area ZU k . The useful area ZUo designates the useful area of ​​the image l0, the useful area ZUi designates the useful area of ​​the image h, etc. This useful area is not necessarily the entire image. This useful area corresponds to the area over which the points (pixels of the image) extend which will be necessary for the implementation of the third step 30) of the method 1 according to the invention, these points being nothing other than the graphical representation of the output data obtained during the second step 20). In other words, the useful area ZU k is the surface of the image l k delimited by the extreme points corresponding to the output data represented in the image. It therefore includes the extreme points (pixels) themselves, but also points (pixels) located between these extreme points, in particular all the points located between these extreme points. Thus, this useful area can be the image l k complete as just a part of the image l k , as appropriate. The identification of the useful area can be carried out by associating a pixel with a value, the absence of a pixel being characterized by the value 0.

[0094] In a third step 30), we estimate the area A k of each of the useful zones ZU k in order to obtain a function f(k) = A k representative of the evolution of the area as a function of time.

[0095] From the two-dimensional output data we obtain by measuring the area A k of each of the useful zones ZU k an indicator giving a photograph of the evolution of the deformations of the external surface as a function of time. Thus, when area A k reaches a threshold value, for example characteristic of damage, it is possible to determine the instant associated with reaching this threshold by identifying the image l k on which this area was measured. However, this area cannot be used as such. Indeed, to be able to estimate whether an area is characteristic of surface damage, it is necessary to determine the threshold from which we consider that there is damage. We will return to this later. In any case, the area constitutes an objective indicator which makes it possible to monitor the propagation of damage on the external surface S of the part in question.

[0096] In a fourth step 40), the area or areas A are determined k for which f(k) is greater than or equal to a threshold “characteristic of damage” and we determine a first instant k ei for which f(k e i) is greater than or equal to ", said first instant k ei being associated with initial damage.

[0097] The threshold "characteristic of damage" (or more simply the "damage threshold") is the area from which damage to the external surface is characterized. In other words, the damage threshold cfi is the area value from which the external surface is damaged, that is to say that it presents damage. A definition of damage was given in the preamble to the description and is not repeated here. Thus, the "damage threshold" is expressed in surface units as the area itself.

[0098] Very advantageously, this damage threshold can be predetermined. Indeed, knowing the nature of the material from which the part is made and having already implemented the method 1 according to the invention at least once, it is possible to determine in advance the damage threshold from which damage to the external surface S has occurred. In this case, the first implementation of the method 1 can be considered as a calibration. However, it is preferable to determine this damage threshold from the data of the test concerned. Indeed, on the one hand this limits the implementation of the method 1 to one use per test and on the other hand this makes it possible to overcome the sliding effects which can be observed after a certain number of tests. In this case, it is appropriate to start the fourth step 40) of the method 1 by determining this threshold. As we will see later, this can be done in two very simple sub-steps 42) and 44).

[0099] Knowing the "damage" threshold, it is sufficient, during a first sub-step or a third sub-step, as the case may be, to compare the values ​​of area A k at the threshold and extract the values ​​greater than or equal to this "damage" threshold. Once the area values ​​A k concerned are extracted, during a second sub-step or a fourth sub-step, an identification of the image l k on which the first area A was measured kei greater than or equal to the “damage threshold” allows us to know the instant kei of the appearance of the damage.

[0100] To the extent that the objective pursued by the operator consists solely of knowing the moment of the appearance of the damage or knowing that there has been damage, the fifth 50) and sixth 60) steps which will be described below do not need to be carried out. The method 1 according to the invention can stop at the end of this fourth step 40). Steps 50) and 60) being optional, they are indicated in the figures by dotted boxes.

[0101] If it is sought to obtain a physical quantity characteristic of the damage, the method 1 according to the invention can advantageously comprise the following steps.

[0102] In a fifth step 50), we analyze the deformation fields associated with the instants k < k ei preceding the first damage and we determine reference parameters representative of the deformation fields before the first damage.

[0103] This fifth sub-step makes it possible to estimate the displacement fields and the deformation fields of the material before mechanical stress. These data therefore make it possible to know the state of the external surface S of the part when it was still healthy. They therefore provide, in a way, reference parameters concerning the material.

[0104] The reference parameters determined during this step depend on the test being implemented. In the case of dynamic fatigue, for example, it will be possible to determine the force, stress, displacement or deformation allowing the endurance limit of the material to be estimated, knowing the instant of appearance of the damage. In this case, it may be useful to check the conditions applied at the instant considered. In the case of a dynamic impact test, the analysis is simplified. In the case of damage on the external face, this means that the local mechanical capacities of the material have been largely exceeded and are translated by a crack. As with a dynamic fatigue test, it is then possible to deduce the energy or the dynamic limit load or the dynamic limit deformation before damage. We will return to this later.

[0105] From then on, it is possible during a sixth step 60) to compare the reference parameters with the deformation fields associated with the instants k > k ei following the first damage in order to determine a physical quantity characteristic of the damage to the part.

[0106] Furthermore, it is also possible, concurrently with the sixth step 60) or after, to carry out a step 70) of comparing the reference parameters with pre-existing parameters representative of a material from which the part is made in order to determine a physical quantity characteristic of the damage to the part.

[0107] These pre-existing parameters representative of the material may have been determined empirically or by theory. The interest of this comparison is simply to verify the qualitative value, and therefore the reliability, of the parameters determined during step 50). This comparison with the external data therefore makes it possible to establish a criterion of consistency of the data obtained by method 1.

[0108] The method according to the invention can be implemented to determine the damage of a variety of composite parts. The composite part can be a turbomachine casing, for example the casing of a turbomachine fan. The composite part can advantageously be made of carbon fibers. The composite part can, alternatively, be made of glass fibers. The composite part can also be a mixture of carbon or glass fibers.

[0109] It should be noted that this method is also applicable with the different possible deformations. With the S-CIN, it is common to obtain the principal deformations of each point of the studied area. It is possible to apply the method 1 according to the invention with the principal deformations 1 and principal deformations 2 instead of those with the deformation fields along X and the deformation fields along Y. These principal deformations can be obtained from an automatic analysis of the Mohr circle of the deformations. The distributions displayed will be visually different but the conclusions will be similar to those detailed above according to the study cases. 3. First embodiment of the invention - Dynamic fatigue

[0110] All comments made previously in relation to the general embodiment apply, except of course where examples relating to the case of dynamic impact are mentioned.

[0111] Referring now to Figure 1 b, in this embodiment, the mechanical stress is carried out during a dynamic fatigue test. The physical quantity that can be determined during step 60) is then used to characterize the dynamic fatigue of the material. In the case of dynamic fatigue, the propagation of the damage is slow and progressive. The function f(k) will change slowly and progressively over time.

[0112] Thus, according to this first embodiment of method 1, step 40) comprises the following sub-steps:

[0113] 42) determine a maximum f(k) = A k ,max of the function f(k) = A xy ,k,

[0114] 44) set the threshold damage as a function of maximum f ma x(k) determined in the previous step,

[0115] 46) determine the first instant k ei for which f(k e (i) is greater than or equal to the “damage” threshold.

[0116] Thus, the "damage threshold" is calculated based on the data relating to the dynamic fatigue test from which the input data were obtained. The maximum f(k) = A k ,max of the function f(k) represents the maximum amplitude of the deformation undergone by the material. It can therefore be used to set the "damage" threshold objectively. Preferably, the "damage" threshold defined in step 44) represents between 20% and 60% of A k , max, i.e. the damage threshold cfi defined in step 44) has a value between 20% and 60% of A k , max- Thus, any area value greater than or equal to a "damage threshold" thus fixed will be associated with damage to the material. In this regard, this allows us to recall that the material can undergo several damages during the test. After having identified the area or areas verifying this condition, all that remains is to identify the image l k associated with this area to know the moment at which the damage(s) occurred.

[0117] Still according to this first embodiment of method 1, step 60) comprises the following sub-steps:

[0118] 62) analyze the image l kei deformation fields associated with the instant k = k ei of the first damage in order to determine the effort, the constraint, a displacement or a deformation of the material, 64) estimate a threshold <j>2 characteristic of an endurance limit from area A kei associated with said first instant k e i.

[0119] Sub-steps 62) and 64) allow the endurance limit of the material to be estimated. To obtain the aforementioned reference parameters, namely the force, the stress, a displacement or a deformation of the material, it is therefore necessary to have previously determined the instant of appearance of the damage, i.e. the instant of the first damage. Therefore, the threshold <j>2nd endurance limit can be estimated. As we saw in the introduction to this description, current methods do not allow this endurance limit to be reliably estimated because the data obtained by these methods are subject to interpretation. Indeed, no indicator exists in the literature to clearly differentiate by an objective criterion the transition of materials from the "healthy" state to the "damaged" state. The method 1 according to the invention allows this endurance limit to be reliably estimated.

[0120] That being said, this does not mean that the endurance limit of the material is higher than the damage to it. Indeed, in most cases, the threshold <j>2nd endurance limit is lower than the "damage" threshold since endurance occurs before damage.

[0121] 4. Second embodiment - Dynamic impact

[0122] All comments made previously in relation to the general method of implementation apply, except of course where examples relating to the case of dynamic fatigue are mentioned.

[0123] Referring now to Figure 1c, in a second embodiment of the invention, the mechanical stress is carried out during a dynamic impact test. Thus, the physical quantity to be determined during step 60) is used to characterize the dynamic impact on the material. In the case of dynamic impact, the propagation of the damage is rapid and sudden. The function f(k) therefore changes in order of magnitude at the time of damage and the instant of damage can therefore be easily identified.

[0124] In this regard, according to this second embodiment of method 1, step 40) comprises the following sub-steps:

[0125] 42') identify the area A, associated with an instant k = k, of an impact,

[0126] 44') set the threshold damage, said threshold damage corresponding to an area A ke i, A kei > A,, whose value differs by at least 25% from A,, 46') identify the first instant k ei for which f(k e i) = 4>i- In this embodiment, the “damage” threshold therefore indicates a change in the order of magnitude of the area A k . It therefore functions as a criterion imaging the damage limit in deformation of the material during a dynamic impact. Preferably, and as indicated above, this change in order of magnitude is evaluated in relation to the area A, associated with the instant k = k, (k, < k e ) of the impact, which makes it possible to have as a reference area an area corresponding to a "healthy" state of the external surface S of the object. In the hypothesis posed here, we consider that the surface is damaged when the value differs by at least 25% compared to A,. This in fact indicates a change in the order of magnitude of the area and therefore a significant increase in the area. This cannot be due to chance. This sudden increase has no physical meaning and necessarily corresponds to damage to the external surface.

[0127] From then on, it is possible to precisely determine the appearance of the first damage by identifying the instant associated with the image l k on which this area A was measured ei . Area A ei and the threshold damage are therefore combined. As mentioned above, this threshold therefore functions as a criterion for assessing the damage limit in deformation of the material during a dynamic impact.

[0128] Still according to this second embodiment of method 1, step 60) comprises the following sub-steps:

[0129] 62') analyze the image l kei deformation fields associated with the instant k = k ei of the first damage,

[0130] 64') estimate a threshold 4>2' of damage characteristic of an energy limit or dynamic load or dynamic deformation of the material.

[0131] Sub-steps 62') and 64') make it possible to estimate the energy or dynamic load of the material. To obtain these aforementioned reference parameters, it is therefore necessary to have previously determined the time of appearance of the damage, i.e. the time of the first damage. From then on, the damage threshold 4>2' characteristic of an energy or dynamic load limit or dynamic deformation of the material can be reliably estimated.

[0132] 5. Examples of implementation of method 1 according to the invention

[0133] For the purposes of dynamic fatigue and dynamic impact testing, a woven composite part was used.

[0134] 5.1 Example 1 - Dynamic Fatigue In this case of dynamic fatigue, the test was carried out progressively. As the study material is a woven composite, damage likely to cause a significant change in mechanical behavior is visible by S-CIN. When the stress conditions change little (or not at all) the first damage is highly likely to be taken up by other elements of the material (other fibers or elements of the weave). Thus, the damage will appear in an image l k to another with jumps in area, with a general tendency of increase.

[0135] Figures 4a, 4b and 4c illustrate the images l k obtained after implementing the method according to the invention, the images l k having been obtained from representative data of the deformation fields at different times during a dynamic fatigue test. The evolution of the distribution at different times can be discerned. Here, only the images l k of 9 different instants are illustrated. Figure 5 illustrates a curve representing the evolution of the function f(k) = A k as a function of time for the same tests as those in Figures 4a to 4c: the horizontal dotted line indicates the position of the damage threshold cfi while the vertical dotted line indicates the time of the start of the fatigue test.

[0136] In the first image h (time k = 1.26 ms), all the points are concentrated. There are no deformations identified. The useful zone ZUi therefore has a small area (Figure 5, first point). During the dynamic fatigue test, an increase in the area of ​​the useful zone is observed. Here, the mechanical stress propagates to the panel. When the damage begins, the useful zone begins to expand. Isolated points or groups of isolated points are visible. This is clearly the case from time k = 1.34 ms which is associated with l4. The area A4 of the distribution increases very significantly, indicating the appearance of damage. In Figure 5, this area clearly exceeds the damage threshold value cfi and is, rightly, associated with the first damage.It is therefore possible to identify three zones of mechanical behavior: 1) the start of the test associated with mechanical loading, 2) a healthy zone corresponding to the minimum values ​​once the load is applied and 3) a damage zone whose start can be identified by the first peak area.

[0137] Therefore, knowing the instant of appearance of the damage, it is sufficient to check the conditions applied at this precise instant during the test to deduce the force, the constraint, the displacement or the deformation allowing the endurance limit of the material to be estimated.

[0138] 5. 2 Example 2 - Dynamic impact In the case of a dynamic impact, the analysis is simplified. In the case of damage to the external face, this means that the local mechanical capacities of the material have been largely exceeded and are translated by a crack. The area of ​​deformations will increase significantly, as seen previously.

[0139] Figures 6a, 6b and 6c illustrate the images l k obtained after implementing the method according to the invention, the images l k having been obtained from representative data of the deformation fields at different times during a dynamic impact test. The evolution of the distribution at different times can be discerned. As in the previous example, only the images l k of 9 different instants are illustrated. Figure 7 illustrates a curve representing the evolution of the function f(k) = A k as a function of time for the same tests as those in Figures 6a to 6c: the horizontal dotted line indicates the position of the “damage” threshold.

[0140] In the first image (time k = 1.32 ms), all the points are concentrated. There are no deformations identified. The useful zone ZUi therefore has a small area (Figure 7, first point). This is still the case for images I2, h and l4. During the dynamic impact test, an increase in the area of ​​the useful zone is observed. This increase is significant and sudden as expected. Isolated points or groups of isolated points are visible. This is clearly the case from time k = 1.42 ms which is associated with l5. The area A5 of the distribution increases very significantly, indicating the appearance of damage. In Figure 7, this area clearly exceeds the damage threshold value and is, rightly, associated with the first damage.

[0141] As with a fatigue test, the external conditions at that moment can be identified to deduce the energy or the dynamic limit load or the dynamic limit deformation before damage during the test.

[0142] The invention finally relates to a device for implementing a method 1 as described previously. This device comprises means for implementing by computer at least steps 20) to 40). These means are for example a processor and a memory.

[0143] The device may also comprise means for measuring movements of the part by stereo-correlation of images of at least one zone S1 of the part to be observed as defined in section “1. Definitions - general comments” of the description. Such means are advantageously:

[0144] - a part comprising an external surface S comprising an area to be observed S1, and preferably a speckle deposited on the surface of the part, - at least two CCD cameras capable of acquiring a few images per second, and more preferably several thousand images per second, to acquire the S-CIN data,

[0145] - several light-emitting diode lamps (LED lamps) to illuminate the acquisition area, - optionally, a means of generating a shot and therefore mechanical stress on the part.< / j> < / j> < / j> < / j> < / j> < / j>

Claims

CLAIMS 1. Method (1) for determining the damage to a part subjected to mechanical stress, said part being defined in a three-dimensional space having orthogonal axes (X, Y, Z) of direction vectors X, Y and Z, the method comprising the following steps, implemented by computer, in this order: 10) providing input data representative of the deformation fields along X, Y and Z obtained by stereo-correlation of images of a zone (S1) of the part to be observed at different times k (k = [0...M], M integer) preceding and following the mechanical stress, 20) transform the input data so as to obtain two-dimensional output data and represent this output data in a two-dimensional representation frame (R1) of direction vectors XÏ, Ÿî so as to obtain a series of images l k representative of the deformation fields, each of the images kincluding a useful area ZU k , 30) estimate an area A k of each of the useful zones ZU k in order to obtain a function f(k) = A k representative of the evolution of the area as a function of time, 40) determine the area or areas A k for which f(k) is greater than or equal to a threshold “characteristic of damage and determine a first instant k ei for which f(k e i) is greater than or equal to ", said first instant k ei being associated with initial damage.

2. Method (1) according to claim 1, further comprising, following step 40), the following steps: 50) analyze the deformation fields associated with the instants k < k ei preceding the first damage and determine reference parameters representative of the deformation fields before the first damage, 60) compare the reference parameters with the deformation fields associated with the instants k > k ei following the first damage in order to determine a physical quantity characteristic of the damage to the object.

3. Method (1) according to claim 2, further comprising, following step 50) or step 60), the following step: 70) compare the reference parameters with pre-existing parameters representative of a material from which the part is made.

4. Method according to any one of claims 2 to 3, in which the mechanical stress is carried out during a dynamic fatigue test, the magnitude physical to be determined during step 60) used to characterize the dynamic fatigue of the material.

5. Method (1) of tracking according to claim 4, in which step 40) comprises the following sub-steps: 42) determine a maximum f(k) = A k, max of the function f(k) = A xy , k , 44) define the damage threshold cj'i as a function of the maximum f ma x(k) determined in the previous step, 46) determine the first instant k ei for which f(k e (i) is greater than or equal to the “damage” threshold.

6. Method (1) according to claim 5, in which the “damage threshold” defined during step 44) represents between 20% and 60% of A k , max- 7. Method (1) according to any one of claims 4 to 6, in which step 60) comprises the following sub-steps: 62) analyze the image l kei deformation fields associated with the instant k = k ei of the first damage in order to determine the stress, the constraint, a displacement or a deformation of the material, 64) estimate a threshold <j>2 characteristic of an endurance limit from area A kei associated with said first instant k e i.

8. Method (1) according to any one of claims 2 to 3, in which the mechanical stress is carried out during a dynamic impact test, the physical quantity to be determined during step 60) serving to characterize the dynamic impact on the material.

9. Method (1) according to claim 8, wherein step 40) comprises the following sub-steps: 42') identify the area A, associated with an instant k = k, (k, < k e ) of an impact, 44') define the “damage” threshold, said “damage” threshold corresponding to an area A ke i, A kei > A,, whose value differs by at least 25% from A,, 46') identify the first instant k ei for which f(k e i) = 4>i- 10. Method (1) according to any one of claims 8 to 9, in which step 60) comprises the following sub-steps: 62') analyze the image l kei deformation fields associated with the instant k = k ei of the first damage, 64') estimate a threshold <j>2' of damage characteristic of an energy limit or dynamic load or dynamic deformation of the material.

11. Device for implementing a method (1) according to one of the preceding claims, comprising means for implementing by computer at least steps 20) to 40).< / j> < / j>

Images