Simulation intention-based CAE geometric model fault tolerance intelligent repair method and system

Abstract

The invention relates to the technical field of crossing of computer-aided engineering and artificial intelligence, discloses a CAE geometric model fault tolerance intelligent repair method and system based on simulation intention, and aims to solve the problems of insufficient precision of a key region and calculation redundancy of a non-key region caused by lack of simulation intention perception in the existing repair technology. The method comprises the following steps: carrying out standardization and defect identification on an imported CAD model; classifying the defects into a key class and a simplified class in combination with a simulation type, a precision requirement and a mechanics principle input by a user; performing high-fidelity repair on the key class, and performing adaptive simplification on the simplifiable class; and iterative optimization is realized through dual verification of geometry and mechanics. The system comprises a model import module, a defect recognition module, an intention analysis module, a layered repair module, a verification module and a result output module. According to the scheme, dynamic balance of precision and efficiency is achieved, and the automation level and simulation reliability of CAE pretreatment are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of computer-aided engineering (CAE) and intelligent geometric modeling, specifically relating to a fault-tolerant intelligent repair method and system for CAE geometric models based on simulation intent. Background Technology

[0002] In the field of Computer-Aided Engineering (CAE) simulation, the quality of the geometric model preprocessing directly determines the reliability and efficiency of subsequent mesh generation and physics field solving. Current mainstream workflows typically import the original geometric model (such as STEP or IGES format) from a Computer-Aided Design (CAD) system into the CAE platform and rely on general repair tools to clean up defects, generating watertight, conflict-free geometry suitable for simulation. For example, in boundary representation (B-Rep) structures, geometric repair operations are often based on topological connectivity and geometric continuity rules to automatically correct anomalies such as gaps, overlapping surfaces, or minor features, usually achieving model closure through local surface fitting or feature deletion. Summary of the Invention

[0003] At least one embodiment of the present invention provides a CAE geometric model fault-tolerant intelligent repair method based on simulation intent, comprising: acquiring the geometric model to be repaired; classifying defects in the geometric model according to preset mechanical and geometric rules, and determining whether they belong to critical defects; performing high-fidelity repair processing on areas determined to be critical defects, including curve fitting for gap-type topological defects and generating transition surfaces to achieve precise stitching, and NURBS surface reconstruction of local geometric gaps based on adjacent multi-ring topological elements to form continuous transition surfaces; further determining whether non-critical defects belong to simplifiable defects, and performing adaptive simplification processing on simplifiable defects; performing topological integrity verification, Hausdorff distance calculation, and low-density mesh parameterized simulation comparison analysis on the repaired model, and if the preset accuracy parameters are not met, reclassifying the corresponding defects and repeating the repair process.

[0004] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in one embodiment of the present invention, the classification of defects in the geometric model includes: extracting the topological information and corresponding geometric parameters of the vertices, edges, faces and loops of the geometric model under the B-Rep structure; quantitatively evaluating each defect based on stress concentration region identification, neutral axis position relationship, bending moment distribution or membrane stress distribution, combined with feature size and curvature index; marking defects that meet any of the following conditions as critical defects: the defect is located in a stress concentration region and the curvature is greater than a set threshold; for bending members, the defect is located within the range of ±1 / 3 of the cross-sectional height of the neutral axis or in the region of maximum bending moment; for tension / compression members, the membrane stress in the region where the defect is located exceeds a preset proportion of the material yield strength.

[0005] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in one embodiment of the present invention, the step of curve fitting and generating transition surfaces for gap-like topological defects to achieve precise stitching includes: retrieving all non-watertight edges in the B-Rep structure, identifying paired boundary edges with a spatial distance less than a preset threshold; extracting the three-dimensional geometric expression of each boundary edge and unifying the parameter directions of the two edges; performing equal arc length discrete sampling on the two boundary edges according to the parameter length to obtain the corresponding point sequence; calculating the midpoint of each pair of corresponding points, constructing a cubic NURBS intermediate curve using the midpoint set as interpolation points, and applying endpoint tangential constraints to make it consistent with the tangential direction of the adjacent parent surface boundary; using the dual-track sweep or Coons surface construction method, generating a transition surface with the original boundary curve and intermediate curve as guide curves, and applying G1 continuity conditions at the boundary; deleting the original non-watertight edges, adding the generated transition surface to the topological structure, updating the edge-face adjacency relationship, and verifying whether Euler's formula V-E+F satisfies the solid condition.

[0006] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in one embodiment of the present invention, the step of reconstructing a local geometric gap using NURBS surface based on adjacent multi-ring topological elements to form a continuous transition surface includes: locating the surface where the gap is located according to the defect database, and extracting three consecutive adjacent rings of topological elements around the gap to construct a local sub-topological region; extracting the set of closed boundary curves of the sub-region, parameterizing the boundary curves and unifying the parameter range; establishing a regular parameter grid within the sub-region, generating an initial control point grid based on the boundary curves and their tangential constraints, wherein the boundary control points coincide with the boundary curves, and the inner control points of the first layer are arranged along the boundary normal; constructing a bicubic NURBS surface, using the least squares fitting method to fit the surface to the original geometric trend within the sub-region, and applying continuity conditions of positional continuity and consistent first derivative direction at the boundary; calculating the principal curvature distribution of the reconstructed surface, adjusting the weights of corresponding control points for abnormal curvature points, and repeating the iteration until the curvature change meets the preset geometric tolerance.

[0007] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in an embodiment of the present invention, the further determination of whether non-critical defects belong to the simplifiable defect category includes: performing a geometric determination on each non-critical defect; if the proportion of its feature size to the model's reference size is less than a set threshold and it is located in a non-stressed area, then proceeding to the next determination; performing a mechanical prediction; if the defect is located outside the area of ​​the neutral axis ±2 / 3 section height of the bending member and the bending moment at its location is less than a set proportion of the maximum bending moment, then proceeding to the next determination; performing a simulation sensitivity analysis; verifying through parametric simulation whether the change in overall mechanical performance after feature deletion is less than a set percentage; if all three determinations are satisfied, then the defect is marked as a simplifiable defect.

[0008] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in an embodiment of the present invention, the adaptive simplification processing for simplifiable defects includes: for the rounded corner region of the non-stressed end, merging the rounded corner surface with the adjacent plane into a single plane. The merging operation is achieved by projecting the rounded corner surface onto the adjacent plane and resampling control points. After merging, the number of meshes in the region is reduced by a set ratio, and the maximum distance deviation between the merged plane and the original feature does not exceed a set tolerance. For the small gaps in the non-stressed region, if the pre-simulation shows that its impact on the overall mechanical performance is less than a set threshold, the repair operation is skipped, and the original geometric state is retained.

[0009] For example, in the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in one embodiment of the present invention, the step of performing topological integrity verification, Hausdorff distance calculation, and low-density mesh parameterized simulation comparative analysis on the repaired model includes: traversing all edges and faces of the repaired model to verify whether there are suspended edges, non-manifold vertices, or non-closed bodies; calculating the Hausdorff distance between the models before and after repair to determine whether the geometric deviation is within the allowable range; generating low-density tetrahedral meshes for the models before and after repair under the same boundary conditions and performing linear static simulations to extract key response quantities such as maximum displacement and maximum stress for comparison; if any indicator exceeds the preset accuracy parameter, the corresponding defect is reclassified as a critical category or a simplifiable category, and the process returns to the high-fidelity repair or adaptive simplification steps.

[0010] For example, in one embodiment of the present invention, the CAE geometric model fault-tolerant intelligent repair method based on simulation intent also includes a model preprocessing step: importing a CAD geometric model in STEP or IGES format; converting the geometric model into a boundary representation B-Rep structure; extracting the topological connection relationships and corresponding geometric parameters of all vertices, edges, faces, and loops; deleting duplicate or redundant topological elements; and unifying the model unit system to the International System of Units (SI) or user-specified units.

[0011] For example, in one embodiment of the present invention, the CAE geometric model fault-tolerant intelligent repair method based on simulation intent also includes a defect identification database construction step: identifying non-watertight edges, isolated surfaces, or invalid loops based on topological connectivity analysis; determining whether there are geometric discontinuities or micro gaps based on the angle between the normal vectors of adjacent surfaces and the shortest edge-edge distance; determining whether there are micro features based on the ratio of feature size to the overall size of the model; extracting stress concentration coefficients based on finite element pre-analysis results and identifying high stress gradient regions; and storing the above defect types and their quantitative indicators in a structured database for subsequent defect classification and repair strategy selection.

[0012] At least one embodiment of the present invention provides a CAE geometric model fault-tolerant intelligent repair system based on simulation intent, comprising: an acquisition unit, a critical defect judgment unit, a high-fidelity repair processing unit, a simplifiable defect judgment unit, and an adaptive simplification processing unit. The acquisition unit is used to import and parse the CAD geometric model to generate a B-Rep structure; the critical defect judgment unit is used to classify defects according to mechanical and geometric rules; the high-fidelity repair processing unit is used to perform gap stitching and notch surface reconstruction on critical defects; the simplifiable defect judgment unit is used to perform secondary classification on non-critical defects; and the adaptive simplification processing unit is used to perform geometric simplification or skip repair on simplifiable defects.

[0013] At least one embodiment of the present invention provides an electronic device, including a CAE geometric model fault-tolerant intelligent repair system based on simulation intent provided in any embodiment of the present invention.

[0014] At least one embodiment of the present invention provides an electronic device, including a processor and a memory. The memory stores at least one computer program, which, when executed by the processor, implements the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in any embodiment of the present invention.

[0015] At least one embodiment of the present invention provides a non-transitory computer-readable storage medium for non-transitory storage of computer-readable instructions, which, when executed by a computer, implement the CAE geometric model fault-tolerant intelligent repair method based on simulation intent provided in any embodiment of the present invention. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of the CAE geometric model fault-tolerant intelligent repair method based on simulation intent in an embodiment of the present invention. Detailed Implementation

[0017] This invention provides a method and system for intelligent fault-tolerant repair of CAE geometric models based on simulation intent. The specific embodiments of this invention are described in detail below with reference to the accompanying drawings. Figure 1 As shown, the overall process of this invention includes original geometric model input, B-Rep structure conversion module, defect identification and database construction module, key class defect judgment module, high-fidelity repair processing module, simplified class defect judgment module, adaptive simplification processing module, post-repair model verification module, feedback reclassification interface, and final repair model. The modules are sequentially connected via data flow, forming a closed-loop processing logic: the original geometric model is first converted into a boundary representation (B-Rep) structure by the B-Rep structure conversion module; then it is sent to the defect identification and database construction module for topological and geometric defect identification, and a structured defect database is established; this database is passed as input to the critical defect judgment module, which performs preliminary classification of defects according to mechanical and geometric rules; the parts judged as critical defects enter the high-fidelity repair processing module for precise repair; non-critical defects are passed to the simplifiable defect judgment module for secondary judgment; if they are marked as simplifiable defects, the adaptive simplification processing module performs simplification operations; all repaired or simplified results are summarized to the repaired model verification module for multi-dimensional verification; if any verification index exceeds the preset accuracy parameter, the corresponding defect is reclassified through the feedback reclassification interface and returned to the high-fidelity repair processing module or the adaptive simplification processing module until the accuracy requirements are met, and finally the final repaired model is output.

[0018] During the model preprocessing stage, the original geometric model is imported into the system in STEP or IGES format, and the B-Rep structure conversion module 2 parses it into a standard B-Rep structure. This structure contains four basic topological elements: vertices, edges, faces, and loops. Each element is associated with its corresponding geometric parameters (such as curve type, surface type, control points, weights, node vectors, etc.). The B-Rep structure conversion module also performs redundant topological element removal operations, such as merging collinear edges, removing duplicate vertices, eliminating zero-area patches, and unifying the model's unit system to the International System of Units (SI) or user-specified units to ensure consistency and numerical stability in subsequent processing.

[0019] After receiving the B-Rep structure, the defect identification and database construction module first identifies non-watertight edges (i.e., edges not shared by two faces), isolated faces (faces not forming a closed body with other faces), or invalid loops (such as self-intersecting loops or empty loops) based on topological connectivity analysis. Secondly, it determines the existence of geometric discontinuities or micro-gaps based on the angle between the normal vectors of adjacent faces and the shortest edge-to-edge distance: when the angle between the normal vectors of two adjacent faces at a common edge is greater than a set threshold (e.g., 5 degrees) and the shortest edge-to-edge distance is less than 0.1 mm, it is marked as a geometric discontinuity defect. Thirdly, it calculates the ratio of feature dimensions (e.g., chamfer radius, hole diameter, boss height) to the overall model reference dimensions (e.g., the length of the bounding box diagonal); if the ratio is less than 0.001, it is determined to be a micro-feature. Finally, it calls the lightweight finite element solver to perform pre-analysis of the model, applies typical loads under fixed boundary conditions, extracts the Stress Concentration Factor (SCF), and marks regions with an SCF greater than 3.0 as high stress gradient regions. All the above-mentioned defect types and their quantitative indicators (including location coordinates, topology ID, geometric parameters, and mechanical response values) are stored in a structured database as the basis for subsequent classification and repair strategy selection.

[0020] After receiving the defect database, the critical defect judgment module performs a three-stage judgment logic for each defect. First, it extracts the B-Rep topology information and geometric parameters of the defect's location; then, it performs a quantitative evaluation based on the simulation intent: if the defect is located in a high stress gradient region and its local curvature is greater than 0.5 mm... -1 If a defect is located within ±1 / 3 of the section height of the neutral axis, or within the region of maximum bending moment (determined by obtaining the bending moment distribution cloud map through pre-analysis), it is marked as a critical defect. For tension / compression members, if the membrane stress in the defect area exceeds 70% of the material's yield strength (this percentage is configurable), it is also marked as a critical defect. The above criteria are combined using "OR" logic; satisfying any one of them triggers the critical defect marking.

[0021] The high-fidelity repair module performs two types of repair operations for critical defects. For gap-type topological defects (such as openings formed by non-watertight edges), it first searches for all non-watertight edges in the B-Rep structure, identifying paired boundary edges with a spatial Euclidean distance of less than 0.5 mm. For each boundary edge, its 3D NURBS curve representation is extracted, and the parameter directions of the two edges are unified through parameter remapping (making the start points correspond to the start points and the end points to the end points). Then, equal-arc-length discrete sampling is performed on the two boundary edges according to the parameter length. The number of sampling points is dynamically determined based on the gap length (one point is taken every 1 mm, with a minimum of 5 points), resulting in two sets of corresponding point sequences P1(i) and P2(i). The midpoint M(i) of each set of corresponding points is calculated as M(i) = (P1(i) + P2(i)) / 2. Using the set of midpoints as interpolation points, a cubic NURBS intermediate curve C is constructed. m (u), and apply tangential constraints at the endpoints: C m '(0) = (T1(0) + T2(0)) / |T1(0) + T2(0)|, C m '(1) = (T1(1) + T2(1)) / |T1(1) + T2(1)|, Where T1 and T2 are the tangent vectors of the two original boundary edges at their endpoints, respectively; then, a dual-track sweep method is used, with the original boundary curves C1(u) and C2(u) and the intermediate curve C m (u) is the guiding curve. A transition surface S(u,v) is generated, and G1 continuity conditions are applied at the boundary: ∂S / ∂v|{v=0} is coplanar with the principal normal vector of C1(u), and ∂S / ∂v|{v=1} is coplanar with the principal normal vector of C2(u). Finally, the original non-watertight edges are deleted, the generated transition surface is added to the B-Rep topology, the edge-face adjacency table is updated, and the Euler formula V-E+F=2 (for simply connected entities) is verified. If it is not valid, the control point weights are adjusted and the surface is regenerated.

[0022] For local geometric gaps (such as holes caused by missing surfaces), the high-fidelity repair module first locates the surface where the gap is located based on the defect database, and extracts three consecutive adjacent rings of topological elements around the gap: the first ring consists of edges and surfaces directly connected to the gap boundary, the second ring consists of surfaces adjacent to the first ring, and the third ring consists of surfaces adjacent to the second ring, thus constructing a local sub-topological region; extracts the set of closed boundary curves {C_bj} of this sub-region, performs parameterization on each boundary curve (using chord length parameterization), and unifies the parameter range to [0,1]; establishes an m×n regular parameter grid (m,n≥4) within the sub-region, and generates an initial control point grid based on the boundary curves and their tangential constraints: the boundary control points are strictly located on the boundary curves, and the first layer of inner control points are offset by a distance d=0.1×L (L is the local feature size) along the average normal direction of the boundary curve; constructs a bicubic NURBS surface S(u,v), and uses the least squares fitting method to fit the original geometric trend within the sub-region (i.e., fit the geometric shape represented by the third ring of topological elements), and applies positional continuity (C) at the boundary. 0 And the direction of the first derivative is consistent (G) 1 The continuity condition of the surface is then determined; subsequently, the principal curvature distributions κ1(u,v) and κ2(u,v) of the reconstructed surface are calculated, and if |κ1-κ2|>0.3 mm, the condition is met. -1 For abnormal curvature points, adjust the weights of the corresponding control points (increase the weights to enhance local rigidity), and repeat the surface generation and curvature calculation process iteratively until the curvature change meets the preset geometric tolerance (e.g., maximum curvature gradient ≤ 0.1 mm). -2 ).

[0023] Non-critical defects are passed to the defect judgment module that can simplify the judgment process, such as... Figure 1 As shown. This module performs three judgments sequentially for each non-critical defect: First, a geometric judgment is performed. If the ratio of the defect feature size (such as the chamfer radius r) to the model's reference size D (such as the maximum contour size) r / D < 0.005, and it is located in a non-stressed area (confirmed by pre-analysis that the von Mises stress in this area is <10 MPa), then proceed to the next step; Second, a mechanical judgment is performed. If the defect is located outside the area of ​​the bending member's neutral axis ±2 / 3 section height, and the bending moment M at its location is < If the value is 0.2 × M_max (where M_max is the maximum bending moment), proceed to the next step. Finally, perform a simulation sensitivity analysis: Perform linear static simulations on the original model and the simplified model after removing the defect in a low-density mesh (the element size is 1 / 5 of the feature size). Calculate the relative change rates of key response quantities (such as maximum displacement δ and maximum stress σ) |(δ_simp-δ_orig) / δ_orig| and |(σ_simp-σ_orig) / σ_orig|. If both are less than 2%, all three criteria are met, and the defect is marked as a simplifiable defect.

[0024] The adaptive simplification module performs differentiated processing on simplifiable defects. For non-load-bearing end fillet regions (such as shaft end chamfers), the fillet surface and adjacent planes are merged into a single plane: first, all control points of the fillet surface are projected onto the adjacent plane along its average normal to obtain a projection point set; then, the projection point set is delaunay triangulated, and resampled to generate a new planar control mesh, reducing the number of control points by more than 50%; after merging, the number of meshes in this region is significantly reduced, and the maximum Hausdorff distance between the merged plane and the original fillet feature does not exceed 0.05 mm (this tolerance is configurable). For small gaps in non-load-bearing areas (such as assembly gaps), if pre-simulation shows that their impact on overall mechanical properties is less than 1.5% (e.g., stiffness change <1.5%), the repair operation is skipped, the original geometry is preserved, and unnecessary geometric modifications are avoided to prevent the introduction of new errors.

[0025] The post-repair model verification module performs triple verification on the repaired model. First, it checks topological integrity: traversing all edges to check if each edge is shared by exactly two faces; traversing all vertices to check for non-manifold vertices (i.e., edges with ≠2 connected faces intersect at the vertex); and checking for volume closure (i.e., no overhanging edges or open shells). If a topological error is found, it returns to the high-fidelity repair module. Second, it calculates the Hausdorff distance between the pre- and post-repair models: using the pre-repair model as a reference, it calculates the shortest distance from each point on the post-repair model to the reference model, taking the maximum value as the Hausdorff distance. If this distance > 0.1 mm (or the user-defined tolerance), the geometric deviation is considered excessive. Finally, it performs low-density mesh parametric simulation comparison analysis: under the same boundary conditions and loads, it generates tetrahedral meshes (unit size uniformly 5 mm) for both the pre- and post-repair models, performs linear statics solutions, and extracts key response quantities (such as maximum displacement, maximum principal stress, and strain energy). If the relative error of any response quantity > 3%, the mechanical performance deviation is considered excessive. If any verification fails, the corresponding defect area will be reclassified through the feedback reclassification interface: if the geometric deviation is large but the mechanical impact is small, it may be classified as a simplified class; if the mechanical impact is large, it will be forcibly classified as a critical class and returned to the corresponding repair module for reprocessing until all verification indicators meet the preset accuracy parameters. Finally, the final repair model will be output for subsequent CAE mesh generation and simulation analysis.

[0026] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principles of this invention are further supplemented below with a specific application scenario.

[0027] In the CAE simulation preprocessing of the automotive engine bracket, the original geometric model was imported into the system in STEP format. This model contained multiple casting fillets, assembly gaps, and missing local surfaces due to CAD modeling errors. First, the B-Rep structure conversion module parsed the original geometric model into a standard boundary representation structure, identifying four types of topological elements: vertices, edges, faces, and loops. Simultaneously, redundancy cleanup was performed: for example, merging two collinear edges with a sum of lengths less than 0.01 mm into a single edge, removing zero-area triangular faces remaining from Boolean operations, and unifying all geometric units to millimeters to ensure the stability of subsequent numerical calculations.

[0028] Subsequently, the defect identification and database construction module performed a multi-dimensional scan of the converted B-Rep structure. At the topological level, three non-watertight edges (i.e., open edges referenced by only one face) were detected in the support base area, forming an open-type gap. At the geometric level, a pair of adjacent faces were found at the connection between the support rib and the main shell, with the angle between the normal vectors on both sides of their common edge reaching 8.2 degrees, and the minimum edge-to-edge distance being 0.06 mm, satisfying the geometric discontinuity defect criterion. At the same time, there were multiple chamfers with a radius of 0.15 mm at the top of the support, while the diagonal length of the model bounding box was 320 mm, with a scale of 0.00047, which was below the 0.001 threshold and was marked as a minor feature. In addition, after applying typical suspension loads through a lightweight finite element solver, the stress concentration factor (SCF) around the bolt holes of the support was identified as reaching 3.8, exceeding the 3.0 threshold. Therefore, this area was marked as a high stress gradient region, and its topological IDs were recorded as Face_217 to Face_223.

[0029] After receiving the above defect data, the critical defect judgment module classifies it. For the minor chamfer defects around the bolt holes, because they are located in a high stress gradient region with SCF > 3.0, and the local curvature is calculated to be 0.62 mm... -1 (greater than 0.5 mm) -1 The defects that meet the first criterion are marked as critical defects. As for the gap formed by the non-watertight edge of the base, although it is in a low-stress area (pre-analysis shows von Mises stress < 5 MPa), it is also classified as a critical defect because it is located on the main load-bearing path of the support, and the bending moment distribution cloud map shows that the bending moment value in this area is 0.85 times M_max, exceeding the 0.2 × M_max threshold. Furthermore, it is within ±1 / 3 of the height of the neutral axis of the bending section. Both types of defects are then transferred to the high-fidelity repair module.

[0030] The high-fidelity repair processing module includes: firstly, identifying two non-watertight boundary edges, Edge_89 and Edge_92, with a spatial distance of 0.06 mm, and extracting their NURBS expressions C1(u) and C2(u) respectively; secondly, remapping the parameters so that the starting points of the two edges correspond to the center direction of the bolt holes and the ending points correspond to the outer edge of the bracket; thirdly, sampling one point every 1 mm, taking 5 corresponding point pairs on a 0.42 mm long gap, and calculating the midpoint sequence M(i); and finally, constructing a cubic NURBS intermediate curve C with endpoint tangential constraints based on this sequence. m (u), where the endpoint tangents are synthesized from the original boundary edge derivative vectors; subsequently, C1, C2, C m To guide the trajectory, a dual-track sweep is used to generate the transition surface S(u,v), and at the boundaries of v=0 and v=1, ∂S / ∂v is forced to be coplanar with the principal normal of the original boundary, thus realizing G. 1 The surface is continuous; finally, the new surface is added to the B-Rep structure, the original open edges are deleted, and the Euler formula V-E+F=2 is verified to confirm the restoration of the solid closure.

[0031] For the local geometric gaps around the bolt holes caused by surface defects, the high-fidelity repair module performs NURBS reconstruction: starting from the gap boundary, it expands outward to extract three rings of topological surfaces (a total of 12 adjacent surfaces) to construct a local sub-region; it extracts the closed boundary curve set {C_b1, C_b2} and parameterizes the chord length to [0,1]; it establishes a 6×6 control mesh, with the boundary control points strictly falling on the original boundary, and the first inner layer control points offset along the average normal by d=0.1×L=0.03mm (L is taken as the local hole diameter of 3mm); after constructing the bicubic NURBS surface, the principal curvature field is calculated, and two locations |κ1-κ2|=0.35 mm are found. -1 The outlier was identified, and the corresponding control point weight was increased from 1.0 to 1.8 to enhance local rigidity; after two iterations, the maximum curvature gradient decreased to 0.08 mm. -2 It satisfies ≤0.1 mm -2 Tolerance requirements must be met to achieve high-fidelity restoration.

[0032] Meanwhile, several chamfers with r=0.15mm at the top of the support were transferred to the simplified defect judgment module. Geometrically, r / D=0.00047<0.005, and pre-analysis showed that the stress in this area was <8 MPa; mechanical prediction indicated that it was located outside the neutral axis ±2 / 3 of the height, and the bending moment was only 0.12 times M_max; further comparison of the original and simplified models under a low-density mesh (element size 0.03mm) showed that the maximum displacement change rate was 1.3% and the maximum stress change rate was 1.7%, both less than the 2% threshold, so it was marked as a simplified defect.

[0033] The adaptive simplification module performs a plane merging operation on this type of chamfer: all NURBS control points of the rounded corner surface are projected along their average normal to the adjacent principal plane to obtain a set of projected points; after Delaunay triangulation of this set of points, it is resampled into a 4×4 planar control mesh, reducing the number of control points from the original 25 to 16, a reduction of 36%; according to Hausdorff distance calculation, the maximum deviation between the simplified plane and the original chamfer is 0.042mm, which is less than the tolerance of 0.05mm and meets the geometric fidelity requirements.

[0034] Finally, the repaired model verification module performed triple verification: topology verification confirmed that all edges were shared by two key points and there were no non-manifold vertices; the maximum Hausdorff distance was 0.09 mm, which did not exceed the 0.1 mm limit; simulation with a low-density mesh (5 mm element size) showed that the relative error of the maximum principal stress before and after repair was 2.4%, and the strain energy error was 1.9%, both below the 3% threshold. Therefore, there was no need to trigger the feedback reclassification interface, and the final repaired model was directly output for subsequent high-precision mesh generation.

[0035] All content not described in detail in this specification belongs to existing technology known to those skilled in the art, and the implementation of each algorithm module can be based on commercial geometric kernels such as OpenCASCADE, ACIS, or Parasolid. The finite element solver involved in this technical solution can adopt open source or commercial software interfaces, and its specific model parameters are not limited; conventional configurations are sufficient. 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault-tolerant intelligent repair method for CAE geometric models based on simulation intent, characterized in that, Includes the following steps: Obtain the geometric model that needs to be repaired; Determine whether a defect in a geometric model is a critical defect based on the following rules: a) Defect is located in a stress concentration area with characteristic dimension < 3% of the model base dimension but curvature > 0.5 mm -1 , the curvature is based on B-Rep face principal curvature calculation; b) For bending members, the defect is located within the range of ±1 / 3 of the section height of the neutral axis, or in the region of maximum bending moment; c) For tension / compression members, the defect is located in the region where the membrane stress is greater than 30% of the material's yield strength; Perform high-fidelity repair on critical defects, including: Curve fitting is performed on gap-type topological defects, and transition patch surfaces are generated to achieve precise stitching. The local geometric gaps are reconstructed using NURBS surfaces based on adjacent multi-cycle topological elements to form a continuous transition surface.

2. The method according to claim 1, characterized in that, The method further includes: Determine whether a defect in a geometric model is a simplifiable defect based on the following rules: a) Geometric determination: The feature size accounts for less than 3% of the model's base size and is located in a non-stressed area; b) Mechanical prediction: For bending members, the defect is located outside the area of ​​±2 / 3 of the section height of the neutral axis, and the bending moment is less than 10% of the maximum bending moment; c) Simulation sensitivity analysis: Parametric simulation verified that the change in mechanical properties after feature deletion was <2%; Perform adaptive simplification processing on simplifiable defects, including: For the non-stressed end, the rounded corner is merged with the adjacent plane into a single plane. After merging, the number of meshes in the region is reduced by 60%, and the maximum distance deviation between the plane and the original feature is ≤0.02mm. For minor gaps in non-stressed areas, if the pre-simulation shows that their impact on mechanical properties is less than 1%, then no repair will be performed.

3. The method according to claim 1, characterized in that, The method also includes verification and iteration of the repair results: The repaired model is subjected to topological integrity verification and Hausdorff distance calculation. The model before and after repair is compared and analyzed by parametric simulation with low-density grid. When the classification result does not meet the preset accuracy parameters, the corresponding defect is reclassified.

4. The method according to claim 1, characterized in that, The method also includes a model preprocessing step: Import the CAD geometric model in STEP or IGES format, convert the geometric model into a boundary representation B-Rep structure, extract the topological information and corresponding geometric parameters of vertices, edges, faces and loops, delete duplicate topological elements and unify the model units.

5. The method according to claim 1, characterized in that, The method also includes a defect identification database construction step: Determine whether topological defects exist based on topological connectivity analysis; Determine whether there are geometric defects based on the angle between the normal vectors of adjacent faces and the shortest distance between edges; Determine the presence of minor feature defects based on the feature size ratio; Determine whether there are defects in critical areas based on the stress concentration factor; Based on historical data, a defect database containing defect types and their quantitative indicators will be established.

6. The method according to claim 1, characterized in that, The process of curve fitting and generating transition patch surfaces for gap-like topological defects to achieve precise stitching includes: Step A1: Extracting the gap boundary In the B-Rep structure, all non-watertight edges are retrieved, and boundary edges that exist in pairs and whose spatial distance is less than a preset threshold are identified. The three-dimensional geometric representation of each edge is extracted, and the orientation of the two edges is unified to make their parameter orientation consistent. Step A2: Establishment and sampling of corresponding points The two sides are sampled discretely with equal arc length according to the parameter length to obtain the corresponding point sequence P1ᵢ = C1(uᵢ), P2ᵢ = C2(vᵢ), i=1...n; the set of corresponding point pairs {(P1ᵢ, P2ᵢ)} is established; the distance dᵢ between corresponding points is calculated for subsequent fitting constraints. Step A3: Constructing the best-fit intermediate curve Calculate the midpoint Mᵢ = (P1ᵢ + P2ᵢ) / 2 for each set of corresponding points; take the point set {Mᵢ} as the interpolation points, and construct a cubic NURBS curve C using the least square method m (w); set the end point tangential constraint so that C m (w) is consistent with the adjacent parent surface boundary at both ends. Step A4: Generating Transition Surfaces Three guide curves are constructed: C1(u), C2(v), C m (w); the transition surface S(u, v) is generated by using the double rail sweep or Coons surface construction method; the boundary constraint condition is applied when the surface is generated: S(u,0)=C1(u) S(u,1)=C2(u) The surface satisfies the G1 continuity condition at the boundary; Step A5: Topology Update and Stitching Delete the existing non-watertight edges; add the generated transition surface to the topology; update the edge-face adjacency relationship; perform the B-Rep reconstruction operation to verify whether the Euler relation V-E+F satisfies the solid condition.

7. The method according to claim 1, characterized in that, The method of reconstructing local geometric gaps using NURBS surfaces based on adjacent multi-cycle topological elements to form a continuous transition surface includes: Step B1: Identification and Expansion of the Gap Region Locate the surface where the gap is located based on the defect database; extract three consecutive adjacent topological elements around the gap to construct a local sub-topological region Ω; Step B2: Boundary Curve Extraction Extract the set of boundary closed curves of the sub-region Ω; parameterize the boundary curves and unify the parameter range; calculate the normal and tangent vectors of the boundary curves at the connection points; Step B3: Control Point Mesh Generation Establish a regular parameter mesh within the sub-region Ω; generate an initial control point mesh based on the boundary curve and its tangential constraints; the control point layout follows the following: boundary control points coincide with the boundary curve; the first layer of inner control points are arranged along the boundary normal direction. Step B4: NURBS Surface Fitting Construct a bicubic NURBS surface S(u,v); employ the least squares fitting method to fit the original geometric trend within the sub-region of the surface; apply continuity conditions at the boundary: maintain positional continuity with adjacent parent surfaces; control the direction of the first derivative to be consistent; Step B5: Curvature Smoothing and Error Control Calculate the principal curvature distribution of the reconstructed surface; adjust the weights of the corresponding control points for abnormal curvature points; repeat the iteration until the curvature change meets the preset geometric tolerance.

8. A CAE geometric model fault-tolerant intelligent repair system based on simulation intent, characterized in that, include: The acquisition unit is used to acquire the geometric model that needs to be repaired; The critical defect judgment unit determines whether a defect in the geometric model is a critical defect based on the following rules: a) The defect is located in a stress concentration region, with a feature size less than 3% of the model's base size but a curvature greater than 0.5 mm. -1 The curvature is calculated based on the principal curvature of the B-Rep surface; b) For bending members, the defect is located within the range of ±1 / 3 of the section height of the neutral axis, or in the region of maximum bending moment; c) For tension / compression members, the defect is located in the region where the membrane stress is greater than 30% of the material's yield strength; The high-fidelity repair processing unit performs high-fidelity repair processing on critical defects, including: Curve fitting is performed on gap-type topological defects, and transition patch surfaces are generated to achieve precise stitching. The local geometric gaps are reconstructed using NURBS surfaces based on adjacent multi-cycle topological elements to form a continuous transition surface.

9. The system according to claim 8, characterized in that, The system also includes: The simplifiable defect judgment unit determines whether a defect in the geometric model is a simplifiable defect based on the following rules: a) Geometric determination: The feature size accounts for less than 3% of the model's base size and is located in a non-stressed area; b) Mechanical prediction: For bending members, the defect is located outside the area of ​​±2 / 3 of the section height of the neutral axis, and the bending moment is less than 10% of the maximum bending moment; c) Simulation sensitivity analysis: Parametric simulation verified that the change in mechanical properties after feature deletion was <2%; An adaptive simplification processing unit performs adaptive simplification processing on simplifiable defects, including: For the non-stressed end, the rounded corner is merged with the adjacent plane into a single plane. After merging, the number of meshes in the region is reduced by 60%, and the maximum distance deviation between the plane and the original feature is ≤0.02mm. For minor gaps in non-stressed areas, if the pre-simulation shows that their impact on mechanical properties is less than 1%, then no repair will be performed.

Images