Interface in-situ failure analysis method based on deep optical flow and physical constraint

Abstract

This invention relates to the field of material interface reliability evaluation technology, and provides an in-situ interface failure analysis method based on depth optical flow and physical constraints. The method includes: acquiring sequential images of the metal-polymer joint during SEM shear testing; performing preprocessing, spatial calibration, and rigid body drift correction; identifying the interface and constructing a geometry-aware mask; reconstructing a pixel-level dense displacement field using depth optical flow to obtain the physical displacement and strain fields; extracting local kinematic indices based on the interface geometric reference, and calculating local matrix energy storage and local interface dissipation indices in conjunction with material and interface physical parameters; and finally constructing a normalized energy competition factor to determine the failure mode tendency of local regions. This invention solves the problem of existing technologies' difficulty in quantitatively characterizing interface damage evolution, achieving a quantitative conversion from in-situ images to failure mechanism determination, and possessing high precision, strong robustness, and clear physical significance.

Description

Technical Field

[0001] This invention belongs to the field of material interface reliability evaluation technology, and particularly relates to an in-situ interface failure analysis method based on deep optical flow and physical constraints. Background Technology

[0002] Metal-polymer hybrid structures combine the high strength and stiffness of metals with the lightweight, corrosion resistance, and ease of forming of polymers, and have been widely used in automotive lightweighting, aerospace, rail transportation, and electronic packaging. However, due to significant differences in physical properties such as elastic modulus and coefficient of thermal expansion between metals and polymers, their interfaces are prone to damage phenomena during service, including stress concentration, localized slippage, interface debonding, and crack initiation and propagation. Accurately identifying the interface damage evolution process is crucial for evaluating joint reliability and optimizing structural design.

[0003] Currently, research on metal-polymer lap joints mainly relies on macroscopic mechanical testing, fracture surface analysis, SEM observation, and digital image correlation (DIC). These methods can evaluate the overall performance or final failure morphology of the joint, but they are difficult to reveal the initiation, propagation, and evolution of damage during loading. In-situ SEM testing can continuously acquire high-resolution images of the interface region during loading, providing an effective means to observe crack propagation, interface debonding, and local deformation. However, relying on manual observation makes it difficult to obtain quantitative information such as the relative displacement between the two sides of the interface and local strain concentration. Traditional DIC methods are prone to matching failures in areas of interface debonding, crack propagation, and displacement discontinuity, making it difficult to accurately characterize complex interface damage processes.

[0004] Although the depth optical flow method developed in recent years can achieve pixel-level dense displacement estimation and provide a new approach for full-field motion analysis of in-situ SEM images, there are still shortcomings when applied directly: in-situ SEM images are often accompanied by platform drift, imaging noise and brightness changes, which can easily affect the accuracy of displacement calculation; at the same time, without interface geometric constraints and mechanical analysis, it is difficult to accurately identify interface damage areas and their failure mechanisms. Summary of the Invention

[0005] The purpose of this invention is to provide an in-situ interface failure analysis method based on deep optical flow and physical constraints, aiming to solve the problems mentioned in the background art.

[0006] The present invention is implemented as follows: an in-situ interface failure analysis method based on depth optical flow and physical constraints, comprising the following steps:

[0007] Step 1: Obtain sequential images of the metal-polymer heterogeneous material lap joint during in-situ SEM push-shear testing;

[0008] Step 2: Perform preprocessing, spatial calibration, and rigid body drift correction on the sequence of images to obtain a sequence of images under a unified spatial reference;

[0009] Step 3: Identify the metal-polymer interface in the sequence images, establish the interface geometric reference, and construct a geometry-aware mask based on the interface geometric reference;

[0010] Step 4: Input the sequence images of adjacent load stages into the depth optical flow full-field displacement reconstruction module to obtain the pixel-level dense displacement field between adjacent load stages, and obtain the physical displacement field and strain field based on the spatial calibration results;

[0011] Step 5: Based on the interface geometric reference, extract local kinematic indices in the region adjacent to the interface. The local kinematic indices include at least one of local tangential displacement, tangential displacement difference, shear strain distribution, and effective shear influence width.

[0012] Step 6: Calculate the local matrix energy storage index and the local interface dissipation index by combining the local kinematic indices and the material and interface physical parameters.

[0013] Step 7: Construct a normalized energy competition factor based on the local matrix energy storage index and the local interface dissipation index, and determine the failure mode tendency of the local region based on the normalized energy competition factor.

[0014] A further technical solution, the specific steps of step 1 are as follows:

[0015] First, a metal-polymer heterostructure lap joint specimen was prepared. The specimen was then installed in an in-situ SEM micro push-shear loading device so that the metal-polymer interface or its cross-section was within the SEM observation field. During the loading process, a push-shear load was applied to the specimen, and high-resolution SEM images of the adjacent area of ​​the interface were continuously acquired at different loading stages.

[0016] Different load stages include the initial unloaded stage, low load stage, medium load stage, high load stage, and near-instability stage; the acquired sequence images are denoted as: ,in, Indicates the first SEM images under each load stage Indicates the first SEM images under each load stage This is the payload stage number corresponding to the last image in the image sequence.

[0017] In a further technical solution, in step 2, the preprocessing includes one or more of the following: image grayscale conversion, image denoising, brightness normalization, contrast adjustment, and region of interest cropping.

[0018] For spatial calibration, the spatial calibration coefficients between pixel size and actual physical size are determined based on the scale bar in the SEM image, the known magnification, or the calibration sample. Spatial calibration coefficients Used to convert pixel displacements obtained from depth optical flow calculations into actual physical displacements;

[0019] For rigid body drift correction, a region far from the interface damage evolution region and which does not deform during loading is selected as the rigid body reference region. Based on the rigid body reference region, rigid body drift correction is performed on the sequence images or the initial displacement field. Rigid body drift correction is achieved through image registration or average displacement subtraction of the reference region.

[0020] A further technical solution, the specific steps of step 3 are as follows:

[0021] Metal-polymer interface identification was performed on the processed SEM image to obtain the interface curve, and the interface curve was used as the interface geometric reference.

[0022] An interface response map is constructed based on the grayscale, texture, or gradient information of the SEM image, and a continuous interface curve is obtained by using energy minimization, dynamic programming, edge tracking, manual calibration, or a combination thereof; subsequently, a geometry-aware mask is constructed based on the interface curve.

[0023] The geometry-aware mask includes at least a metal region mask, a polymer region mask, and an interface adjacent region mask; the geometry-aware mask is used for subsequent division of the metal region, polymer region, and interface region, and is also used to define the interface tangential direction, interface normal direction, local analysis path, and local analysis window.

[0024] In a further technical solution, in step 4, the depth optical flow full-field displacement reconstruction module includes a feature extraction unit, a correlation calculation unit, a context information extraction unit, and a displacement iteration update unit.

[0025] The feature extraction unit is used to extract multi-scale features of SEM images from adjacent load stages.

[0026] The correlation calculation unit is used to establish the matching relationship between the features of two SEM images;

[0027] The context information extraction unit is used to provide local texture, boundary, and region continuity information;

[0028] The displacement iteration update unit is used to iteratively update the displacement field and output a pixel-level dense displacement field between adjacent load stages.

[0029] In a further technical solution, step 4 includes the following specific steps:

[0030] Step 4.1: Combine two SEM images from adjacent load stages. and Input the depth optical flow full-field displacement reconstruction module to obtain the pixel-level dense displacement field from the k-th load stage to the (k+1)-th load stage, denoted as: ,in, Indicates the horizontal pixel displacement. Indicates pixel displacement in the vertical direction. and Indicates spatial location in the image. Indicates the load stage;

[0031] Step 4.2: Based on the spatial calibration coefficients The pixel-level displacement field output by the depth optical flow is converted into a physical displacement field; the physical displacement field includes horizontal physical displacement and vertical physical displacement, and its calculation method is as follows:

[0032] (1)

[0033] In the formula, and These represent the horizontal and vertical pixel displacements of the depth optical flow output, respectively. and These are the corresponding horizontal and vertical physical displacements, respectively.

[0034] Step 4.3: Calculate the two-dimensional strain field based on the physical displacement field, wherein the two-dimensional strain field includes horizontal normal strain. Vertical normal strain and shear strain The calculation method is as follows:

[0035] (2)

[0036] in, Used to characterize the local shear response in the region adjacent to the interface.

[0037] A further technical solution, the specific steps of step 5 are as follows:

[0038] One or more local analysis paths are deployed in the region adjacent to the interface. The local analysis paths are analysis probes that are deployed along the interface normal direction and cross the region adjacent to the interface.

[0039] For a selected analysis location on the interface curve, define the interface tangential unit vector at that location. and normal coordinates ;No. Under each load stage, the normal position Local tangential displacement at the location Calculate according to the following formula:

[0040] (3)

[0041] In the formula, Normal position The physical displacement vector at that location. Represents the dot product of vectors;

[0042] Select the proximal normal position on the local analysis path and the far-end normal position Calculate the tangential displacement difference It is used to characterize the local slip intensity in the region adjacent to the interface, and its calculation method is as follows:

[0043] (4)

[0044] in, For the first Far-end normal position under each load stage Local tangential displacement at the location, For the first Proximal normal position under each load phase Local tangential displacement at the location;

[0045] Based on the shear strain distribution within the local analysis path or local analysis region Determine the effective shear influence width The calculation method is as follows:

[0046] (5)

[0047] In the formula, The threshold coefficient is, and ; For the first The maximum shear strain in the local analysis path or local analysis area under each load stage; This represents the width of the normal region that meets the conditions.

[0048] A further technical solution, the specific steps of step 6 are as follows:

[0049] Step 6.1: Set up a local matrix analysis window in the polymer matrix region adjacent to the interface. According to the local matrix analysis window The equivalent strain and matrix strain energy density function within the matrix are used to calculate the local matrix energy storage index. This characterizes the changes in continuous deformation of the polymer matrix and energy storage demand in a local region, and its calculation method is as follows:

[0050] (6)

[0051] In the formula, This represents the area of ​​the local matrix analysis window. The matrix strain energy density function; No. Position under each load stage Equivalent change at the location.

[0052] Step 6.2: Set up a local interface analysis window in the adjacent area of ​​the interface. According to the local interface analysis window Calculate the local interface dissipation index based on the tangential relative displacement and the tangential traction-displacement relationship of the interface. This characterizes the changes in relative interface slip and interface dissipation demand in a local region, and its calculation method is as follows:

[0053] (7)

[0054] In the formula, Analyze the length of the local interface window; Coordinates along the interface; For the first Interface coordinates under each load stage Local tangential relative displacement; The interface tangential traction-displacement relationship. This represents the integral variable of the tangential relative displacement.

[0055] A further technical solution, the specific steps of step 7 are as follows:

[0056] Local matrix energy storage index and local interface dissipation index Normalization was performed to obtain the normalized local matrix energy storage index. and normalized local interface dissipation index Based on the normalized local matrix energy storage index and local interface dissipation index, a normalized energy competition factor is constructed. The calculation method is as follows:

[0057] (8)

[0058] In the formula, The value range is from 0 to 1; when Greater than the preset threshold When this occurs, the local area is determined to be a dominant interface failure; when Less than the preset threshold When this occurs, the local area is determined to be a matrix-dominated failure; when equal to the preset threshold When the local area is determined to have a mixed failure tendency, the analysis object includes at least two local analysis areas, and at least one local analysis area... Greater than the preset threshold At the same time, at least another local analysis region Less than the preset threshold When the analysis object is determined to have a failure path differentiation trend, it is determined that the analysis object has a failure path differentiation trend.

[0059] The interface in-situ failure analysis method based on depth optical flow and physical constraints provided in this invention has the following beneficial effects:

[0060] (1) Quantitative analysis: In-situ SEM sequence images are transformed into quantifiable full-field displacement field, strain field and failure mode determination results, overcoming the limitations of manual observation.

[0061] (2) High robustness full-field reconstruction: The deep optical flow method has a stronger full-field kinematic characterization capability in areas of displacement discontinuity such as crack propagation and interface debonding compared with the traditional DIC method.

[0062] (3) Eliminate non-realistic interference: Through rigid body drift correction and credibility evaluation, the interference of platform drift, field of view shift and other factors on the calculation results is effectively reduced.

[0063] (4) Clear physical meaning: The image analysis results are combined with physical parameters such as material elastic modulus and interface stiffness to give the analysis results a clear mechanical meaning.

[0064] (5) Process judgment capability: By constructing a normalized energy competition factor, dynamic judgment of the local failure mode tendency is realized during the loading process, realizing the transformation from "final state result judgment" to "loading process judgment". Attached Figure Description

[0065] Figure 1 The flowchart shows the overall process of the interface in-situ failure analysis method based on depth optical flow and physical constraints provided in the embodiments of the present invention.

[0066] Figure 2 A schematic diagram of in-situ SEM push-cut test and sequence image acquisition;

[0067] Figure 3 The diagram shows the process of metal-polymer interface recognition and geometry-sensing mask construction (where a is a SEM sequence image, b is coarse recognition, c is fine adsorption, and d is the geometry mask).

[0068] Figure 4 The displacement and strain fields obtained from depth optical flow reconstruction are shown in the figure (where a is the transverse displacement field, b is the longitudinal displacement field, and c is the strain field).

[0069] Figure 5 A schematic diagram is set up for the local analysis path, matrix analysis window, and near-interface analysis window;

[0070] Figure 6 The figure shows the evolution of the normalized energy competition factor with different load stages under different local paths.

[0071] In the attached figures: SEM chamber 1; in-situ SEM push-shear test instrument 2; uniaxial loading platform 3; objective lens 4; detector 5; metal-polymer overlap area 6; metal 7; polymer 8. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0073] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0074] like Figure 1 As shown, an embodiment of the present invention provides a method for in-situ interface failure analysis based on depth optical flow and physical constraints, comprising the following steps:

[0075] Step 1: Obtain sequential images of the metal-polymer heterogeneous material lap joint during in-situ SEM push-shear testing;

[0076] First, a metal-polymer heterogeneous material lap joint specimen is prepared. The metal-polymer heterogeneous material lap joint can be an aluminum alloy-polymer, steel-polymer, titanium alloy-polymer, magnesium alloy-polymer, or other metal-polymer composite structures; the polymer can be polyamide, polyphenylene sulfide, polypropylene, polyetheretherketone, or their fiber-reinforced composites. Preferably, the joint is an aluminum alloy-glass fiber reinforced polyamide injection-molded lap joint. The specimen is mounted in an in-situ SEM micro-push-shear loading device, so that the metal-polymer interface or its cross-section is within the SEM observation field. During loading, a push-shear load is applied to the specimen, and high-resolution SEM images of the area adjacent to the interface are continuously acquired at different loading stages.

[0077] Different load stages can include the initial unloaded stage, low-load stage, medium-load stage, high-load stage, and near-instability stage. For example, SEM images can be acquired at load levels of 0%, 40%, 60%, 80%, 90%, 95%, and 98% of the maximum load. The acquired image sequence is denoted as: ,in, Indicates the first SEM images under each load stage Indicates the first SEM images under each load stage This is the payload stage number corresponding to the last image in the image sequence.

[0078] Step 2: Perform preprocessing, spatial calibration, and rigid body drift correction on the sequence images to obtain a sequence of images under a unified spatial reference;

[0079] The preprocessing includes one or more of the following: image grayscale conversion, image denoising, brightness normalization, contrast adjustment, and region of interest cropping.

[0080] For spatial calibration, the spatial calibration coefficients between pixel size and actual physical size are determined based on the scale bar in the SEM image, the known magnification, or the calibration sample. The spatial calibration coefficients The unit can be μm / pixel, used to convert the pixel displacement obtained from depth optical flow calculation into actual physical displacement.

[0081] For rigid body drift correction, during in-situ SEM push-shear testing, non-real rigid body motion may exist in the sequence images due to slight movement of the loading platform, microscope imaging drift, or field of view shift. To reduce the impact of such non-real motion on the displacement field calculation results, a region far from the interface damage evolution region and which does not deform significantly during loading is selected as the rigid body reference region.

[0082] Rigid body drift correction is performed on sequential images or initial displacement fields based on a rigid body reference region. Rigid body drift correction can be achieved through image registration, reference region average displacement subtraction, or other equivalent methods. After correction, images from different load stages are placed under a unified spatial reference, thereby reducing the impact of platform drift, field of view shift, and imaging drift on subsequent displacement and strain field calculations.

[0083] Step 3: Identify the metal-polymer interface in the sequence images, establish the interface geometric reference, and construct a geometry-aware mask based on the interface geometric reference;

[0084] Metal-polymer interface identification was performed on the processed SEM image to obtain the interface curve, which was then used as the interface geometric reference.

[0085] An interface response map is constructed based on the grayscale, texture, or gradient information of the SEM image, and a continuous interface curve is obtained using energy minimization, dynamic programming, edge tracking, manual calibration, or a combination thereof. Subsequently, a geometry-aware mask is constructed based on the interface curve.

[0086] The geometry-aware mask includes at least a metal region mask, a polymer region mask, and an interface proximity region mask. The geometry-aware mask is used for subsequent delineation of the metal, polymer, and interface regions, and also for defining the interface tangential direction, interface normal direction, local analysis path, and local analysis window.

[0087] Step 4: Input the sequence images of adjacent load stages into the depth optical flow full-field displacement reconstruction module. This module includes a feature extraction unit, a correlation calculation unit, a context information extraction unit, and a displacement iteration update unit. The feature extraction unit is used to extract multi-scale features of SEM images of adjacent load stages. The correlation calculation unit is used to establish the matching relationship between the features of two SEM images. The context information extraction unit is used to provide local texture, boundary, and regional continuity information. The displacement iteration update unit is used to iteratively update the displacement field and output the pixel-level dense displacement field between adjacent load stages. The physical displacement field and strain field are obtained based on the spatial calibration results.

[0088] Step 4.1: Combine two SEM images from adjacent load stages. and Inputting the depth optical flow full-field displacement reconstruction module yields a pixel-level dense displacement field from the k-th load stage to the (k+1)-th load stage, which can be represented as follows: ,in, Indicates the horizontal pixel displacement. Indicates pixel displacement in the vertical direction. and Indicates spatial location in the image. This indicates the load stage.

[0089] Step 4.2: Based on the spatial calibration coefficients This converts the pixel-level displacement field output by the depth optical flow into a physical displacement field. The physical displacement field includes horizontal and vertical physical displacements, and its calculation method is as follows:

[0090] (1)

[0091] In the formula, and These represent the horizontal and vertical pixel displacements of the depth optical flow output, respectively. and These are the corresponding horizontal and vertical physical displacements, respectively.

[0092] Step 4.3: Calculate the two-dimensional strain field based on the physical displacement field, wherein the two-dimensional strain field includes horizontal normal strain. Vertical normal strain and shear strain The calculation method is as follows:

[0093] (2)

[0094] in, This is used to characterize the local shear response in the region adjacent to the interface. Through the above calculations, the full-field displacement and strain fields in the region adjacent to the metal-polymer interface under different loading stages can be obtained.

[0095] Step 5: Based on the interface geometric reference, extract local kinematic indices in the region adjacent to the interface. The local kinematic indices include at least one of local tangential displacement, tangential displacement difference, shear strain distribution, and effective shear influence width.

[0096] Specifically, one or more local analysis paths are deployed in the region adjacent to the interface. These local analysis paths are preferably analysis probes deployed along the interface normal direction and traversing the region adjacent to the interface.

[0097] For a selected analysis location on the interface curve, define the interface tangential unit vector at that location. and normal coordinates . No. Under each load stage, the normal position Local tangential displacement at the location Calculate according to the following formula:

[0098] (3)

[0099] In the formula, Normal position The physical displacement vector at that location. This represents the vector dot product. Further, the proximal normal position is selected on the local analysis path. and the far-end normal position Calculate the tangential displacement difference The calculation method is as follows:

[0100] (4)

[0101] in, For the first Far-end normal position under each load stage Local tangential displacement at the location, For the first Proximal normal position under each load phase The local tangential displacement at that location. The tangential displacement difference... Used to characterize the local slip intensity in the region adjacent to the interface. When When the shear strain increases rapidly with increasing load, it indicates the presence of strong tangential relative slip or localized deformation concentration in that region. Furthermore, this is determined based on the shear strain distribution within the local analysis path or region. Determine the effective shear influence width The calculation method is as follows:

[0102] (5)

[0103] In the formula, The threshold coefficient is, and ; For the first The maximum shear strain in the local analysis path or local analysis area under each load stage; This represents the width of the normal region that meets the conditions. The effective shear influence width... Used to characterize the extent of the significant shear response along the interface normal direction. When When the shear strain is smaller and the peak shear strain is higher, it indicates that the shear action is more concentrated; when A larger value indicates that the shearing action extends over a wider range.

[0104] Step 6: Calculate the local matrix energy storage index and the local interface dissipation index by combining the local kinematic indices and the material and interface physical parameters.

[0105] Step 6.1: To characterize the local continuous deformation and energy storage requirements on the polymer matrix side, a local matrix analysis window is set in the polymer matrix region near the interface. According to the local matrix analysis window The equivalent strain and matrix strain energy density function within the matrix are used to calculate the local matrix energy storage index. The calculation method is as follows:

[0106] (6)

[0107] In the formula, This represents the area of ​​the local matrix analysis window. The matrix strain energy density function; For the first Position under each load stage Equivalent change at the location.

[0108] pass It can characterize the changes in continuous deformation of the polymer matrix and energy storage demand in a local region.

[0109] Step 6.2: To characterize the relative slip and interfacial dissipation requirements at the metal-polymer interface, a local interface analysis window is set up in the region near the interface. According to the local interface analysis window Calculate the local interface dissipation index based on the tangential relative displacement and the tangential traction-displacement relationship of the interface. The calculation method is as follows:

[0110] (7)

[0111] In the formula, Analyze the length of the local interface window; Coordinates along the interface; For the first Interface coordinates under each load stage Local tangential relative displacement; The interface tangential traction-displacement relationship. This represents the integral variable of the tangential relative displacement.

[0112] Step 7: Construct a normalized energy competition factor based on the local matrix energy storage index and the local interface dissipation index, and determine the failure mode tendency of the local region based on the normalized energy competition factor.

[0113] Due to local matrix energy storage index and local interface dissipation index The dimensions, orders of magnitude, or range of change may differ. This invention normalizes both to obtain normalized local matrix energy storage indices. and normalized local interface dissipation index Based on the normalized local matrix energy storage index and local interface dissipation index, a normalized energy competition factor is constructed. The calculation method is as follows:

[0114] (8)

[0115] In the formula, The value range is from 0 to 1. When Greater than the preset threshold When it is determined that interface dissipation is dominant in this local region, this region is more prone to interface-dominated failure; when Less than the preset threshold When it is determined that matrix energy storage dominates in this local region, this region is more prone to matrix-dominated failure; when equal to the preset threshold When the local area is determined to have a mixed failure tendency, the analysis object includes at least two local analysis areas, and at least one local analysis area... Greater than the preset threshold At the same time, at least another local analysis region Less than the preset threshold At that time, it is determined that the analyzed object exhibits a failure path differentiation trend. In one embodiment, Take 0.5.

[0116] Through the above steps, the present invention can output the horizontal displacement field, vertical displacement field, horizontal normal strain, vertical normal strain, shear strain, local tangential displacement, tangential displacement difference, effective shear influence width, local matrix energy storage index, local interface dissipation index, normalized energy competition factor, and failure mode determination results under different load stages.

[0117] The failure mode determination results include matrix-dominated failure, interface-dominated failure, and mixed failure. The output results can be displayed or saved in the form of contour plots, graphs, tables, determination labels, or data files.

[0118] In a preferred embodiment of the present invention, in step 4, to improve the reliability of the physical displacement field and strain field, a credibility evaluation is further performed on the depth optical flow reconstruction results. The credibility evaluation includes, but is not limited to, image reconstruction error evaluation, forward and reverse consistency error evaluation, rigid body reference region residual displacement evaluation, and pseudo-strain noise evaluation.

[0119] When the image reconstruction error, forward and reverse consistency error, residual displacement, or pseudo-strain noise of the corresponding region meet the preset conditions, the optical flow reconstruction result of that region is deemed reliable; when the corresponding indicators do not meet the preset conditions, the region is marked, removed, or its weight is reduced. This step is used to reduce the impact of image noise, local gray-level abrupt changes, plateau drift, and abnormal matching in crack areas on the displacement and strain field results.

[0120] In a preferred embodiment of the present invention, in step 6.1, the matrix strain energy density function It can be determined based on the elastic modulus, yield strength, hardening modulus, tensile test results, indentation test results, or material constitutive model of the polymer matrix. In one embodiment, An elastic-plastic segmented form can be adopted, as follows:

[0121] (9)

[0122] in, This represents the elastic modulus of the PA6 matrix. The matrix yield strength, For plastic hardening modulus, The yield strain threshold, and ; For the local equivalent strain of the polymer matrix The strain energy density function under the given conditions, This is a local equivalent change.

[0123] In a preferred embodiment of the present invention, in step 6.2, the interface tangential traction-displacement relationship... It can be determined based on interface bond strength, interface stiffness, shear test results, fracture test results, or interface constitutive model. In one embodiment, A linear stiffness-strength constrained description can be used, as follows:

[0124] (10)

[0125] in, For the maximum bonding strength of the interface, For the apparent stiffness of the interface, This refers to the critical tangential relative displacement at which the interfacial tangential traction reaches the maximum interfacial bonding strength. Through... It can characterize the changes in relative interface slip and interface dissipation demand in local regions.

[0126] Based on the above method, the following is an in-situ SEM push-shear test sequence image analysis of aluminum alloy-glass fiber reinforced polyamide single lap joint.

[0127] Taking a metal-polymer single lap joint formed by 6061-T6 aluminum alloy and 30 wt% glass fiber reinforced polyamide 6 (GF30%-PA6) as the object, the above method is used to quantitatively analyze the displacement field, strain field, local slip behavior and failure mode tendency of the joint in the interface adjacent region during the push-shear loading process.

[0128] (1) Sample composition: The metal material used was 6061-T6 aluminum alloy, and the polymer material used was GF30%-PA6. First, the aluminum alloy surface was mechanically ground, cleaned, alkali etched, neutralized and de-dusted, treated with hot water and plasma activated to improve the bonding ability between the metal surface and the polymer. Then, the treated aluminum alloy insert was placed in an injection mold, and an aluminum alloy-GF30%-PA6 single lap joint was prepared by injection molding.

[0129] A single lap joint comprises a metal region, a polymer region, and a metal-polymer interface between them. To accommodate the space constraints of in-situ SEM push-shear testing, the macroscopic lap joint specimen is further processed into a small-sized specimen suitable for clamping and observation by a micro-loading device within the SEM chamber, exposing the cross-section of the metal-polymer interface and placing the area adjacent to the interface within the SEM observation field of view.

[0130] (2) In-situ SEM push-cut test and sequence image acquisition;

[0131] Combination Figure 2The aluminum alloy-GF30%-PA6 single-lapped sample was mounted in an in-situ SEM push-shear testing instrument 2. The in-situ SEM push-shear testing instrument 2 was placed inside the SEM chamber 1, and the sample was fixed on a uniaxial loading platform 3. Objective lens 4 and detector 5 were used to acquire high-resolution images of the region adjacent to the metal-polymer interface. During loading, a push-shear load was applied to the sample along a preset loading direction, and SEM images at different loading stages were acquired simultaneously to obtain an in-situ SEM push-shear test sequence image.

[0132] In this embodiment, the acquired sequence of images can be represented as the first... Frame SEM image, the first Frame SEM image up to the 1st Frame SEM images. The sequence of images corresponds to the interface morphology of the specimen under different shear loading stages, and can be used for subsequent full-field displacement reconstruction of depth optical flow, strain field calculation, and local failure mode determination.

[0133] Figure 2 This embodiment illustrates the in-situ SEM shearing test and sequence image acquisition process, including the SEM chamber 1, objective lens 4, detector 5, in-situ SEM shearing test instrument 2, uniaxial loading platform 3, metal-polymer overlap region 6, metal 7, polymer 8, loading direction, and SEM images of different frames. This step allows for the acquisition of the required raw sequence image input.

[0134] (3) Image preprocessing, spatial calibration and interface recognition;

[0135] The obtained SEM image sequence is preprocessed, including one or more of the following: grayscale conversion, noise suppression, brightness normalization, contrast adjustment, image registration, and region of interest (ROI) cropping. Spatial calibration coefficients are determined based on the scale bar in the SEM image. It is used to convert the pixel displacement output by the depth optical flow method into the actual physical displacement.

[0136] Combination Figure 3 The metal-polymer interface is identified in the preprocessed SEM image. First, the metal-polymer interface is roughly identified based on the differences in grayscale, texture, or gradient between the metal and polymer regions in the SEM image to obtain an initial interface curve. Then, local interface refinement is performed near the initial interface curve to obtain a fine interface curve that more closely approximates the actual material boundary. Finally, a geometry-aware mask is constructed based on the fine interface curve.

[0137] The geometry-sensing mask is used to distinguish between metal regions, polymer regions, and regions adjacent to the interface, and is subsequently used to define the interface tangential direction, interface normal direction, local analysis path, and local matrix analysis window. and local interface analysis window This step provides a unified interface geometric benchmark for subsequent full-field displacement reconstruction, strain field calculation, and local energy competition analysis.

[0138] Figure 3 This illustration demonstrates the interface recognition and geometry-aware mask construction process in this embodiment, including SEM image sequence, coarse recognition, interface refinement, and the final generated geometry mask. Figure 3 a to Figure 3 As can be seen, this method can extract continuous metal-polymer bonding interfaces from complex SEM images and generate geometric masks that can be used for region segmentation.

[0139] (4) Depth optical flow full-field displacement reconstruction and strain field calculation;

[0140] Input the SEM images of adjacent load stages into the depth optical flow full-field displacement reconstruction module to obtain the result from the first load stage. Frame image to the Pixel-level dense displacement field between frames. Based on spatial calibration coefficients. The pixel-level displacement field is converted into an actual physical displacement field, yielding the lateral and longitudinal displacement fields. Further, the strain field is calculated based on the physical displacement field to obtain the shear strain field in the region adjacent to the interface.

[0141] Combination Figure 4 In this embodiment, the output is the first... Frame image and the first The frame corresponds to the transverse displacement field, longitudinal displacement field, and shear strain field. (From...) Figure 4 a to Figure 4 As can be seen from c, with the increase of shear load, the displacement response and shear strain response in the region near the metal-polymer interface change significantly, especially the local shear concentration phenomenon near the interface and in the region near polymer matrix defects. This result shows that the proposed method can transform the morphological changes in in-situ SEM sequence images into physically meaningful full-field kinematic data.

[0142] The transverse displacement field characterizes the local displacement accumulation along the shear direction in the vicinity of the interface, the longitudinal displacement field characterizes the local motion response in the vertical direction, and the shear strain field characterizes the degree of local shear deformation concentration in the vicinity of the interface. These results enable a quantitative characterization of the damage evolution process at the metal-polymer interface.

[0143] (5) Local analysis path and local analysis window settings;

[0144] Combination Figure 5Based on the metal-polymer interface identified in step 3, two representative local analysis paths are selected in the region adjacent to the interface, denoted as Analysis Path 1 and Analysis Path 2, respectively. The analysis paths are laid out along the normal direction of the interface and cross the region adjacent to the metal-polymer interface.

[0145] A local matrix analysis window and a local interface analysis window are set near each analysis path. The local matrix analysis window is used to analyze the continuous deformation and local energy storage requirements on the polymer matrix side; the local interface analysis window is used to analyze the relative slip and interface dissipation requirements near the metal-polymer interface.

[0146] For each analysis path, the local tangential displacement, tangential displacement difference, and effective shear influence width are extracted under different load stages. The local tangential displacement characterizes the local motion along the tangential direction of the interface; the tangential displacement difference characterizes the local slip intensity in the region adjacent to the interface; and the effective shear influence width characterizes the extent of the significant shear response along the normal direction of the interface.

[0147] This step allows the full-field displacement and strain fields to be further transformed into kinematic indices that can be used to determine local failure modes.

[0148] (6) Local energy competition analysis and failure mode determination;

[0149] Within the local matrix analysis windows corresponding to analysis paths one and two, the local matrix energy storage index is calculated based on the strain field and matrix strain energy density function obtained from depth optical flow reconstruction. Within the local interface analysis windows corresponding to analysis paths one and two, the local interface dissipation index is calculated based on the local tangential relative displacement and the interface tangential traction-displacement relationship.

[0150] In this embodiment, the physical parameters required for local energy competition analysis include the polymer matrix elastic modulus, yield strength, hardening modulus, maximum interfacial shear strength, and interfacial stiffness. These parameters can be obtained through material testing, shear testing, indentation testing, or literature data.

[0151] Furthermore, the local matrix energy storage index and the local interface dissipation index are normalized to obtain the normalized local matrix energy storage index and the normalized local interface dissipation index, and the normalized energy competition factor is calculated. .when Less than the preset threshold When this occurs, it indicates that matrix energy storage dominates in that local region, and that region is more prone to matrix-dominated failure; when Greater than the preset threshold When this occurs, it indicates that interface dissipation is dominant in that local area, and that area is more prone to interface-dominated failure. In this embodiment, a preset threshold is used. Take 0.5.

[0152] Combination Figure 6 Analyze the normalized energy competition factor of path one The normalized energy competition factor was above 0.5 during the lower load phase, then decreased significantly with increasing load, and fell below 0.5 during the medium-to-high load phase. This indicates that the response in the local region analyzed by Pathway 1 gradually shifted to matrix energy storage dominance, and this region was more prone to matrix-dominated failure. The normalized energy competition factor for Pathway 2... Throughout the loading process, the value remained above 0.5, indicating that the interface dissipation accounted for a relatively high proportion in the local area where analysis path two is located, and that this area is more prone to interface-dominated failure.

[0153] As can be seen from the above analysis, although the two local regions in this embodiment are located within the same metal-polymer lap joint, their normalized energy competition factors exhibit different evolution trends due to differences in local defect states, interface integrity, and local shear transmission modes, ultimately showing different failure mode tendencies.

[0154] (7) Effects: As can be seen from this embodiment, this method can transform the sequence images collected during the in-situ SEM shear test into a full-field displacement field and strain field with physical meaning, and further obtain local tangential displacement, tangential displacement difference, effective shear influence width, local matrix energy storage index, local interface dissipation index and normalized energy competition factor.

[0155] Compared to methods that rely solely on manual observation of SEM images or final fracture morphology analysis, this invention can quantitatively identify deformation concentration, local slippage, and failure path evolution trends in the vicinity of the interface during loading. Compared to traditional DIC analysis methods, this invention can utilize depth optical flow to reconstruct pixel-level dense displacement fields, exhibiting better full-field kinematic characterization capabilities in areas of interface debonding, crack propagation, abrupt displacement gradient changes, and rapid texture changes.

[0156] This embodiment demonstrates that for aluminum alloy-GF30%-PA6 single lap joints, this method can not only obtain the transverse displacement field, longitudinal displacement field, and shear strain field under different load stages, but also distinguish the failure mode tendencies of different local regions through local energy competition factors. Therefore, this invention can achieve quantitative analysis and failure mode determination of the interface failure process of metal-polymer heterogeneous material lap joints, and can be used for joint interface reliability evaluation, surface treatment process comparison, material combination screening, and structural optimization design.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for in-situ interface failure analysis based on depth optical flow and physical constraints, characterized in that, Includes the following steps: Step 1: Obtain sequential images of the metal-polymer heterogeneous material lap joint during in-situ SEM push-shear testing; Step 2: Perform preprocessing, spatial calibration, and rigid body drift correction on the sequence of images to obtain a sequence of images under a unified spatial reference; Step 3: Identify the metal-polymer interface in the sequence images, establish the interface geometric reference, and construct a geometry-aware mask based on the interface geometric reference; Step 4: Input the sequence images of adjacent load stages into the depth optical flow full-field displacement reconstruction module to obtain the pixel-level dense displacement field between adjacent load stages, and obtain the physical displacement field and strain field based on the spatial calibration results; Step 5: Based on the interface geometric reference, extract local kinematic indices in the region adjacent to the interface. The local kinematic indices include at least one of local tangential displacement, tangential displacement difference, shear strain distribution, and effective shear influence width. Step 6: Calculate the local matrix energy storage index and the local interface dissipation index by combining the local kinematic indices and the material and interface physical parameters. Step 7: Construct a normalized energy competition factor based on the local matrix energy storage index and the local interface dissipation index, and determine the failure mode tendency of the local region based on the normalized energy competition factor.

2. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 1, characterized in that, The specific steps of step 1 are as follows: First, a metal-polymer heterostructure lap joint specimen was prepared. The specimen was then installed in an in-situ SEM micro push-shear loading device so that the metal-polymer interface or its cross-section was within the SEM observation field. During the loading process, a push-shear load was applied to the specimen, and high-resolution SEM images of the adjacent area of ​​the interface were continuously acquired at different loading stages. Different load stages include the initial unloaded stage, low load stage, medium load stage, high load stage, and near-instability stage; the acquired sequence images are denoted as: ,in, Indicates the first SEM images under each load stage Indicates the first SEM images under each load stage This is the payload stage number corresponding to the last image in the image sequence.

3. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 2, characterized in that, In step 2, the preprocessing includes one or more of the following: image grayscale conversion, image denoising, brightness normalization, contrast adjustment, and region of interest cropping. For spatial calibration, the spatial calibration coefficients between pixel size and actual physical size are determined based on the scale bar in the SEM image, the known magnification, or the calibration sample. Spatial calibration coefficients Used to convert pixel displacements obtained from depth optical flow calculations into actual physical displacements; For rigid body drift correction, a region far from the interface damage evolution region and which does not deform during loading is selected as the rigid body reference region. Based on the rigid body reference region, rigid body drift correction is performed on the sequence images or the initial displacement field. Rigid body drift correction is achieved through image registration or average displacement subtraction of the reference region.

4. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 3, characterized in that, The specific steps of step 3 are as follows: Metal-polymer interface identification was performed on the processed SEM image to obtain the interface curve, and the interface curve was used as the interface geometric reference. An interface response map is constructed based on the grayscale, texture, or gradient information of the SEM image, and a continuous interface curve is obtained by using energy minimization, dynamic programming, edge tracking, manual calibration, or a combination thereof. Subsequently, a geometry-aware mask is constructed based on the interface curve; The geometry-aware mask includes at least a metal region mask, a polymer region mask, and an interface adjacent region mask; the geometry-aware mask is used for subsequent division of the metal region, polymer region, and interface region, and is also used to define the interface tangential direction, interface normal direction, local analysis path, and local analysis window.

5. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 1, characterized in that, In step 4, the depth optical flow full-field displacement reconstruction module includes a feature extraction unit, a correlation calculation unit, a context information extraction unit, and a displacement iteration update unit. The feature extraction unit is used to extract multi-scale features of SEM images from adjacent load stages. The correlation calculation unit is used to establish the matching relationship between the features of two SEM images; The context information extraction unit is used to provide local texture, boundary, and region continuity information; The displacement iteration update unit is used to iteratively update the displacement field and output a pixel-level dense displacement field between adjacent load stages.

6. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 4, characterized in that, Step 4 includes the following specific steps: Step 4.1: Combine two SEM images from adjacent load stages. and Input the depth optical flow full-field displacement reconstruction module to obtain the pixel-level dense displacement field from the k-th load stage to the (k+1)-th load stage, denoted as: ,in, Indicates the horizontal pixel displacement. Indicates pixel displacement in the vertical direction. and Indicates spatial location in the image. Indicates the load stage; Step 4.2: Based on the spatial calibration coefficients The pixel-level displacement field output by the depth optical flow is converted into a physical displacement field; the physical displacement field includes horizontal physical displacement and vertical physical displacement, and its calculation method is as follows: （1） In the formula, and These represent the horizontal and vertical pixel displacements of the depth optical flow output, respectively. and These are the corresponding horizontal and vertical physical displacements, respectively. Step 4.3: Calculate the two-dimensional strain field based on the physical displacement field, wherein the two-dimensional strain field includes horizontal normal strain. Vertical normal strain and shear strain The calculation method is as follows: （2） in, Used to characterize the local shear response in the region adjacent to the interface.

7. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 6, characterized in that, The specific steps of step 5 are as follows: One or more local analysis paths are deployed in the region adjacent to the interface. The local analysis paths are analysis probes that are deployed along the interface normal direction and cross the region adjacent to the interface. For a selected analysis location on the interface curve, define the interface tangential unit vector at that location. and normal coordinates ;No. Under each load stage, the normal position Local tangential displacement at the location Calculate according to the following formula: （3） In the formula, Normal position The physical displacement vector at that location. Represents the dot product of vectors; Select the proximal normal position on the local analysis path and the far-end normal position Calculate the tangential displacement difference It is used to characterize the local slip intensity in the region adjacent to the interface, and its calculation method is as follows: （4） in, For the first Far-end normal position under each load stage Local tangential displacement at the location, For the first Proximal normal position under each load phase Local tangential displacement at the location; Based on the shear strain distribution within the local analysis path or local analysis region Determine the effective shear influence width The calculation method is as follows: （5） In the formula, The threshold coefficient is, and ; For the first The maximum shear strain in the local analysis path or local analysis area under each load stage; This represents the width of the normal region that meets the conditions.

8. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 7, characterized in that, The specific steps of step 6 are as follows: Step 6.1: Set up a local matrix analysis window in the polymer matrix region adjacent to the interface. According to the local matrix analysis window The equivalent strain and matrix strain energy density function within the matrix are used to calculate the local matrix energy storage index. This characterizes the changes in continuous deformation of the polymer matrix and energy storage demand in a local region, and its calculation method is as follows: （6） In the formula, This represents the area of ​​the local matrix analysis window. The matrix strain energy density function; No. Position under each load stage Equivalent change at the point; Step 6.2: Set up a local interface analysis window in the adjacent area of ​​the interface. According to the local interface analysis window Calculate the local interface dissipation index based on the tangential relative displacement and the tangential traction-displacement relationship of the interface. This characterizes the changes in relative interface slip and interface dissipation demand in a local region, and its calculation method is as follows: （7） In the formula, Analyze the length of the local interface window; Coordinates along the interface; For the first Interface coordinates under each load stage Local tangential relative displacement; The interface tangential traction-displacement relationship. This represents the integral variable of the tangential relative displacement.

9. The in-situ interface failure analysis method based on depth optical flow and physical constraints according to claim 8, characterized in that, The specific steps of step 7 are as follows: Local matrix energy storage index and local interface dissipation index Normalization was performed to obtain the normalized local matrix energy storage index. and normalized local interface dissipation index Based on the normalized local matrix energy storage index and local interface dissipation index, a normalized energy competition factor is constructed. The calculation method is as follows: （8） In the formula, The value range is from 0 to 1; when Greater than the preset threshold When this occurs, the local area is determined to be a dominant interface failure; when Less than the preset threshold When this occurs, the local area is determined to be a matrix-dominated failure; when equal to the preset threshold When the local area is determined to have a mixed failure tendency, the analysis object includes at least two local analysis areas, and at least one local analysis area... Greater than the preset threshold At the same time, at least another local analysis region Less than the preset threshold When the analysis object is determined to have a failure path differentiation trend, it is determined that the analysis object has a failure path differentiation trend.

Images