Surface flaw analysis method combined with metal material fatigue life prediction

Abstract

The invention provides a surface flaw analysis method combined with metal material fatigue life prediction, which comprises the following steps: carrying out grid optimization and refinement processing on an initial three-dimensional model, and carrying out refinement reconstruction on a local area to obtain a three-dimensional geometric model of the surface of a metal material, the three-dimensional geometric model comprising micro-scale surface flaw information; carrying out region division on the three-dimensional geometric model, identifying and extracting a surface defect region, carrying out edge detection on the surface defect region, and determining an accurate geometric contour and spatial position information of a defect; using finite element analysis software to import the three-dimensional geometric model and the flaw mathematical description model of the metal material, setting material attributes, boundary conditions and load conditions, and establishing a finite element analysis model containing microscopic flaws; and carrying out stress-strain analysis on the finite element analysis model of the microscopic flaws, calculating stress distribution and stress concentration coefficients around the flaws, and evaluating the influence of the flaws on fatigue crack initiation and expansion by adopting a fracture mechanics theory.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a surface defect analysis method combined with fatigue life prediction of metal materials. Background Art

[0002] In the field of fatigue life prediction of metal materials, the accurate characterization of surface defects has always been a thorny problem. Traditional two-dimensional surface detection methods are difficult to fully capture the spatial geometric characteristics of defects, resulting in limited accuracy of fatigue life prediction models. Although reverse engineering provides the possibility of three-dimensional reconstruction of defects, it still faces many challenges in practical applications. For micron-level surface defects, existing three-dimensional reconstruction algorithms often find it difficult to accurately capture their fine structures. The processing of three-dimensional point cloud data is time-consuming and difficult to meet the needs of rapid detection. In addition, how to seamlessly integrate the reconstructed three-dimensional defect model into finite element analysis to achieve efficient stress concentration calculation is also an urgent problem to be solved. Finally, the evolution of defects in metal materials during service, how to introduce dynamic defect evolution mechanisms into fatigue life prediction models to more accurately reflect the actual use of materials, is an important direction of current research. The solution of these technical problems will directly affect the accuracy and reliability of fatigue life prediction, and is of great significance to improving the safety and service life of metal components. Summary of the invention

[0003] The present invention provides a surface defect analysis method combined with fatigue life prediction of metal materials, which mainly includes:

[0004] The surface of the metal material is scanned by a 3D laser scanner to obtain the surface morphology point cloud data. According to the density and distribution characteristics of the surface morphology point cloud data, an initial 3D model of the metal material surface is constructed by a triangular mesh reconstruction algorithm.

[0005] The initial 3D model is meshed and refined, and the local area is reconstructed in detail to obtain a 3D geometric model of the metal material surface. The 3D geometric model contains microscopic surface defect information.

[0006] Divide the 3D geometric model into regions, identify and extract surface defect regions, perform edge detection on surface defect regions, and determine the precise geometric contour and spatial location information of the defects;

[0007] According to the geometric profile and spatial position information of the defect, a parameterized defect model is established, including the key dimension parameters of the defect's depth, width, and length. The mathematical description model of the defect is obtained through the least squares fitting method.

[0008] Using finite element analysis software, the three-dimensional geometric model of metal materials and the mathematical description model of defects are imported, material properties, boundary conditions and load conditions are set, and a finite element analysis model including microscopic defects is established;

[0009] Perform stress-strain analysis on the finite element analysis model of micro-defects, calculate the stress distribution and stress concentration factor around the defects, and use fracture mechanics theory to evaluate the impact of defects on fatigue crack initiation and propagation;

[0010] According to the evaluation results, based on the Paris formula and cumulative damage theory, combined with the stress analysis results and material fatigue parameters, a fatigue life prediction model combining the influence of defects is established to calculate the crack propagation life, and realize the surface defect analysis of metal material fatigue life prediction.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a surface defect analysis method combined with fatigue life prediction of metal materials. The method obtains metal surface morphology point cloud data through three-dimensional laser scanning, and constructs a three-dimensional geometric model containing micro-defect information. The model is divided into regions and edge detected, the precise geometric contour and spatial position information of the defects are extracted, and a parameterized defect model is established. The geometric model and the defect model are imported into finite element analysis software, stress-strain analysis is performed, and the influence of defects on fatigue cracks is evaluated. Based on the analysis results, combined with the Paris formula and cumulative damage theory, a fatigue life prediction model considering the influence of defects is established. The present invention realizes the accurate characterization and quantitative analysis of micro-defects on the surface of metal materials, improves the accuracy of fatigue life prediction, and provides a new technical means for the safety assessment and life prediction of metal components. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a surface defect analysis method combined with metal material fatigue life prediction.

[0014] Figure 2 It is a schematic diagram of a surface defect analysis method combined with fatigue life prediction of metal materials according to the present invention.

[0015] Figure 3 It is another schematic diagram of a surface defect analysis method combined with metal material fatigue life prediction of the present invention. DETAILED DESCRIPTION

[0016] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0017] like Figure 1 -3. In this embodiment, a surface defect analysis method combined with metal material fatigue life prediction may specifically include:

[0018] S101. Scan the surface of the metal material using a 3D laser scanner to obtain surface morphology point cloud data, and construct an initial 3D model of the metal material surface using a triangular mesh reconstruction algorithm based on the density and distribution characteristics of the surface morphology point cloud data.

[0019] A high-precision three-dimensional laser scanner is used to perform high-density scanning on the surface of the metal material to obtain high-resolution surface morphology point cloud data, which contains the three-dimensional coordinate information of the surface of the metal material. For the point cloud data, the effective point cloud data set is screened, and the abnormal points and noise points are eliminated to obtain the optimized point cloud data set. According to the optimized point cloud data set, the principal component analysis method is used to calculate the normal vector of each point, and the curvature is calculated using the differential geometry method. The point cloud is feature segmented according to the curvature threshold to identify the plane area, edge area and curved surface area. Based on the feature segmentation results, an initial triangular mesh is constructed, and the mesh quality is optimized by edge flipping and vertex insertion. The local quadratic surface fitting method is used to improve the model accuracy and generate a surface three-dimensional model. The surface three-dimensional model is calibrated to ensure that the model is consistent with the actual surface size of the metal material, and the surface area and volume of the model are calculated to obtain standardized metal material surface three-dimensional model data.

[0020] Specifically, a high-precision 3D laser scanner is used to perform high-density scanning on the surface of metal materials, and appropriate scanning resolution and scanning step length are set to obtain high-resolution surface morphology point cloud data. According to the density and spatial distribution characteristics of the point cloud data, the effective point cloud data set is screened, and abnormal points and noise points are eliminated to obtain an optimized point cloud data set. For the optimized point cloud data set, the KD tree algorithm is used for spatial indexing to improve the retrieval efficiency of the point cloud data. The principal component analysis method is used to calculate the normal vector of each point, and the differential geometry method is used to calculate the curvature. The point cloud is feature segmented according to the curvature threshold to identify the plane area, edge area and surface area. The initial triangular mesh is constructed based on the point cloud data, and the mesh quality is optimized by edge flipping and vertex insertion. The local quadratic surface fitting method is used to improve the model accuracy, and the mesh density is adjusted to balance the model accuracy and data volume to generate a surface 3D model with uniform mesh distribution. The generated surface 3D model is calibrated to ensure that the model is consistent with the actual metal material surface size, and the surface area and volume of the model are calculated to obtain standardized metal material surface 3D model data. When using a high-precision 3D laser scanner to scan the surface of metal materials, the scanning resolution is set to 0.1mm and the scanning step is set to 0.05mm to ensure the acquisition of high-quality point cloud data. During the scanning process, the laser beam is emitted at a frequency of 50kHz, and the reflected signal is processed to form a point cloud data set containing spatial coordinates and reflection intensity information. The original point cloud data is preprocessed, and the statistical outlier filtering algorithm is used to remove abnormal points. The neighborhood radius is set to 1mm, the standard deviation threshold is set to 2.5, and the noise points removed account for about 3% of the total data volume. The KD tree algorithm is used to perform spatial indexing on the optimized point cloud data, and the three-dimensional space is divided into multiple cubic grids, each with a side length of 2mm, to improve data retrieval efficiency. When calculating the normal vector of each point, a point set with a neighborhood radius of 1.5mm is selected, and the eigenvector of the covariance matrix is ​​solved by the principal component analysis method. The curvature is calculated using the differential geometry method, and the neighborhood radius is taken as 1mm to calculate the Gaussian curvature and the mean curvature. The point cloud was segmented according to the curvature threshold of 0.01mm^-1, and the plane area, edge area and curved area were identified. The initial triangular mesh was constructed based on the segmentation results. The improved Ball-Pivoting algorithm was used to generate the initial mesh with the ball radius set to 0.5mm. The mesh quality was optimized by edge flipping and vertex insertion. The edge flip angle threshold was set to 15° and the vertex insertion spacing threshold was set to 0.3mm. The local quadratic surface fitting method was used to improve the model accuracy. The least squares fitting was performed on each triangular face and its adjacent face, and the fitting error threshold was set to 0.05mm. The mesh density was adjusted to control the side length of the triangular face between 0.2mm and 1mm, and a uniformly distributed surface three-dimensional model was generated. The generated model was calibrated, and the feature points on the model were selected for comparison with the actual measurement values. The error was controlled within ±0.1mm.

[0021] S102, performing mesh optimization and refinement processing on the initial three-dimensional model, and reconstructing the local area in detail to obtain a three-dimensional geometric model of the metal material surface, wherein the three-dimensional geometric model includes microscopic-scale surface defect information.

[0022] The initial three-dimensional model is obtained, and the mesh topology, geometric features and quality indicators of the initial three-dimensional model are calculated according to the preset threshold value, wherein the quality indicators include the mesh side length ratio, dihedral angle and area ratio; it is judged whether the quality indicators meet the preset conditions, and if not, the mesh units that need to be optimized are screened out. The model surface is divided according to the curvature value of the initial three-dimensional model to obtain low curvature areas and high curvature areas; wherein, the area with a curvature less than the preset first threshold is the low curvature area, and the area with a curvature greater than the preset second threshold is the high curvature area. The quadratic error metric algorithm is used to simplify the mesh of the low curvature area, and the redundant triangular facets are reduced by the edge collapse operation; the Loop subdivision algorithm is used to locally encrypt the high curvature area and the feature boundary to obtain the optimized global mesh. The identified surface defect area is refined and reconstructed, and the mesh density is dynamically adjusted according to the local curvature value; the local mesh density is increased by using the adaptive mesh refinement method, and the defect edge is accurately described in combination with the non-uniform rational B-spline surface fitting algorithm to obtain a high-precision local model containing micro-scale surface defect information. The optimized global grid is merged with the high-precision local model, and the regional growing algorithm is used to achieve seamless connection between the local fine model and the global model. A progressive grid density adjustment is adopted in the transition area to adjust the grid density and topological structure of the transition area, so as to obtain a three-dimensional geometric model of the metal material surface that retains the overall morphology and contains microscopic defect information.

[0023] Specifically, the mesh quality of the initial three-dimensional model is evaluated, and the topological structure, geometric characteristics and quality indicators of the mesh are calculated, including the mesh side length ratio, dihedral angle and area ratio. The mesh units that need to be optimized are screened out according to the preset threshold, and the mesh side length ratio is set to be between 0.5 and 2, and the dihedral angle is set to be between 30° and 150°. According to the evaluation results, the model surface is divided into a low curvature area with a curvature less than 0.1mm^-1 and a high curvature area with a curvature greater than 1mm^-1. The quadratic error metric algorithm is used to simplify the mesh in the low curvature area, and the redundant triangular facets are reduced by edge collapse operation while keeping the overall shape of the model unchanged. The Loop subdivision algorithm is used for local encryption of high curvature areas and feature boundaries. The identified surface defect area is refined and reconstructed, and the mesh density is dynamically adjusted according to the local curvature value. The larger the curvature, the denser the mesh. The adaptive mesh refinement method is used to increase the local mesh density, and the defect edge is accurately described in combination with the non-uniform rational B-spline surface fitting algorithm to generate a high-precision local model containing micro-scale surface defect information. The optimized global mesh is fused with the refined local model, and the regional growing algorithm is used to achieve seamless connection between the local refined model and the global model. The progressive mesh density adjustment is used in the transition area to adjust the mesh density and topological structure of the transition area, so as to obtain a three-dimensional geometric model of the metal material surface that retains the overall morphology and contains microscopic defect information. When evaluating the mesh quality of the initial three-dimensional model, the Hausdorff distance is used to calculate the mesh side length ratio, and the threshold is set to 0.5-2; the cosine theorem is used to calculate the dihedral angle, which is controlled within the range of 30°-150°; the area ratio is calculated by the area ratio of adjacent triangles, which is limited to 0.2-5. According to the evaluation results, the area with a curvature less than 0.1mm^-1 is divided into a low curvature area, and the area with a curvature greater than 1mm^-1 is divided into a high curvature area. The quadratic error metric algorithm is applied to the low curvature area to simplify the mesh, and the maximum edge collapse error is set to 0.01mm, and the number of triangles is reduced by 5% each iteration. For high curvature areas and feature boundaries, the Loop subdivision algorithm is used for local encryption, and the subdivision threshold is set to 0.5mm edge length, and two iterative subdivisions are performed. When the surface defect area is finely reconstructed, the curvature-based adaptive mesh refinement method is used. When the local curvature is greater than 2mm^-1, the mesh density is doubled. The third-order non-uniform rational B-spline surface fitting algorithm is used to describe the defect edge. The number of control points is set to 16×16, and the fitting accuracy reaches 0.001mm. When fusing the global mesh and the local fine model, the regional growing algorithm is applied, with the local model boundary as the seed point, and 5-10 layers of triangular facets are expanded outward. The linear interpolation method is used in the transition area to adjust the mesh density so that the mesh edge length gradually transitions from 0.05mm in the local model to 0.5mm in the global model.

[0024] Based on the acquired initial 3D model, a mesh quality analysis is performed on the 3D model, and the triangular facets are optimized according to the obtained mesh quality evaluation results.

[0025] Obtain mesh data of a three-dimensional model, and for each triangular face in the mesh data, calculate the side length ratio, dihedral angle, and area ratio of the triangular face to obtain geometric feature parameters of the triangular face; according to the geometric feature parameters, use a discrete differential geometry method to calculate the Gaussian curvature and mean curvature of each vertex to obtain a mesh quality evaluation index; combine the side length ratio, the dihedral angle, and the area ratio to calculate a comprehensive quality index, and if the comprehensive quality index is less than a preset multi-level threshold, it is determined to be a low-quality mesh unit; for the low-quality mesh unit, use a Laplace smoothing algorithm to perform local optimization, and improve the shape of the triangular face by adjusting the vertex position; recalculate the mesh quality evaluation index of the locally optimized three-dimensional model, and if the mesh quality evaluation index is greater than a preset threshold, it is determined that the local optimization is effective; generate a mesh quality analysis report including a mesh quality distribution histogram, a quality improvement percentage, and a comparison chart before and after local optimization.

[0026] Specifically, according to the mesh data of the initial three-dimensional model, the geometric characteristic parameters of each triangular facet are calculated, including the side length ratio, dihedral angle, and area ratio, and the mesh density is counted. The Gaussian curvature and mean curvature of each vertex are calculated using the discrete differential geometry method to obtain the mesh quality evaluation index. The comprehensive quality index is calculated by combining the side length ratio, dihedral angle, and area ratio, and a multi-level threshold is set to grade the mesh quality. Triangular faces with a side length ratio less than 0.2 or greater than 5, a dihedral angle less than 30 degrees or greater than 150 degrees, and an area ratio less than 0.1 or greater than 10 are marked as low-quality mesh units. The Laplace smoothing algorithm is used to locally optimize the low-quality mesh units, and the shape of the triangular facets is improved by adjusting the vertex position. Boundary constraints and volume preservation conditions are set to ensure that the overall shape of the model is not changed during the optimization process. The optimization is iterated until the mesh quality index meets the preset threshold, and the optimization results are preliminarily evaluated. The mesh quality evaluation index of the optimized 3D model is recalculated and compared with the initial model to verify the mesh optimization effect. A mesh quality analysis report containing a mesh quality distribution histogram, quality improvement percentage, and a comparison chart before and after local optimization is generated. The data contains mesh quality statistics and distribution charts before and after optimization. Based on the mesh data of the initial 3D model, the geometric characteristics of each triangular face are analyzed to calculate the ratio of the longest side to the shortest side, the angle between adjacent facets, and the ratio of the area of ​​adjacent facets. The k-nearest neighbor algorithm is used to count the mesh density, with k=6, and the average distance of the 6 adjacent vertices around each vertex is calculated. The curvature is calculated using the discrete differential geometry method, and a 1-ring neighborhood is taken for each vertex to construct a local coordinate system, fit the quadratic surface, and solve the Gaussian curvature and mean curvature. The calculation formula of the comprehensive quality index Q is Q=w1ER+w2EA+w3EV, where ER, EA, and EV are the normalized errors of the edge length ratio, dihedral angle, and area ratio, respectively, and the weights w1=0.4, w2=0.4, and w3=0.2. Set three levels of quality thresholds: Q<0.6 for low quality, 0.6≤Q<0.8 for medium quality, and Q≥0.8 for high quality. Apply the Laplace smoothing algorithm to low-quality mesh elements, and the iterative formula is Vi'=Vi+λΣ(Vj-Vi) / n, where Vi is the vertex to be optimized, Vj is the adjacent vertex, n is the number of adjacent vertices, and λ is the smoothing factor, which is 0.5. Set the constraints of fixed position of boundary vertices and volume change rate less than 0.1%. Perform a quality assessment every 100 iterations during the optimization process, and stop the iteration when the change rate of the evaluation results is less than 1% for three consecutive times. Finally, generate a mesh quality analysis report, including the quality distribution histogram before and after optimization, the quality improvement percentage, and the comparison chart before and after optimization of 5 typical local areas.

[0027] Identify and locate the micro defect areas on the surface of metal materials, determine the refinement range and subdivision level of the local defect area according to the scale and distribution characteristics of the micro defect area, and dynamically adjust the grid density according to the refinement range and subdivision level of the local defect area to perform fine reconstruction of the local defect area.

[0028] The Gaussian curvature and the average curvature of each point on the surface of the metal material are calculated, and the maximum value of the Gaussian curvature and the average curvature is taken as the local curvature value; a curvature threshold is set according to the local curvature value, and the area exceeding the curvature threshold is marked as a potential defect area; the potential defect area is analyzed for a connected domain by an 8-neighborhood seed growing algorithm to obtain the boundary and area of ​​the defect; the geometric features of each defect are calculated for the boundary and area of ​​the defect, and the geometric features include area, perimeter, aspect ratio and direction; the aspect ratio and direction of the defect are calculated by a rectangular fitting algorithm, and the defects are classified and sorted; Priority is sorted according to the weighted sum of the defect area and the curvature value to determine the priority of the defect area that needs to be refined and reconstructed; for the defect area that needs to be refined and reconstructed, a multi-level refinement range is set, expanding from the center of the defect outward and divided into a core area, a transition area and an edge area; the subdivision level of each area is determined according to the scale and complexity of the defect; based on the refinement range and subdivision level, the local defect area is adaptively meshed; the mesh density is dynamically adjusted using an octree algorithm, starting from the top-level octree, and whether to continue subdividing is determined according to the curvature value in each sub-area until a preset maximum subdivision level is reached.

[0029] Specifically, the Gaussian curvature and mean curvature of each point on the surface of the metal material are calculated using the discrete differential geometry method, and the maximum value is taken as the local curvature value. The curvature threshold is set, and the area exceeding the threshold is marked as a potential defect area. The connected domain analysis is performed through the 8-neighborhood seed growth algorithm to determine the boundary and area of ​​the defect. According to the identified defect area, the geometric features of each defect are calculated, including area, perimeter, aspect ratio, and direction. The aspect ratio and direction of the defect are calculated using the rectangular fitting algorithm. The defects are classified and sorted, and the priority is sorted according to the weighted sum of the defect area and curvature value to determine the priority of the defect area that needs to be refined and reconstructed. For each defect area to be reconstructed, a multi-level refinement range is set, which expands outward from the defect center and is divided into a core area, a transition area, and an edge area. The radius of the core area is 1.2 times the equivalent diameter of the defect, 2 times the transition area, and 3 times the edge area. The subdivision level of each area is determined according to the scale and complexity of the defect. Based on the set refinement range and subdivision level, the local area of ​​the defect is adaptively refined, and the mesh density is dynamically adjusted using the octree algorithm. Starting from the top octree, the curvature value in each sub-area is used to decide whether to continue subdividing until the preset maximum subdivision level is reached. The core area has the highest mesh density, the transition area gradually decreases, and the edge area and the surrounding mesh transition smoothly. In the process of identifying micro-defects on the surface of metal materials, the local curvature of the surface point cloud data is first calculated using the discrete differential geometry method. For each point, a spherical neighborhood with a radius of 0.5 mm is selected, and the quadratic surface is fitted using the least squares method to calculate the Gaussian curvature and the average curvature. The maximum value of the two is taken as the local curvature value, and the curvature threshold is set to 0.5 mm^-1. For areas exceeding the threshold, the 8-neighborhood seed growth algorithm is used for connected domain analysis, and the minimum connected area is set to 0.1 mm2 to obtain the preliminary defect area. Then the geometric features of each defect are calculated, and the major and minor axes are calculated using the principal component analysis method to obtain the aspect ratio and direction. The defect priority is determined by the weighted sum of the area (weight 0.6) and the maximum curvature value (weight 0.4). For the highest priority defect, the radius of the core area is set to 1.2 times the equivalent diameter of the defect, 2 times in the transition area, and 3 times in the edge area. The octree algorithm is used for adaptive mesh refinement, with an initial mesh size of 1mm and a minimum mesh size of 0.01mm. Starting from the top layer, if the maximum curvature value in the sub-area is greater than the threshold corresponding to the current level, the subdivision continues, and the maximum subdivision level is set to 6. The maximum subdivision levels of the core area, transition area, and edge area are 6, 4, and 2, respectively, to ensure that the mesh density gradually decreases from the center to the outside to achieve a smooth transition.

[0030] S103, dividing the three-dimensional geometric model into regions, identifying and extracting surface defect regions, performing edge detection on the surface defect regions, and determining the precise geometric contours and spatial position information of the defects.

[0031] The discrete Laplace-Beltrami operator is used to calculate the vertex Gaussian curvature and mean curvature of the three-dimensional geometric model mesh to obtain vertex curvature data; based on the vertex curvature data, the model surface is segmented by a region growing algorithm, and if the curvature difference between adjacent vertices is less than a preset threshold, they are merged to obtain a set of regions with similar geometric features; for the set of regions, the local curvature histogram and shape index are extracted as feature vectors, and the support vector machine algorithm is used to extract and classify features of each region to determine whether it is a potential defect area; for the identified surface defect area, the three-dimensional Canny edge detection algorithm is used to extract the edge, and the boundary point set of the defect area is obtained by calculating the normal vector and the gradient of the curvature; for the boundary point set, the cubic B-spline interpolation method is used to perform curve fitting, the precise geometric contour of the defect is constructed, the spatial coordinates and geometric parameters of the contour are calculated, and the precise geometric contour and spatial position information of the defect are obtained.

[0032] Specifically, the mesh curvature analysis of the three-dimensional geometric model is performed, and the discrete Laplace-Beltrami operator is used to calculate the Gaussian curvature and mean curvature of each vertex. According to the distribution characteristics of the curvature value, the regional growing algorithm is used to perform preliminary segmentation of the model surface. The growth criterion is set to merge adjacent vertices when the curvature difference is less than the preset threshold, and a set of regions with similar geometric features is obtained. Based on the segmentation results, the local curvature histogram and shape index are extracted as feature vectors, and the support vector machine algorithm is used to extract and classify the features of each region. The region with abnormal geometric features is marked as a potential defect region. The final surface defect region is screened out by setting a threshold, and the centroid coordinates and main direction of the defect region are calculated. For the identified surface defect region, the three-dimensional Canny edge detection algorithm is used for edge extraction. The specific steps include Gaussian filtering, gradient calculation, non-maximum suppression, double threshold detection and edge connection. By calculating the normal vector and the gradient of the curvature, the boundary point set of the defect region is identified to form a preliminary defect contour. The obtained defect boundary point set was subjected to curve fitting, and the cubic B-spline interpolation method was used to construct the precise geometric contour of the defect. The control point interval was set to 1 / 3 of the average distance of the boundary point set. The spatial coordinates and geometric parameters of the contour, including the perimeter, area, and maximum diameter, were calculated, and the parametric equation of the defect contour on the model surface was calculated to obtain the precise geometric contour and spatial position information of the defect. In the mesh curvature analysis of the three-dimensional geometric model, the discrete Laplace-Beltrami operator was used to calculate the vertex curvature, and the vertex 1-ring neighborhood was selected to construct the covariance matrix, and the eigenvalue was solved to obtain the principal curvature. The curvature threshold was set to 0.5mm^-1, and the region growing algorithm was used for segmentation. The growth criterion was that the adjacent vertices were merged when the curvature difference was less than 0.1mm^-1. The features of the segmented region were extracted, and the 20-dimensional local curvature histogram and the 10-dimensional shape index were calculated to form a 30-dimensional feature vector. The support vector machine with RBF kernel function was used for classification, and the number of training samples was 1000. The optimal parameters C=10 and gamma=0.1 were selected by cross-validation. The threshold for determining the defect area is set to 0.8, and the coordinates of the center of gravity and the main direction eigenvector are calculated. In edge detection, the three-dimensional Canny algorithm is used, the Gaussian filter kernel size is 5×5×5, sigma=1.4, the gradient threshold low=0.1, high=0.3. The boundary point set uses cubic B-spline interpolation, and the control point interval is 1 / 3 of the average distance, about 0.1mm. The fitting accuracy is controlled within 0.01mm, and the contour perimeter, area, and maximum diameter are calculated. Finally, the least squares method is used to fit the parametric equation of the defect contour on the model surface, such as u(t)=a0+a1t+a2t^2, v(t)=b0+b1t+b2t^2, parameter t∈[0,1], which accurately describes the geometric contour and spatial position of the defect.

[0033] S104, establishing a parameterized defect model based on the geometric profile and spatial position information of the defect, including key dimension parameters of the depth, width and length of the defect, and obtaining a mathematical description model of the defect through a least squares fitting method.

[0034] The key feature point set of the defect is obtained by using the curvature extreme point detection algorithm, and the key feature point set includes the contour boundary point, the deepest point, the widest point and the endpoint. According to the key feature point set, principal component analysis is performed to determine the main direction of the defect, and a local coordinate system is established with the deepest point as the origin and the main direction as the z-axis. The defect contour is uniformly sampled using the local coordinate system to obtain dense three-dimensional point cloud data. For the three-dimensional point cloud data, the surface equation of the defect is fitted using the weighted least squares method, where the weight is inversely proportional to the distance from the point to the boundary. Based on the surface equation, the depth, width and length of the defect are calculated, and the parameter value is optimized by the gradient descent method to obtain an accurate parameterized defect model. The parameterized defect model is aligned with the original three-dimensional geometric model, and the iterative closest point algorithm is used for accurate alignment to calculate the position and direction of the defect in the global coordinate system. According to the position and direction, a 4x4 transformation matrix is ​​used to describe the position and orientation of the defect in the global coordinate system, and a complete defect mathematical description model is constructed in combination with key size parameters.

[0035] Specifically, according to the geometric contour and spatial position information of the defect, the curvature extreme point detection algorithm is used to extract the key feature point set of the defect, including the contour boundary point, the deepest point, the widest point and the endpoint. The main direction of the defect is determined by principal component analysis. The deepest point is taken as the origin and the main direction is the z-axis. A right-handed coordinate system is established as the local coordinate system. In the local coordinate system, the defect contour is uniformly sampled to obtain dense three-dimensional point cloud data. The surface equation of the defect is fitted by the weighted least squares method. The weight is inversely proportional to the distance from the point to the boundary. A suitable basis function, such as a quadratic surface or a B-spline surface, is selected. The root mean square error is used to evaluate the fitting accuracy to obtain a preliminary mathematical description of the defect. Based on the fitted surface equation, the depth, width and length of the defect are calculated. The bounding box size of the contour is used as the initial value. The parameter value is optimized by the gradient descent method. The error change is less than 0.1% for 5 consecutive iterations or the maximum number of iterations is 100 as the stop condition to minimize the fitting error and obtain an accurate parameterized defect model. The parametric defect model is registered with the original 3D geometric model, and the iterative closest point algorithm is used for accurate registration. The position and direction of the defect in the global coordinate system are calculated. The 4x4 transformation matrix is ​​used to describe the position and posture of the defect in the global coordinate system. Combined with the key size parameters, a complete defect mathematical description model is constructed, including position, direction and shape parameters. In the process of defect parametric modeling, the product of Gaussian curvature and mean curvature is first used as the curvature index, the threshold is set to 0.5mm^-2, and the curvature extreme point is extracted as the key feature point. The eigenvalue and eigenvector are calculated by principal component analysis, and the eigenvector corresponding to the maximum eigenvalue is selected as the main direction. The right-handed coordinate system is established with the deepest point as the origin (0, 0, 0) and the main direction as the z axis (0, 0, 1). In the local coordinate system, uniform sampling is performed at intervals of 0.1mm along the x-y plane to obtain a dense point cloud. The weighted least squares method was used to fit the quadratic surface z = ax^2 + by^2 + cxy + dx + ey + f, with weight w = 1 / (1 + d^2), and d was the distance from the point to the boundary. The Levenberg-Marquardt algorithm was used to solve the optimal parameters, with the maximum number of iterations set to 100 and the convergence threshold set to 1e-6. The root mean square error was calculated to evaluate the fitting accuracy. If it was greater than 0.05 mm, the sampling density was increased to 0.05 mm intervals and refitted. Based on the fitting equation, the extreme points of z were solved numerically to obtain the depth, and the maximum distance on the x-y plane was used to obtain the length and width. The gradient descent method was used to optimize these parameters, with the learning rate set to 0.01, and the error change was less than 0.1% for 5 consecutive iterations or when 100 iterations were reached. Finally, the iterative closest point algorithm was used for global registration, with the maximum number of iterations set to 50 and the convergence threshold set to 0.01 mm. A 4x4 transformation matrix T is obtained, in which a 3x3 submatrix R represents rotation and a 3x1 vector t represents translation, thus constructing a complete defect mathematical description model.

[0036] S105. Using finite element analysis software, the three-dimensional geometric model of the metal material and the mathematical description model of defects are imported, material properties, boundary conditions and load conditions are set, and a finite element analysis model including microscopic defects is established.

[0037] The three-dimensional geometric model of the metal material is obtained and imported into the finite element analysis software, and the feature recognition algorithm is used to identify the tiny faces and edges in the three-dimensional geometric model; if the size of the identified tiny faces and edges is less than the preset threshold, the tiny faces and edges are removed to obtain the optimized model topology. According to the optimized model topology and the preset defect mathematical description model, the micro defects are located on the three-dimensional geometric model; the micro defects are subtracted from the three-dimensional geometric model by CSG difference operation to obtain a complete geometric model containing micro defects. The complete geometric model containing micro defects is meshed; the local mesh encryption technology is used in the micro defect area, and the local mesh encryption technology includes: setting the minimum mesh size at the defect boundary, and increasing the mesh size of each layer outward by a specified multiple until the global mesh size is reached. The preset material property parameters are obtained, including elastic modulus and Poisson's ratio; the material property parameters are assigned to the meshed geometric model to obtain a finite element model with material properties. For the finite element model with material properties, a fixed constraint is applied at the bottom of the model; and a uniform pressure is applied at the top of the model to obtain a complete finite element analysis model containing microscopic defects.

[0038] Specifically, the three-dimensional geometric model of the metal material is imported into the finite element analysis software, and the feature recognition algorithm is used to identify and remove tiny faces and edges less than 0.1mm, optimize the model topology, eliminate repeated boundaries, small faces and sharp corners, and improve the geometric quality of the model. According to the defect mathematical description model, the micro defects are accurately located and constructed on the three-dimensional geometric model, and the defect geometry is subtracted from the original model using CSG difference operation to generate a complete geometric model containing micro defects, and the defect boundary curvature and depth distribution are recorded. The geometric model containing micro defects is meshed, and the local mesh encryption technology is used in the defect area. The minimum mesh size is set to 0.01mm at the defect boundary, and the mesh size of each layer is increased by 1.2 times outward until the global mesh size is reached. Set material properties, boundary conditions and load conditions, define the elastic modulus of the material as 200GPa and the Poisson's ratio as 0.3, assign these parameters to the mesh unit, apply fixed constraints at the bottom of the model, apply a uniform pressure of 100MPa at the top, define the load type and size, and complete the establishment of the finite element analysis model containing micro defects. In the process of establishing the finite element analysis model, the 3D geometric model in STL format was first imported, and the feature recognition algorithm was used to identify the tiny faces and edges. The threshold was set to 0.1 mm, and 18 tiny faces and 32 edges were removed. Then, according to the defect mathematical description model,

[0039] z=0.05*(x^2+y^2)-0.2, unit is mm, a micro defect is constructed at the position (50, 50, 0) on the model surface, and the defect body is subtracted from the original model using CSG difference operation. The maximum curvature of the defect boundary is recorded as 0.2mm^-1, and the maximum depth is 0.2mm. When meshing, the minimum grid size is set to 0.01mm at the defect boundary, and the grid size of each layer increases by 1.2 times outward until the global grid size reaches 0.5mm, generating a total of about 1 million tetrahedral units. When setting material properties, the elastic modulus is defined as 200GPa, the Poisson's ratio is 0.3, and the density is 7850kg / m^3. In terms of boundary conditions, a fixed constraint is applied at the bottom of the model to restrict all degrees of freedom; a uniform pressure of 100MPa is applied at the top. The load condition was set as static analysis, considering the large deformation effect, using the Newton-Raphson iteration method to solve the nonlinear equations, and the convergence criterion was set to force residual less than 0.1%. The final finite element analysis model contains 104,568 nodes and 573,921 units, ready for stress analysis and fatigue life assessment.

[0040] According to the nonlinear behavior of metal materials, the stress and strain data of the materials are obtained, the constitutive relationship model of the materials is established, and the material properties are obtained. Based on the three-dimensional geometric model containing defect information and material properties,

[0041] Set the boundary conditions. If the complexity of the boundary conditions is higher than the threshold, use the multi-point constraint method to process them and obtain the complete boundary conditions.

[0042] The stress and strain experimental data of metal materials in the material database are obtained, and the experimental data are smoothed using the Savitzky-Golay filtering algorithm to obtain the filtered stress-strain data. According to the filtered stress-strain data, the Levenberg-Marquardt algorithm is used for nonlinear regression fitting to determine the constitutive relationship model parameters of the material. The constitutive relationship model is combined with a three-dimensional geometric model containing defect information, and the initial boundary conditions are set to obtain a boundary condition description. The complexity of the boundary condition description is evaluated, and the ratio of the number of boundary nodes to the total number of nodes and the diversity index of the constraint type are calculated to determine whether the complexity is higher than the preset threshold. If the complexity is higher than the preset threshold, a multi-point constraint method is used to connect the complex boundary nodes with the reference nodes using the RBE2 unit to obtain an equivalent simplified constraint.

[0043] Specifically, the stress and strain experimental data of metal materials are obtained from the material database, and the Savitzky-Golay filtering algorithm is used to smooth the original data, remove outliers and noise, and obtain a reliable stress-strain curve. Based on the processed stress-strain data, the Levenberg-Marquardt algorithm is used for nonlinear regression fitting of the material constitutive relationship model, and a suitable nonlinear model function is selected, such as the Ramberg-Osgood model or the Johnson-Cook model. The model parameters are optimized by the genetic algorithm to obtain the material constitutive relationship expression and specific parameter values. The obtained material constitutive relationship model is combined with the three-dimensional geometric model containing defect information, and the symmetric boundary conditions are set according to the symmetry of the model. Uniform pressure is applied on the loading surface, and displacement constraints are set at the fixed end to complete the initial boundary condition setting. The complexity of the set boundary conditions is evaluated, and the ratio of the number of boundary nodes to the total number of nodes and the diversity index of the constraint type are calculated. If the complexity is higher than the preset threshold, the multi-point constraint method is used for processing, and the complex boundary nodes are connected to the reference nodes using the RBE2 unit, and the complex boundary conditions are converted into equivalent simplified constraints, and finally a complete boundary condition description is obtained. In the implementation process, 1000 sets of stress-strain experimental data points were first extracted from the material database, and the Savitzky-Golay filtering algorithm was used for smoothing. The window size was set to 15, the polynomial order was 3, and 950 sets of valid data points were obtained after processing. The Levenberg-Marquardt algorithm was used to fit the Ramberg-Osgood model, and the initial parameters were set as follows: elastic modulus E = 200 GPa, yield strength σ_y = 300 MPa, and hardening exponent n = 10. Through genetic algorithm optimization, the population size was set to 100, and 50 generations were iterated. Finally, the optimized parameters were obtained: E = 205 GPa, σ_y = 315 MPa, and n = 8.5. The constitutive model was combined with a three-dimensional geometric model. The model size was 100 mm × 50 mm × 20 mm, the defect was located in the center, the diameter was 2 mm, and the depth was 1 mm. Symmetric boundary conditions were set in the XY plane, the bottom surface was fixed, and a uniform pressure of 100 MPa was applied to the top surface. In the boundary condition complexity evaluation, the boundary node ratio is 0.15 and the constraint type diversity index is 0.8, which is higher than the preset threshold of 0.6. RBE2 units are used for multi-point constraint processing. One master node is selected for every 10 nodes in the complex boundary area. A total of 50 RBE2 units are set to reduce the number of boundary nodes from 15,000 to 1,500 and simplify the boundary condition description.

[0044] S106. Perform stress-strain analysis on the finite element analysis model of micro-defects, calculate the stress distribution and stress concentration factor around the defects, and use fracture mechanics theory to evaluate the impact of defects on fatigue crack initiation and propagation.

[0045] The finite element analysis model of micro-defects is meshed, and an adaptive mesh refinement technique based on stress gradient is used. When the unit stress gradient exceeds a preset threshold, the mesh is refined. According to the meshing result, the stress distribution field around the defect is calculated by the Newton-Raphson iterative solver. From the stress distribution field, the stress values ​​of the defect tip and the surrounding area are extracted, and the stress concentration coefficient K_t is calculated, where K_t is equal to σ_max divided by σ_nom, σ_max is the maximum local stress, and σ_nom is the nominal stress. The stress intensity factor at the defect tip is calculated using the J integral method. If the stress intensity factor exceeds the crack growth threshold ΔK_th of the material, the crack growth rate is calculated using the Paris formula. Based on the stress intensity factor and crack growth rate, combined with the fracture toughness and fatigue limit of the material, the Morrow equation is used to predict the fatigue crack initiation life. According to the fatigue crack initiation life, the crack growth life is calculated in combination with the S-N curve and Miners linear cumulative damage theory, and the evaluation results of the impact of defects on the fatigue performance of components are obtained.

[0046] Specifically, the finite element analysis model of micro-defects is meshed and solved. The adaptive mesh refinement technology based on stress gradient is used. When the unit stress gradient exceeds the threshold, the mesh is refined. The mesh is encrypted in the area around the defect. The stress distribution field around the defect is calculated by the Newton-Raphson iterative solver. According to the calculated stress distribution field, the stress values ​​of the defect tip and the surrounding area are extracted, and the stress concentration coefficient K_t=σ_max / σ_nom is calculated, where σ_max is the maximum local stress and σ_nom is the nominal stress. The maximum principal stress criterion or VonMises stress criterion is used to determine the stress concentration position. The stress intensity factor of the defect tip is calculated using the J integral method, and the maximum circumferential stress criterion is used to determine the crack propagation direction. When the stress intensity factor exceeds the crack propagation threshold ΔK_th of the material, it is considered that the crack begins to propagate, and the crack propagation rate is calculated using the Paris formula. Based on the calculated stress intensity factor and crack growth rate, combined with the fracture toughness and fatigue limit of the material, the Morrow equation is used to predict the fatigue crack initiation life, and the crack growth life is calculated by combining the S-N curve and Miners linear cumulative damage theory to evaluate the influence of defects on the fatigue performance of components. In the finite element analysis of micro-defects, the model is first meshed, and the initial mesh size is set to 0.5mm. Adaptive refinement based on stress gradient is adopted. When the unit stress gradient exceeds 200MPa / mm, refinement is performed, and the minimum mesh size is limited to 0.01mm. The Newton-Raphson iterative solver is used, and the convergence threshold is set to 0.1% and the maximum number of iterations is 100. After calculating the stress distribution around the defect, the stress value within 50μm of the defect tip is extracted, and the stress concentration coefficient K_t is calculated. The maximum stress at the defect is 450MPa, and the nominal stress in the far field is 150MPa, and K_t=3 is obtained. The VonMises stress criterion is used to determine the stress concentration position. The J integral was calculated using the virtual displacement method, and the integral path selected three concentric circles around the defect tip, with radii of 0.05mm, 0.1mm and 0.15mm respectively. The stress intensity factor KI = 25MPa√m was calculated, which exceeded the crack extension threshold ΔK_th = 15MPa√m of the material, and it was determined that the crack began to extend. The crack extension rate was calculated using the Paris formula da / dN = C(ΔK)^m, where C = 1e-11 and m = 3. Based on the Morrow equation and the material S-N curve, the fatigue crack initiation life was predicted to be 5e5 times. Combined with Miners linear cumulative damage theory, the total fatigue life was calculated to be 7.2e5 times, which was 40% lower than the life of defect-free components.

[0047] S107. According to the evaluation results, based on the Paris formula and cumulative damage theory, combined with the stress analysis results and material fatigue parameters, a fatigue life prediction model combining the influence of defects is established to calculate the crack propagation life, and realize the surface defect analysis of metal material fatigue life prediction.

[0048] According to the stress distribution data around the defect, the J integral value is extracted, and the stress intensity factor K is obtained through the plane strain relationship; if the J integral value is greater than the preset threshold, the stress intensity factor range ΔK at the defect tip is calculated; the crack growth rate model is established using the parameters C and m of the material Paris formula; the expansion life of the crack from the initial length a0 to the critical length ac is calculated using the adaptive Simpson integral method, and the expansion life is determined by the crack growth rate model; when the stress intensity factor range ΔK is less than the material crack expansion threshold ΔKth, the integral calculation is stopped; the crack initiation life is obtained by combining the material S-N curve and the Miner linear cumulative damage criterion, and the crack initiation life is determined by the stress amplitude and the actual number of cycles; the crack initiation life is added to the expansion life to obtain the total fatigue life; a fatigue life prediction model combined with the influence of surface defects is established, and the prediction model takes the defect size, shape factor and stress level as independent variables and the logarithmic life as the dependent variable; the prediction model parameters are determined by regression analysis; and the accuracy of the prediction model is evaluated by K-fold cross validation.

[0049] Specifically, according to the stress analysis results, the stress distribution data around the defect is extracted, and the J integral value is extracted by finite element post-processing. It is converted into the stress intensity factor K through the plane strain relationship, and the stress intensity factor range ΔK at the defect tip is calculated. Combined with the Paris formula parameters C and m of the material, a crack growth rate model is established. The adaptive Simpson integral method is used to calculate the crack growth life from the initial length a0 to the critical length ac based on the crack growth rate model, considering the influence of the stress amplitude and stress ratio in the load spectrum on the crack growth rate. When ΔK is less than the crack growth threshold ΔKth of the material, the integral calculation is stopped. Using the cumulative damage theory, combined with the S-N curve of the material and the Miner linear cumulative damage criterion, the stress amplitude is used to query the S-N curve to obtain the corresponding fatigue life, and the damage amount is calculated in combination with the actual number of cycles to obtain the crack initiation life. The total fatigue life is obtained by adding the crack initiation life and the propagation life. A fatigue life prediction model considering the influence of surface defects is established. A multivariate nonlinear regression model is established. The independent variables include defect size, shape factor and stress level, and the dependent variable is logarithmic life. The model parameters are determined by regression analysis. The prediction accuracy of the model is evaluated by K-fold cross validation to achieve quantitative prediction of fatigue life of metal materials. In the process of establishing the fatigue life prediction model, the stress distribution data within 50μm around the defect is first extracted from the finite element analysis results. The J integral value is calculated using the virtual displacement method. Three integral paths are selected with radii of 0.05mm, 0.1mm and 0.15mm respectively. After taking the average value, it is converted into stress intensity factor K through the plane strain relationship K=√(EJ / (1-ν^2)), where E is the elastic modulus 210GPa and ν is the Poisson's ratio 0.3. The stress intensity factor range ΔK=25MPa√m was calculated. Combined with the material Paris formula parameters C=1e-11 and m=3, the crack growth rate model da / dN=1e-11(ΔK)^3 was established. The adaptive Simpson integral method was used, with the initial step size set to 0.1mm and the error threshold set to 1e-6. The life of the crack from the initial length of 0.5mm to the critical length of 5mm was calculated, considering the influence of the stress ratio R=0.1. The integration was stopped when ΔK was less than the crack growth threshold ΔKth=5MPa√m, and the growth life was 1.5e5 times. Using the material S-N curve, the life corresponding to the stress amplitude of 350MPa was 2e5 times. Combined with the Miner criterion to calculate the damage amount of 0.75, the crack initiation life was 1.5e5 times. The total fatigue life was 3e5 times. A multivariate nonlinear regression model log(Nf)=β0+β1a+β2F+β3σ+β4(a*σ) was established, where a is the defect size, F is the shape factor, and σ is the stress level. The Levenberg-Marquardt algorithm was used to optimize the parameters, and a 10-fold cross validation was used, with an average prediction error of 15%.

[0050] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A surface defect analysis method combined with fatigue life prediction of metal materials, characterized in that: The method comprises: The surface of the metal material is scanned by a 3D laser scanner to obtain the surface morphology point cloud data. According to the density and distribution characteristics of the surface morphology point cloud data, an initial 3D model of the metal material surface is constructed by a triangular mesh reconstruction algorithm. The initial 3D model is meshed and refined, and the local area is reconstructed in detail to obtain a 3D geometric model of the metal material surface. The 3D geometric model contains microscopic surface defect information. Divide the 3D geometric model into regions, identify and extract surface defect regions, perform edge detection on surface defect regions, and determine the precise geometric contour and spatial location information of the defects; According to the geometric profile and spatial position information of the defect, a parameterized defect model is established, including the key dimension parameters of the defect's depth, width, and length. The mathematical description model of the defect is obtained through the least squares fitting method. Using finite element analysis software, the three-dimensional geometric model of metal materials and the mathematical description model of defects are imported, material properties, boundary conditions and load conditions are set, and a finite element analysis model including microscopic defects is established; Perform stress-strain analysis on the finite element analysis model of micro-defects, calculate the stress distribution and stress concentration factor around the defects, and use fracture mechanics theory to evaluate the impact of defects on fatigue crack initiation and propagation; According to the evaluation results, based on the Paris formula and cumulative damage theory, combined with the stress analysis results and material fatigue parameters, a fatigue life prediction model combining the influence of defects is established to calculate the crack propagation life, and realize the surface defect analysis of metal material fatigue life prediction.

2. The method according to claim 1, wherein: The method comprises: scanning the surface of the metal material with a three-dimensional laser scanner to obtain surface morphology point cloud data, and constructing an initial three-dimensional model of the metal material surface through a triangular mesh reconstruction algorithm according to the density and distribution characteristics of the surface morphology point cloud data, including: A high-precision three-dimensional laser scanner is used to perform high-density scanning on the surface of the metal material to obtain high-resolution surface morphology point cloud data, wherein the point cloud data includes three-dimensional coordinate information of the surface of the metal material; For the point cloud data, a valid point cloud data set is screened out, abnormal points and noise points are removed, and an optimized point cloud data set is obtained; According to the optimized point cloud data set, a principal component analysis method is used to calculate the normal vector of each point, a differential geometry method is used to calculate the curvature, and feature segmentation is performed on the point cloud according to the curvature threshold to identify the plane area, edge area and curved surface area; Based on the feature segmentation results, an initial triangular mesh is constructed, the mesh quality is optimized by edge flipping and vertex insertion, the model accuracy is improved by using a local quadratic surface fitting method, and a surface three-dimensional model is generated; The three-dimensional surface model is calibrated to ensure that the model is consistent with the actual surface size of the metal material, and the surface area and volume of the model are calculated to obtain standardized three-dimensional surface model data of the metal material.

3. The method according to claim 1, wherein: The initial three-dimensional model is meshed and refined, and the local area is reconstructed in detail to obtain a three-dimensional geometric model of the metal material surface. The three-dimensional geometric model contains microscopic surface defect information, including: Acquire an initial three-dimensional model, and calculate the mesh topological structure, geometric features and quality indicators of the initial three-dimensional model according to a preset threshold, wherein the quality indicators include mesh side length ratio, dihedral angle and area ratio; Determine whether the quality index meets the preset conditions, and if not, select the grid units that need to be optimized; Dividing the model surface according to the curvature value of the initial three-dimensional model to obtain a low curvature area and a high curvature area; The area with a curvature less than a preset first threshold is a low curvature area, and the area with a curvature greater than a preset second threshold is a high curvature area. Using a quadratic error metric algorithm to simplify the mesh of the low curvature region, and reducing redundant triangles through edge collapse operations; The high curvature area and feature boundaries are locally encrypted using a Loop subdivision algorithm to obtain an optimized global grid; Refined reconstruction is performed on the identified surface defect areas, and the mesh density is dynamically adjusted according to the local curvature value; Adaptive mesh refinement method is used to increase local mesh density, and non-uniform rational B-spline surface fitting algorithm is used to accurately describe the defect edge, thus obtaining a high-precision local model containing micro-scale surface defect information. The optimized global grid is merged with the high-precision local model, and a regional growing algorithm is used to achieve a seamless connection between the local fine model and the global model; In the transition area, the mesh density and topological structure of the transition area are adjusted by progressive mesh density adjustment to obtain a three-dimensional geometric model of the metal material surface that retains the overall morphology and contains microscopic defect information; It also includes: based on the acquired initial 3D model, performing mesh quality analysis on the 3D model, optimizing the triangular facets according to the obtained mesh quality evaluation results; identifying and locating the micro defect area on the surface of the metal material, determining the refinement range and subdivision level of the local defect area according to the scale and distribution characteristics of the micro defect area, and dynamically adjusting the mesh density according to the refinement range and subdivision level of the local defect area to perform fine reconstruction of the local defect area; The method of performing mesh quality analysis on the three-dimensional model based on the acquired initial three-dimensional model and optimizing the triangular facets according to the obtained mesh quality evaluation results specifically includes: acquiring mesh data of the three-dimensional model, and calculating the side length ratio, dihedral angle, and area ratio of each triangular facet in the mesh data to obtain geometric feature parameters of the triangular facet; calculating the Gaussian curvature and mean curvature of each vertex using a discrete differential geometry method according to the geometric feature parameters to obtain a mesh quality evaluation index; calculating a comprehensive quality index based on the side length ratio, the dihedral angle, and the area ratio, and determining the mesh unit to be a low-quality mesh unit if the comprehensive quality index is less than a preset multi-level threshold; performing local optimization on the low-quality mesh unit using a Laplace smoothing algorithm to improve the shape of the triangular facet by adjusting the vertex position; recalculating the mesh quality evaluation index of the locally optimized three-dimensional model, and determining the local optimization to be effective if the mesh quality evaluation index is greater than a preset threshold; and generating a mesh quality analysis report including a mesh quality distribution histogram, a quality improvement percentage, and a comparison chart before and after local optimization; The method of identifying and locating micro defect areas on the surface of metal materials, determining the refinement range and subdivision level of the local defect area according to the scale and distribution characteristics of the micro defect area, and dynamically adjusting the grid density according to the refinement range and subdivision level of the local defect area to perform fine reconstruction of the local defect area specifically includes: calculating the Gaussian curvature and the average curvature of each point on the surface of the metal material, taking the maximum value of the Gaussian curvature and the average curvature as the local curvature value; setting a curvature threshold according to the local curvature value, marking the area exceeding the curvature threshold as a potential defect area; performing a connected domain analysis on the potential defect area by using an 8-neighborhood seed growth algorithm to obtain the boundary and area of ​​the defect; calculating the geometric features of each defect according to the boundary and area of ​​the defect, and the geometric features The defects include area, perimeter, aspect ratio and direction; the aspect ratio and direction of the defects are calculated by a rectangular fitting algorithm, and the defects are classified and sorted; priority is sorted according to the weighted sum of the defect area and curvature value, and the priority of the defect area that needs to be refined and reconstructed is determined; for the defect area that needs to be refined and reconstructed, a multi-level refinement range is set, which expands outward from the center of the defect and is divided into a core area, a transition area and an edge area; the subdivision level of each area is determined according to the scale and complexity of the defect; based on the refinement range and subdivision level, the local area of ​​the defect is adaptively meshed; the octree algorithm is used to dynamically adjust the mesh density, starting from the top-level octree, and judging whether to continue subdividing according to the curvature value in each sub-area until the preset maximum subdivision level is reached.

4. The method according to claim 1, wherein: The method of dividing the three-dimensional geometric model into regions, identifying and extracting surface defect regions, performing edge detection on the surface defect regions, and determining the precise geometric contour and spatial position information of the defects includes: The discrete Laplace-Beltrami operator is used to calculate the vertex Gaussian curvature and mean curvature of the three-dimensional geometric model mesh to obtain vertex curvature data; According to the vertex curvature data, the model surface is segmented using a region growing algorithm, and adjacent vertices are merged if the curvature difference is less than a preset threshold, to obtain a set of regions with similar geometric features; For the region set, extract the local curvature histogram and shape index as feature vectors, use the support vector machine algorithm to extract features and classify each region to determine whether it is a potential defect region; For the identified surface defect area, the 3D Canny edge detection algorithm is used to extract the edge, and the boundary point set of the defect area is obtained by calculating the normal vector and the gradient of the curvature; For the boundary point set, a cubic B-spline interpolation method is used to perform curve fitting, construct an accurate geometric contour of the defect, calculate the spatial coordinates and geometric parameters of the contour, and obtain the accurate geometric contour and spatial position information of the defect.

5. The method according to claim 1, wherein: According to the geometric profile and spatial position information of the defect, a parameterized defect model is established, including key dimension parameters of the depth, width and length of the defect, and a mathematical description model of the defect is obtained by the least squares fitting method, including: A key feature point set of the defect is obtained by using a curvature extreme point detection algorithm, wherein the key feature point set includes a contour boundary point, a deepest point, a widest point, and an endpoint; Perform principal component analysis based on the key feature point set to determine the main direction of the defect, and establish a local coordinate system with the deepest point as the origin and the main direction as the z-axis; The local coordinate system is used to uniformly sample the defect contour to obtain dense three-dimensional point cloud data; For the three-dimensional point cloud data, a weighted least square method is used to fit the surface equation of the defect, wherein the weight is inversely proportional to the distance from the point to the boundary; Based on the surface equation, the depth, width and length of the defect are calculated, and the parameter values ​​are optimized by the gradient descent method to obtain an accurate parameterized defect model; Aligning the parameterized defect model with the original three-dimensional geometric model, using an iterative closest point algorithm for accurate alignment, and calculating the position and direction of the defect in a global coordinate system; According to the position and direction, a 4x4 transformation matrix is ​​used to describe the position and posture of the defect in the global coordinate system, and a complete defect mathematical description model is constructed in combination with key size parameters.

6. The method according to claim 1, wherein: The finite element analysis software is used to import the three-dimensional geometric model of the metal material and the defect mathematical description model, set the material properties, boundary conditions and load conditions, and establish a finite element analysis model containing microscopic defects, including: Obtaining a three-dimensional geometric model of a metal material and importing it into a finite element analysis software, and using a feature recognition algorithm to identify tiny faces and edges in the three-dimensional geometric model; If the size of the identified tiny faces and edges is smaller than a preset threshold, the tiny faces and edges are removed to obtain an optimized model topology structure; Locating microscopic defects on the three-dimensional geometric model according to the optimized model topology structure and the preset defect mathematical description model; Subtracting the microscopic defects from the three-dimensional geometric model through CSG difference operation to obtain a complete geometric model including the microscopic defects; Meshing the complete geometric model including the microscopic defects; A local mesh encryption technique is used in the micro defect area, wherein the local mesh encryption technique includes: setting a minimum mesh size at the defect boundary, and increasing the mesh size of each layer outward by a specified multiple until the global mesh size is reached; Get preset material property parameters, including elastic modulus and Poisson's ratio; Assigning the material property parameters to the meshed geometric model to obtain a finite element model with material properties; For the finite element model with material properties, a fixed constraint is applied to the bottom of the model; Apply uniform pressure on the top of the model to obtain a complete finite element analysis model including microscopic defects; The method also includes: obtaining stress and strain data of the material, establishing a constitutive relationship model of the material, and obtaining material properties according to the nonlinear behavior of the metal material; setting boundary conditions according to the three-dimensional geometric model containing defect information and the material properties; if the complexity of the boundary conditions is higher than a threshold, processing them by a multi-point constraint method to obtain complete boundary conditions, specifically including: obtaining stress and strain experimental data of the metal material in the material database, smoothing the experimental data by a Savitzky-Golay filtering algorithm, and obtaining filtered stress-strain data; performing nonlinear regression fitting by a Levenberg-Marquardt algorithm according to the filtered stress-strain data to determine the constitutive relationship model parameters of the material; combining the constitutive relationship model with the three-dimensional geometric model containing defect information, setting initial boundary conditions, and obtaining a boundary condition description; evaluating the complexity of the boundary condition description, calculating the ratio of the number of boundary nodes to the total number of nodes and the diversity index of the constraint type, and judging whether the complexity is higher than a preset threshold; if the complexity is higher than the preset threshold, using a multi-point constraint method, using RBE2 units to connect complex boundary nodes with reference nodes, and obtaining equivalent simplified constraints.

7. The method according to claim 1, wherein: The finite element analysis model of the micro-defect is subjected to stress-strain analysis, the stress distribution and stress concentration factor around the defect are calculated, and the influence of the defect on the initiation and propagation of fatigue cracks is evaluated by using fracture mechanics theory, including: The finite element analysis model of micro-defects is meshed, and adaptive mesh refinement technology based on stress gradient is used to refine the mesh when the element stress gradient exceeds the preset threshold; According to the meshing result, the stress distribution field around the defect is calculated by Newton-Raphson iterative solver; From the stress distribution field, the stress values ​​at the defect tip and the surrounding area are extracted to calculate the stress concentration factor K_t, where K_t is equal to σ_max divided by σ_nom, σ_max is the maximum local stress, σ_nom is the nominal stress; The stress intensity factor at the flaw tip is calculated using the J-integral method. If the stress intensity factor exceeds the material crack growth threshold ΔK_th, the crack growth rate is calculated using the Paris formula. Based on the stress intensity factor and crack growth rate, combined with the fracture toughness and fatigue limit of the material, the Morrow equation is used to predict the fatigue crack initiation life; According to the fatigue crack initiation life, the crack propagation life is calculated in combination with the S-N curve and Miners linear cumulative damage theory, and the evaluation result of the influence of defects on the fatigue performance of components is obtained.

8. The method according to claim 1, wherein: According to the evaluation results, based on the Paris formula and cumulative damage theory, combined with the stress analysis results and material fatigue parameters, a fatigue life prediction model combining the influence of defects is established to calculate the crack propagation life, and the surface defect analysis for predicting the fatigue life of metal materials is realized, including: According to the stress distribution data around the defect, the J integral value is extracted, and the stress intensity factor K is obtained through the plane strain relationship; If the J integral value is greater than a preset threshold, then calculating the defect tip stress intensity factor range ΔK; The crack growth rate model is established using the material Paris formula parameters C and m; The adaptive Simpson integral method is used to calculate the crack extension life from the initial length a0 to the critical length ac, where the extension life is determined by the crack growth rate model; When the stress intensity factor range ΔK is less than the material crack extension threshold ΔKth, the integral calculation is stopped; Combining the material S-N curve and Miner's linear cumulative damage criterion, the crack initiation life is obtained, and the crack initiation life is determined by the stress amplitude and the actual number of cycles; Adding the crack initiation life and the extension life to obtain a total fatigue life; Establishing a fatigue life prediction model incorporating the influence of surface defects, wherein the prediction model uses defect size, shape factor and stress level as independent variables and logarithmic life as a dependent variable; Determining the prediction model parameters by regression analysis; K-fold cross validation was used to evaluate the accuracy of the prediction model.

Images