CAD model through-hole feature recognition and processing method based on graph neural network

Abstract

The application discloses a CAD model through-hole feature recognition and processing method based on a graph neural network, comprising the following steps: constructing a surface adjacency graph according to the B-Rep representation of a CAD model; extracting differential geometric information and attribute information of each surface to construct a node initial feature vector; classifying each surface by using a graph neural network to determine a through-hole surface recognition result; performing triangular subdivision on the CAD model, deleting all grids located on the through-hole surface, applying a hole filling algorithm to repair the grid gap, determining a final grid model with the through-hole feature removed, or associating the surface label with a grid generation algorithm to guide grid generation and determine the finally generated anisotropic grid. The application comprehensively improves the recognition accuracy, processing efficiency and output quality while maintaining the lightweight of the model, and provides reliable technical support for the CAD / CAE integrated process.

Description

Technical Field

[0001] This application relates to the fields of deep learning and computer graphics, and in particular to a method for identifying and processing through-hole features in CAD models based on graph neural networks. Background Technology

[0002] Through-hole features are widely present in computer-aided design models, serving functions such as component connection, fluid transport, and structural heat dissipation. Their applications span aerospace, automotive manufacturing, and electronics industries. Before conducting computer-aided engineering (CAE) analysis, it is usually necessary to mesh the geometric model. However, the numerous through-hole features in the model can significantly impact the efficiency of mesh generation and simulation calculations. Therefore, how to achieve automatic identification and differentiated processing of through-hole features while maintaining the integrity of geometric and physical representation is a key technical challenge in balancing the accuracy and efficiency of CAE analysis.

[0003] In existing technologies, methods for recognizing rounded corner features are mainly divided into two categories: Traditional rule-based geometric methods rely on manually set geometric thresholds (such as diameter ranges) or simple topological rules (such as cylindrical surfaces) to identify through-holes. While effective for standard through-holes, these methods have fixed identification logic and struggle to accurately identify non-cylindrical through-holes, inclined holes, intersecting holes, or holes located on complex curved surfaces. Faced with the diverse hole structures in modern industrial models, rule-based methods exhibit insufficient coverage and adaptability, resulting in low levels of intelligence.

[0004] General Feature Recognition Methods Based on Deep Learning: In recent years, geometric deep learning methods based on neural networks have made significant progress in CAD feature recognition. However, existing models aim to provide a general framework for recognizing dozens of features simultaneously, but their learning objectives are scattered and fail to perform in-depth modeling and optimization of the semantic and geometric patterns implicit in the specific functional feature of "through holes." This leads to fuzzy criteria for judging through holes, and the recognition accuracy and reliability are insufficient to meet the simplification requirements of demanding CAE.

[0005] In the process of realizing this invention, the inventors discovered that the prior art has at least the following problems: Insufficient recognition accuracy and robustness: Traditional methods rely on manual rules, which are difficult to adapt to complex and varied through-hole structures; general deep learning models have limited ability to distinguish through-hole features; Weak generalization ability: Existing methods are sensitive to the distribution of training data, and the recognition accuracy drops significantly when faced with unseen through-hole types; Lack of intelligent association with CAE objectives: Most existing methods remain at the identification level and fail to intelligently associate the identification results with downstream tasks. Summary of the Invention

[0006] In view of this, embodiments of this application provide a method for identifying and processing through-hole features in CAD models based on graph neural networks, in order to solve the problems existing in the background art.

[0007] According to an embodiment of this application, a method for identifying and processing through-hole features in a CAD model based on a graph neural network includes: S1: Obtain the face adjacency graph constructed from the CAD model B-Rep representation; S2: Based on the surface adjacency graph, extract the differential geometric information and attribute information of each surface to determine the initial feature vector of the node; S3: Based on the initial feature vector of the node and the surface adjacency graph, use a graph neural network to classify each surface and determine the through hole surface identification result; S4: Based on the through-hole surface identification results, triangulate the CAD model, delete all meshes located on the through-hole surfaces, apply a hole filling algorithm to repair mesh gaps, and determine the final mesh model with through-hole features removed; or associate the surface labels with the mesh generation algorithm to guide mesh generation and determine the final anisotropic mesh generated.

[0008] The technical solutions provided by the embodiments of this application may include the following beneficial effects: As can be seen from the above embodiments, the graph neural network designed in this application is specifically designed for through hole recognition. It is trained on a self-built dataset containing various through hole feature types, which significantly improves the model's ability to recognize special through hole topologies and achieves high-precision recognition of through hole features in CAD models. This effectively solves the bottleneck of traditional methods when facing complex through holes.

[0009] The method proposed in this application demonstrates superior model efficiency, with the entire recognition network containing only 0.03M parameters, far fewer than existing state-of-the-art models, achieving a balance between high performance and low computational cost. This method maintains 100% recognition accuracy in multiple complex industrial test cases, exhibiting strong generalization ability and adapting to diverse through-hole structures in real-world industrial scenarios.

[0010] This application also constructs a large-scale and diverse benchmark dataset for through-hole feature recognition, which not only fills the gap in existing data resources, but also provides a reproducible and fair evaluation platform for subsequent research in this field.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0012] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0013] Figure 1 This is a text flowchart illustrating a method for identifying through-hole features in a CAD model based on a graph neural network, according to an exemplary embodiment.

[0014] Figure 2 This is a schematic diagram of a CAD model through-hole feature recognition method based on graph neural networks, according to an exemplary embodiment.

[0015] Figure 3 This is a schematic diagram illustrating the calculation of surface attribute features of a CAD model according to an exemplary embodiment.

[0016] Figure 4 This is a schematic diagram illustrating representative examples of a self-built via feature recognition dataset according to an exemplary embodiment.

[0017] Figure 5 This is a porous assembled disk model illustrated according to an exemplary embodiment, showing the through-hole features in the model.

[0018] Figure 6 This is an example of the through-hole feature recognition effect of a porous assembled disk model, as shown in an exemplary embodiment.

[0019] Figure 7 This is an example of the through-hole surface removal effect of a porous assembly disk model, as shown in an exemplary embodiment.

[0020] Figure 8 This is a comparison diagram of the through-hole feature removal effect and the amount of mesh before and after in a porous assembly disk model according to an exemplary embodiment.

[0021] Figure 9 This is an example of the through-hole feature recognition effect of a porous turbine blade model, as shown in an exemplary embodiment.

[0022] Figure 10 This is a comparison diagram of the anisotropic mesh generation effect and mesh quantity of a porous turbine blade model according to an exemplary embodiment.

[0023] Figure 11 This is a comparison diagram of the through-hole feature recognition performance of commercial software SolidWorks and the present invention, according to an exemplary embodiment. Detailed Implementation

[0024] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0025] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0026] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0027] In this invention, we primarily deal with CAD models represented using B-Rep. A CAD model is a digital three-dimensional solid representation created using computer-aided design software. It describes the shape and structure of an object through precise geometric data and topological information. B-Rep representation defines a three-dimensional solid as a spatial region enclosed by a set of closed surfaces, which are further defined by points and vertices, thus forming a geometric description with a strict hierarchical structure. The through-hole feature recognition method provided by this invention, through high-precision and high-efficiency through-hole recognition and processing, can significantly improve the pre-processing efficiency of CAD models to CAE simulation, and has broad application prospects in multiple industrial fields. 1. Aerospace Field: The surface of aero-engine turbine blades is densely covered with irregularly shaped airflow holes and structural weight reduction holes for cooling. Traditional identification methods based on geometric rules struggle to accurately distinguish these features on complex curved surfaces, often requiring extensive manual intervention for correction. The dedicated graph neural network for through-hole identification proposed in this invention can accurately identify various film cooling holes and effectively handle the topological relationships between hole arrays and complex curved surfaces, achieving fully automated identification and significantly shortening the preprocessing time for aero-engine thermo-mechanical coupling analysis.

[0028] 2. In the automotive manufacturing field: Engine blocks are riddled with through-hole features of various functions, including bolt holes and coolant holes. These features are not only numerous and complex in type, but also frequently intersect with other geometric features. Traditional methods require extensive manual selection and repair work, resulting in low efficiency. This invention can identify all through-hole features in batches at once. Combined with anisotropic mesh generation technology, the system can generate refined boundary layer meshes around key through-holes, reducing the overall mesh size and improving computational efficiency while ensuring connection strength and simulation accuracy.

[0029] 3. Electronic Equipment Field: This invention also demonstrates superior performance in addressing the high-density hole array recognition requirements of electronic equipment. Circuit boards and other devices typically contain thousands of micro-hole arrays, including heat dissipation holes, positioning holes, and various interface sockets. These holes are tiny in size and densely packed, posing a significant challenge to traditional recognition methods. The lightweight network model of this invention contains only 0.03M parameters, enabling extremely rapid processing of large-scale micro-hole arrays, shortening the simulation preprocessing cycle, and greatly improving product development efficiency.

[0030] This invention will use a multi-hole assembly disk model with through-hole features as an exemplary embodiment to explain the specific implementation. The method of this invention is applicable to CAD models with through-hole features, but is not limited to the multi-hole assembly disk model described below.

[0031] Figure 1 This is a flowchart illustrating a method for identifying and processing through-hole features in a CAD model based on a graph neural network, according to an exemplary embodiment. Figure 1 As shown, the method may include the following steps: S1: Obtain the face adjacency graph constructed from the CAD model B-Rep representation; Specifically, the B-Rep data structure of the CAD model is parsed, all surfaces are traversed, and each surface is abstracted as a node in an undirected graph. Then, based on the topological relationships between the surfaces in the B-Rep, undirected edges between nodes in the graph are constructed, ultimately forming a complete graph structure. .

[0032] Taking a multi-hole assembled disk model as an example, such as Figure 5 As shown, the model contains 20 parametric surfaces and 64 edges. In step S1, these 20 surfaces are transformed into 20 nodes in a graph structure, and the adjacency relationships between surfaces are transformed into connecting edges between nodes, thereby constructing the face adjacency graph of the model, laying the topological foundation for subsequent feature extraction and graph neural network analysis.

[0033] S2: Based on the surface adjacency graph, extract the differential geometric information and attribute information of each surface to determine the initial feature vector of the node; this step includes the following sub-steps: S21: Based on the surfaces in the face adjacency graph, uniformly sample on the UV parameter domain of the surfaces to determine the obtained sampling points; Specifically, for each parametric surface, its parameter domain is first normalized to... Within the range. Then generate in the UV space at equal intervals. A sampling point, for example, when At that time, 25 uniformly distributed sampling locations were obtained. For each sampling point... The corresponding three-dimensional coordinates are calculated using the parametric equations of the surface. And its differential geometric properties. This sampling strategy ensures that surfaces of different sizes and shapes can obtain feature representations of a uniform dimension, providing a standardized input format for subsequent neural network processing.

[0034] S22: Based on the sampling points, calculate the Gaussian curvature and average curvature of each sampling point to determine the constructed geometric feature vector; Specifically, at each sampling point, the first fundamental form coefficients of the surface are calculated. E, F, G Second fundamental form coefficient L, M, N Then, according to the differential geometry formulas: Calculate the Gaussian curvature K and the mean curvature H. Then, apply the curvature values ​​of all sampled points according to... Arrange, form The geometric eigenvectors are dimensional. Since curvature is an intrinsic geometric quantity of a surface, this characteristic indicates that the surface remains undeformed under rigid body transformations such as rotation and translation, and can stably distinguish fillet surfaces from other surfaces. Considering that the sign of curvature depends on the parameterization direction, to eliminate this ambiguity, the absolute value of curvature is taken to determine the constructed geometric eigenvectors. .

[0035] S23: Based on the geometric feature vector, extract the area ratio and ring number attribute information to determine the final constructed node initial feature vector.

[0036] Specifically, for each surface The area ratio is calculated by first traversing all the boundary edges of the surface and finding the longest straight edge. .like If there are two adjacent faces, then the other face is denoted as . ;if Adjacent to only one face, or curved surface If there is no straight edge in the middle, then let The formula for calculating the area ratio is: The area ratio attribute can effectively capture the geometric proportions between the through-hole surface and its adjacent surfaces. For example... Figure 3 As shown, Figure 3 In (a) , Figure 3 (b) The number of surface loops is defined as the number of closed boundary loops on a curved surface. In engineering scenarios, through-hole features are mainly used to construct branchless channels, requiring a complete sidewall structure. Therefore, the sidewall of a through-hole typically appears as a complete curved surface without holes, and its number of surface loops is 1.

[0037] S3: Based on the initial feature vectors of the nodes and the surface adjacency graph, a graph neural network is used to classify each surface and determine the through-hole surface identification result; this step includes the following sub-steps: S31: Design a network architecture that includes a preprocessing layer, a fusion layer, a graph convolutional layer, an embedding layer, and a classification layer, and determine the graph neural network model for through-hole recognition; Specifically, such as Figure 2 As shown, the preprocessing layer consists of fully connected layers, responsible for preprocessing geometric and attribute features, mapping them to a 32-dimensional latent space. The fusion layer integrates geometric and attribute features to achieve effective information integration. The graph convolutional layer uses a three-layer graph convolutional structure, with each layer outputting 64 dimensions and employing the Tanh activation function. The remaining modules in the network use the ReLU activation function. The embedding layer performs dimensionality reduction on the features after graph convolution and inputs them into the classification layer. The classification layer outputs the probability of binary classification to guide the final prediction. During training, batch normalization is applied to the graph convolutional layer and subsequent layers to stabilize the training dynamics and accelerate convergence.

[0038] S32: Based on the graph neural network model for through-hole recognition, train the model using a self-built labeled dataset, and determine the parameters of each layer of the network after training. Specifically, the training dataset consists of 9273 meticulously annotated CAD models, including not only standard cylindrical through-holes but also irregularly shaped through-holes such as elliptical cylinders. In addition to common cuboids, it also includes a rich variety of matrix morphologies such as spheres, cylinders, and swept volumes. Figure 4 Representative examples from the dataset are shown. The dataset is randomly divided into training, validation, and test sets in a 6:3:1 ratio. During training, the Adam optimizer, L2 regularization, and Dropout strategy are used to prevent overfitting and improve the model's generalization ability. The parameters of the model that performed best on the validation set are saved during training.

[0039] S33: Based on the trained neural network and the node feature vectors, reasoning is performed to obtain the probability that each surface belongs to a through-hole surface, and the through-hole surface identification result is determined.

[0040] Specifically, the node feature vector is input into the trained through-hole recognition model, and the classification layer of the model outputs the probability that each surface belongs to the through-hole surface and the non-through-hole surface; according to the probability, a label is assigned to each surface, and all surface classification labels are summarized to form a complete surface-level semantic segmentation result for the B-Rep representation of the CAD model, thereby determining the final through-hole surface recognition result.

[0041] Taking the multi-hole assembly disk model as an example, after model prediction, all 13 through-hole surfaces on the 20 curved surfaces of the model were accurately identified. Figure 6 The visualization effect of the model's through-hole surface recognition is shown: the green marked areas in the figure are all through-hole surfaces identified by the model, and the blue marked areas are non-through-hole surfaces, which intuitively verifies the accurate recognition capability of the present invention for through-hole features.

[0042] S4: Based on the through-hole surface identification results, triangulate the CAD model, delete all meshes located on the through-hole surfaces, apply a hole-filling algorithm to repair mesh gaps, and determine the final mesh model with through-hole features removed; or associate surface labels with the mesh generation algorithm to guide mesh generation and determine the final anisotropic mesh. This step includes the following sub-steps: S41: Based on the through-hole surface identification results, mark the geometric surfaces of the CAD model, and determine the mesh generation strategy according to the final goal: if the goal is to remove the through-hole features, then execute steps S42-S44; if the goal is to retain the through-hole features and generate anisotropic meshes, then execute step S45. Specifically, the results of the through-hole surface identification are converted into numerical labels and written into the B-Rep model, where 0 represents a non-through-hole surface and 1 represents a through-hole surface. The corresponding steps are then executed according to the specified through-hole feature processing strategy.

[0043] S42: Perform initial triangulation on the CAD model, delete all triangular mesh elements located on the through-hole surface, thereby forming hole boundaries on the mesh model and determining the mesh after the deletion operation; Specifically, the CAD model surface is first triangulated to generate an initial mesh. Then, based on the face labels, all triangular faces are traversed, and all triangular mesh cells located on the through-hole faces are deleted. After the deletion operation, the areas in the model originally occupied by the through-hole faces will form topologically open hole boundaries.

[0044] Taking a multi-hole assembled disk model as an example, delete the 13 curved surfaces identified as through holes. For example... Figure 7 As shown, the model has many geometric gaps, and the outlines of these gaps are the boundaries of the holes formed by the deletion operation.

[0045] S43: Based on the mesh after the deletion operation, determine all hole boundary rings generated by the deletion operation; Specifically, based on the damaged mesh after the deletion operation, a boundary edge detection algorithm is used to traverse all edges in the mesh. If an edge is referenced by only one triangular facet, it is determined to be a boundary edge. The interconnected boundary edges are then combined in sequence to form a closed hole boundary loop.

[0046] S44: Based on the hole boundary ring, apply the hole filling algorithm to fill each identified hole with a mesh, generate a new mesh patch that smoothly transitions with the surrounding mesh, and determine the final watertight mesh model that removes through-hole features; Specifically, for each hole boundary ring, a hole-filling algorithm based on triangulation is used to generate a new mesh within the ring. The quality of the newly generated mesh is then adjusted to ensure a smooth transition with the surrounding original mesh in terms of vertex density and normal variations. Finally, a global consistency check is performed on the overall model to ensure that the mesh is free from self-intersections and degradation, outputting a final watertight mesh model with through-hole features removed.

[0047] Taking the multi-hole assembly disk model as an example, Figure 8 The final watertight mesh model with through-hole features removed is shown, resulting in a smooth mesh transition. The mesh models of the porous assembly disk model before and after through-hole feature removal are remeshed with the same parameters, and the mesh size is compared. Figure 8 It can be seen that after removing the through-hole feature, the mesh size of the model decreased from 50481 to 16104.

[0048] S45: Based on the CAD model with face labels, associate the face labels with the mesh generation algorithm, automatically refine the mesh at the edge of the through hole feature, generate high aspect ratio cells in the through hole feature normal, maintain a relatively coarse scale in the tangential direction, and determine the final anisotropic mesh model that accurately captures the through hole feature and optimizes the number of cells.

[0049] Specifically, based on the CAD model with face labels, localized density constraints are set at the edges of the through-holes. Boundary layer generation technology is used to generate high aspect ratio elements in the normal direction of the through-hole sidewalls. In the axial direction and circumferential tangential direction of the through-hole, coarser element sizes are used to generate an extruded mesh aligned with the feature. Finally, mesh quality optimization ensures the quality of the element shape, outputting an anisotropic mesh that accurately captures the geometry of the through-hole while significantly reducing the total number of elements.

[0050] Taking a porous turbine blade model as an example, Figure 9 This refers to the effectiveness of the model in recognizing through-hole features. Figure 10 The left side shows the anisotropic mesh generated according to step S45. (And...) Figure 10 Compared to the isotropic mesh on the right, the number of mesh elements is reduced from 367,250 to 151,602.

[0051] As demonstrated by the above embodiments, this application effectively solves the technical challenge of automated identification of through-hole features in CAD models. By innovatively constructing a dedicated graph neural network architecture for through-hole features and combining a feature fusion mechanism of differential geometric features and attribute features, high-precision identification of various through-hole features is achieved. The identification results can directly drive mesh generation strategies, guiding through-hole feature removal and anisotropic mesh generation tasks, significantly improving the efficiency and quality of CAE preprocessing.

[0052] In this embodiment, we compared the through-hole feature recognition performance of our invention with that of mainstream commercial software SolidWorks. For example... Figure 11 The complex aircraft structural model shown has the following results: On the left, SolidWorks displays the recognition results, with yellow highlighting indicating unrecognized through-hole surfaces and green highlighting accurately recognized ones. SolidWorks' recognition accuracy is only 14.9%. On the right, the present invention displays the recognition results, achieving 100% accuracy. This demonstrates that the present invention exhibits stronger robustness and adaptability in handling complex through-hole scenarios, effectively addressing the low recognition accuracy of existing methods in such situations and providing reliable technical support for integrated CAD / CAE workflows.

[0053] In summary, this invention achieves a comprehensive improvement in recognition accuracy, processing efficiency, and output quality while maintaining a lightweight model, providing reliable technical support for integrated CAD / CAE workflows. The verification in Example 1 fully demonstrates the significant advancements and practical value of the method compared to existing technologies.

[0054] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0055] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for identifying and processing through-hole features in CAD models based on graph neural networks, characterized in that, include: S1: Obtain the face adjacency graph constructed from the CAD model B-Rep representation; S2: Based on the surface adjacency graph, extract the differential geometric information and attribute information of each surface to determine the initial feature vector of the node; S3: Based on the initial feature vector of the node and the surface adjacency graph, use a graph neural network to classify each surface and determine the through hole surface identification result; S4: Based on the through-hole surface identification results, triangulate the CAD model, delete all meshes located on the through-hole surfaces, apply a hole filling algorithm to repair mesh gaps, and determine the final mesh model with through-hole features removed; or associate the surface labels with the mesh generation algorithm to guide mesh generation and determine the final anisotropic mesh generated.

2. The method according to claim 1, characterized in that, Based on the surface adjacency graph, the differential geometric information and attribute information of each surface are extracted to determine the initial feature vector of the nodes, including: S21: Based on the surfaces in the face adjacency graph, uniformly sample on the UV parameter domain of the surfaces to determine the obtained sampling points; S22: Based on the sampling points, calculate the Gaussian curvature and average curvature of each sampling point to determine the constructed geometric feature vector; S23: Based on the geometric feature vector, extract the area ratio and ring number attribute information to determine the final constructed node initial feature vector.

3. The method according to claim 1, characterized in that, Based on the initial feature vectors of the nodes and the surface adjacency graph, a graph neural network is used to classify each surface and determine the via surface identification result, including: S31: Design a network architecture that includes a preprocessing layer, a fusion layer, a graph convolutional layer, an embedding layer, and a classification layer, and determine the graph neural network model for through-hole recognition; S32: Based on the graph neural network model for through-hole recognition, train the model using a self-built labeled dataset, and determine the parameters of each layer of the network after training. S33: Based on the trained neural network and the node feature vectors, reasoning is performed to obtain the probability that each surface belongs to a through-hole surface, and the through-hole surface identification result is determined.

4. The method according to claim 1, characterized in that, Based on the through-hole surface identification results, the CAD model is triangulated, all meshes located on the through-hole surface are deleted, the hole filling algorithm is applied to repair the mesh gaps, and the final mesh model with through-hole features removed is determined. Alternatively, surface labels can be associated with the mesh generation algorithm to guide mesh generation and determine the final anisotropic mesh, including: S41: Based on the through-hole surface identification results, mark the geometric surfaces of the CAD model, and determine the mesh generation strategy according to the final goal: if the goal is to remove the through-hole features, then execute steps S42-S44; if the goal is to retain the through-hole features and generate anisotropic meshes, then execute step S45. S42: Perform initial triangulation on the CAD model, delete all triangular mesh elements located on the through-hole surface, thereby forming hole boundaries on the mesh model and determining the mesh after the deletion operation; S43: Based on the mesh after the deletion operation, determine all hole boundary rings generated by the deletion operation; S44: Based on the hole boundary ring, apply the hole filling algorithm to fill each identified hole with a mesh, generate a new mesh patch that smoothly transitions with the surrounding mesh, and determine the final watertight mesh model that removes through-hole features; S45: Based on the CAD model with face labels, associate the face labels with the mesh generation algorithm, automatically refine the mesh at the edge of the through hole feature, generate high aspect ratio cells in the through hole feature normal, maintain a relatively coarse scale in the tangential direction, and determine the final anisotropic mesh model that accurately captures the through hole feature and optimizes the number of cells.

Images