Model quality assessment optimization method, device and server for industrial simulation applications

Abstract

The present invention provides a model quality assessment and optimization method, device, and server for industrial simulation applications, relating to the technical field of universal development of simulation applications. The method comprises: obtaining a geometric model to be processed and determining a target input model by performing format detection, format conversion, and asynchronous processing on the geometric model; performing meshing and finite element analysis on the target input model to determine finite element calculation results, and performing local error estimation and analysis on the finite element calculation results to locally refine the model mesh of the target input model, identify key areas of the model, perform local encryption on the key areas, and determine the model output results; and optimizing the model output results through a hierarchical model quality assessment and optimization system to determine the optimized target model output results. The present invention can significantly improve model quality.

Description

Technical Field

[0001] The present invention relates to the technical field of universal development of simulation applications, and in particular to a model quality assessment optimization method, device and server for industrial simulation applications. Background Art

[0002] At present, relevant technologies have proposed that due to the lack of model quality inspection functions and processes in existing industrial simulation applications, the finite element model quality inspection and meshing quality optimization during model meshing in traditional simulation applications mainly rely on manual settings and adjustments. These processes of manually processing and optimizing the quality of complex models often make it difficult to balance mesh quality and computational efficiency. On the one hand, although high-precision meshes can provide more accurate analysis results, they will lead to a surge in rendering computing power consumption and a drop in frame rate, which will seriously affect the user experience. On the other hand, over-simplified meshes will lose key details, such as stress concentration areas and geometric feature edges, thereby affecting the accuracy and reliability of model visualization. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a model quality assessment optimization method, device and server for industrial simulation applications, which can significantly improve model quality.

[0004] In the first aspect, an embodiment of the present invention provides a model quality assessment and optimization method for industrial simulation applications, the method comprising: obtaining a geometric model to be processed, and determining a target input model by performing format detection processing, format conversion processing, and asynchronous processing on the geometric model; performing meshing processing and finite element analysis processing on the target input model to determine the finite element calculation results, and performing local error estimation and analysis processing on the finite element calculation results to locally refine the model mesh of the target input model, determine the key areas of the model, and perform local encryption processing on the key areas to determine the model output results; optimizing the model output results through a model quality assessment and optimization system with a hierarchical architecture to determine the optimized target model output results, wherein the model quality assessment and optimization system comprises: a gradient-driven data organization module, a multi-dimensional quality assessment module, and an intelligent optimization processing module.

[0005] In one embodiment, the step of determining the target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model includes: performing format detection processing on the geometric model, determining the file extension of the geometric model file, and when it is detected that the geometric model file corresponding to the file extension is a non-target format file, uploading the geometric model file to the cloud for format conversion processing to convert the non-target format file into a target format file; if the file format conversion is successful, deleting the temporary files generated during the format detection processing and format conversion processing, and initializing the database to perform asynchronous processing on the geometric model according to the solution type corresponding to the geometric model in the initialized database to determine the target input model.

[0006] In one embodiment, the finite element calculation results are subjected to local error estimation analysis processing to locally refine the model mesh of the target input model, and the step of determining the key area of ​​the model includes: determining the unit stress and the smooth stress based on the node reconstruction according to the finite element calculation results, and determining the error estimation value based on the unit stress and the smooth stress through an error estimator, so as to determine the key area of ​​the model according to the error estimation value.

[0007] In one embodiment, the step of determining the key area of ​​the model based on the error estimate value includes: when the error estimate corresponding to any area of ​​the target input model is greater than a preset error threshold, it is determined that there is a stress mutation in the area, and the area is marked as the key area of ​​the model, so as to perform local encryption processing on the key area, and re-perform finite element calculation to determine the output result.

[0008] In one embodiment, the model output results are optimized by a model quality assessment and optimization system with a hierarchical architecture, and the steps of determining the optimized target model output results include: utilizing the data gradient dynamic partitioning algorithm and the view cone constraint resolution decision algorithm in the gradient-driven data organization module to dynamically adjust the resolution of the model rendering, perform intelligent rendering processing on the model output results, and determine the optimized model output results; combining the illumination energy calculation algorithm, normal smoothness quantization algorithm, principal curvature calculation algorithm, and mesh defect detection algorithm in the multidimensional quality assessment module to perform multidimensional quality assessment on the surface point radiance, vertex normal smoothness, principal curvature, defect type, and defect location in the optimized model output results; and using the Laplace smoothing algorithm, normal mean filtering algorithm, conformal smoothing algorithm, and mesh subdivision algorithm in the intelligent optimization processing module to smooth, filter, and redirect the normals of the model output results after quality assessment to determine the target model output results.

[0009] In one embodiment, before determining the optimized model output result, the method includes: utilizing a patch visibility determination algorithm in a gradient-driven data organization module to remove invisible patches facing away from the viewpoint during intelligent rendering.

[0010] In one embodiment, before the step of smoothing, filtering and normal redirecting the model output results after quality assessment through the Laplace smoothing algorithm, normal mean filtering algorithm, conformal smoothing algorithm and mesh subdivision algorithm in the intelligent optimization processing module and determining the target model output results, it includes: obtaining a set of adjacent facets of each vertex, and using the feature-aware normal calculation algorithm in the multi-dimensional quality assessment module to analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet; determining the weighted average of the vertex normals and the weighted average of the curvature based on the facet area and the facet normal, and determining the target normal vector based on the weighted average of the vertex normals and the weighted average of the curvature to separate the sharp edges of the model from the normal.

[0011] In the second aspect, an embodiment of the present invention also provides a model quality assessment and optimization device for industrial simulation applications, the device including: a model preprocessing unit, which obtains a geometric model to be processed, and determines a target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model; a local encryption unit, which performs meshing processing and finite element analysis processing on the target input model, determines the finite element calculation results, and performs local error estimation and analysis processing on the finite element calculation results, so as to locally refine the model mesh of the target input model, determine the key areas of the model, and perform local encryption processing on the key areas to determine the model output results; an assessment and optimization unit, which optimizes the model output results through a hierarchical model quality assessment and optimization system to determine the optimized target model output results, wherein the model quality assessment and optimization system includes: a gradient-driven data organization module, a multi-dimensional quality assessment module and an intelligent optimization processing module.

[0012] In a third aspect, an embodiment of the present invention further provides a server, comprising a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement any one of the methods provided in the first aspect.

[0013] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the first aspect.

[0014] The embodiments of the present invention bring the following beneficial effects:

[0015] An embodiment of the present invention provides a model quality assessment and optimization method, device and server for industrial simulation applications. After obtaining a geometric model to be processed, the method determines a target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model. Then, meshing processing and finite element analysis processing are performed on the target input model to determine the finite element calculation results, and local error estimation and analysis processing are performed on the finite element calculation results to locally refine the model mesh of the target input model, determine the key areas of the model, perform local encryption processing on the key areas, determine the model output results, and finally optimize the model output results through a model quality assessment and optimization system with a hierarchical architecture to determine the optimized target model output results. The embodiment of the present invention can achieve a balance between rendering performance and quality by intelligently identifying visual key areas and implementing local enhancement, while ensuring the accuracy of key details and reducing the computational load of non-important areas, thereby improving model quality.

[0016] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A flow chart of a method for optimizing model quality assessment for industrial simulation applications provided by an embodiment of the present invention;

[0020] Figure 2 A schematic flow chart of a model preprocessing method provided in an embodiment of the present invention;

[0021] Figure 3 A schematic diagram of a process for determining a model output result provided by an embodiment of the present invention;

[0022] Figure 4 A schematic diagram of a specific process of a model quality assessment and optimization method for industrial simulation applications provided by an embodiment of the present invention;

[0023] Figure 5 A schematic diagram of the structure of a model quality assessment and optimization device for industrial simulation applications provided by an embodiment of the present invention;

[0024] Figure 6 A schematic diagram of the structure of a server provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] Currently, scenario-based industrial simulation applications are finite element simulation tools designed for customized products for small and medium-sized enterprises. To reduce user operational complexity, the finite element pre-processing, solution, and post-processing are simplified and automated as much as possible. Throughout the entire industrial simulation application process, model quality processing, analysis, and optimization are involved, including finite element model quality inspection and mesh quality optimization during meshing. Since the quality of most user-provided models varies, finite element analysis requires greater attention to their quality. This quality is determined by the quality of the finite element mesh after meshing. This later forms a universal component for model quality mesh independence inspection in industrial simulation applications. This component, combined with advances in post-processing algorithms in industrial simulation applications, can simultaneously optimize the model's post-processing rendering performance. High-precision meshes for complex models can lead to increased rendering computational overhead and decreased frame rates, while overly simplified meshes can lose critical details (such as stress concentration areas and geometric feature edges), compromising visualization. Traditional rendering methods employ a globally unified level-of-detail simplification strategy, failing to dynamically allocate computing resources based on the importance of local areas within the model.

[0027] Specifically, traditional rendering methods use a globally unified level-of-detail simplification strategy, which cannot dynamically allocate computing resources based on the local importance of the model. This makes it difficult to meet the differentiated requirements for detail accuracy and rendering performance in different areas of complex models. Consequently, models in industrial simulation applications suffer from poor rendering effects and a lack of light perception, which can be manifested in the following aspects:

[0028] 1. Low processing efficiency due to fixed resolution: Traditional methods use a global unified resolution for rendering. For complex models containing more than 1 million facets, a single lighting optimization takes more than 2 hours. Layered detail technology relies on predefined levels (usually only 3-5 levels) and cannot dynamically adjust the resolution based on real-time curvature changes and stress distribution. Taking aircraft engine blade simulation as an example, the leading and trailing edges of the blade are processed at the same resolution, resulting in a waste of more than 40% of computing resources.

[0029] 2. Lack of visual rendering technology:

[0030] (1) Lack of backlight culling mechanism: Existing tools do not implement face visibility judgment based on the normal direction, resulting in faces facing away from the viewpoint (accounting for about 50%) still participating in rendering calculations, wasting GPU resources.

[0031] (2) Rough normal calculation: The surface normal is generally assigned directly to the vertex, or the adjacent surface normals are simply averaged without considering the preservation of geometric features (such as the separation of sharp edge normals). This results in "pseudo-angular" models after smoothing, and the light perception does not match the actual geometric features.

[0032] 3. Fragmentation of the evaluation and optimization processes: Existing tools often focus on a single quality dimension (such as surface roughness or lighting effects alone) and fail to establish a comprehensive multi-index evaluation system. For example, lighting parameter adjustments rely on visual feedback and lack quantitative evaluation based on physically based rendering (PBR) theory. Surface smoothing is not combined with curvature distribution and simulation load characteristics, which can easily lead to loss of details or computational redundancy.

[0033] 4. Technical barriers and cost issues: Advanced optimization functions (such as adaptive subdivision and physical lighting rendering) rely on expensive commercial software or complex script development. Small and medium-sized enterprises need to invest a lot of manpower to master the underlying algorithms (such as Laplace smoothing and loop subdivision), making it difficult to popularize the technology.

[0034] Based on this, the model quality assessment optimization method, device and server for industrial simulation applications provided by the present invention can achieve a balance between rendering performance and quality by intelligently identifying key visual areas and implementing local enhancements, while ensuring the accuracy of key details and reducing the computational load of non-important areas, thereby improving model quality.

[0035] See also Figure 1 The flowchart of a model quality assessment optimization method for industrial simulation applications is shown, and the method mainly includes the following steps S102 to S106:

[0036] Step S102, obtain the geometric model to be processed, and determine the target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model. In one embodiment, the geometric model can be subjected to format detection processing to determine the file extension of the geometric model file, and when it is detected that the geometric model file corresponding to the file extension is a non-target format file, the non-target format file is converted into a target format file by uploading the geometric model file to the cloud for format conversion processing. If the file format conversion is successful, the temporary files generated during the format detection processing and format conversion processing are deleted, and the database is initialized to perform asynchronous processing on the geometric model according to the solution type corresponding to the geometric model in the initialized database to determine the target input model.

[0037] Step S104, meshing and finite element analysis are performed on the target input model to determine the finite element calculation results, and local error estimation and analysis are performed on the finite element calculation results to locally refine the model mesh of the target input model, determine the key areas of the model, and perform local encryption on the key areas to determine the model output results. In one embodiment, in order to realize finite element analysis of the three-dimensional model, meshing of the model is an indispensable part. Since the size of the mesh is difficult to define, it is feasible to adopt a strategy of extremely small mesh size to realize meshing. However, endlessly reducing the mesh size will lead to a surge in the number of meshes, which will consume a lot of computing power and increase the subsequent finite element solution time and data transmission time. In addition, it is impossible to obtain reasonable meshing data, which will lead to the same problem after subsequent large-scale calculations and iterations, making it difficult to use effective data for subsequent automatic setting iterations and optimization of the mesh size.

[0038] Therefore, a feedback-based method can be used to verify the mesh independence of industrial simulation applications during runtime, which can be reasonable and minimize computing resources. Due to the strong ease of use of industrial simulation applications, the mesh independence optimization method does not require user interaction. The benchmark information collected comes from the builders of industrial simulation applications based on low-code simulation platforms. When setting parameters for specific industrial simulation applications, the builders need to determine the benchmark size for mesh division. After automatic mesh division and result calculation based on this benchmark size, the global mesh is not encrypted at the same time as traditional methods. Global encryption will excessively increase the number of meshes and does not directly improve the accuracy of large error areas. Here, a local error estimation analysis method is used to locally refine the model mesh, especially in key areas of stress concentration. Local mesh encryption is implemented to obtain a gradually stable and reliable stress distribution, and computing resources are concentrated in the areas that have the greatest impact on the results. These areas are often located at geometric edges, hole edges, sharp corners, or load contact points where the maximum stress values ​​appear. After analysis, the system uses a specific algorithm to identify the error distribution in these areas and perform local encryption. After multiple iterations, the results of each analysis are compared to determine whether the stress is stable and thus whether convergence has been achieved.

[0039] Step S106, optimize the model output results through the hierarchical model quality assessment and optimization system to determine the optimized target model output results, wherein the model quality assessment and optimization system includes: a gradient-driven data organization module, a multi-dimensional quality assessment module and an intelligent optimization processing module. In one embodiment, since the local grid encryption effect is represented by the grid in the stress concentration area in the model becoming gradually denser, while the grid away from the key area becomes relatively sparse and does not change significantly, therefore, if the maximum stress tends to be stable after local encryption and the corresponding regional error index decreases, it means that the model has achieved a sufficient level of accuracy under the current grid. With the development of industrial simulation technology in products With deep integration in R&D, the quality of three-dimensional models has become a key factor in determining the accuracy and efficiency of numerical simulation. In complex scenarios such as aerospace structural strength analysis, automobile collision dynamics simulation, and fluid dynamics (CFD) simulation, parameters such as the model's lighting performance, surface smoothness, and geometric accuracy not only affect the visual interpretation effect, but are also closely related to the convergence of the physical field solution and calculation errors. For example, in finite element analysis, tiny serrations on the model surface may lead to misjudgment of stress concentration, and light distortion will obscure key geometric features, affecting engineers' recognition of model details. However, existing industrial simulation model processing solutions have many limitations and are difficult to meet the needs of efficient, accurate, and intelligent model quality optimization.

[0040] Therefore, a model quality assessment and optimization system with a hierarchical architecture can be constructed, which can realize intelligent processing of the entire process from model import to simulation readiness through gradient-driven data organization modules, multi-dimensional quality assessment modules, and intelligent optimization processing modules.

[0041] Regarding the above step S102, in order to simplify and automate the pre-processing as much as possible, three core pain points need to be addressed: 1. There are significant shortcomings in multi-format compatibility. Niche format files such as .x_t cannot be automatically parsed and need to rely on manual conversion, which greatly reduces design efficiency; 2. The system architecture design is unreasonable, and the front-end interaction and back-end data processing are highly coupled, which can easily cause logical errors in complex face selection operations and parameter binding processes; 3. The job submission process lacks unified standards and specifications, the asynchronous task scheduling mechanism is imperfect, and the error handling capability is weak, which seriously affects the stability and reliability of task processing.

[0042] In response to the above problems, a complete integrated solution is needed, covering the entire process of geometric model uploading, format intelligent processing, visual interactive design and standardized job submission. By building a standardized input interface, deploying an automated processing engine, and developing a visual interactive interface, the system compatibility, stability and usability can be effectively improved, providing users with an efficient, accurate and convenient CAD collaborative design experience. The solution includes the following modules: File processing module: supports multi-format parsing, .x_t file cloud conversion and temporary file management; Database module: stores geometric instance metadata, job status and rendering path, supports asynchronous task triggering and status update; Rendering conversion module: calls the Ares service to generate visual rendering files, and manages unique file names based on the snowflake algorithm; Front-end interaction module: provides model visualization interface, face selection operation and real-time parameter binding function; Job scheduling module: generates standardized job input, creates a workspace and submits it to the solver, supports idempotence verification and error handling, see Figure 2 The flowchart of a model preprocessing method is shown in FIG. The specific preprocessing scheme is as follows (1) to (5):

[0043] (1) File format detection and conversion: Receive multi-part form data, parse and detect uploaded file extensions (.stp, .x_t, .igs, .step, .par), convert .x_t files to .stp format through the Diwei cloud interface, poll to obtain the conversion results and decompress them, and delete temporary files; realize multi-format automated processing: cover mainstream CAD formats, and convert .x_t files in the cloud to reduce manual intervention.

[0044] (2) Database initialization and asynchronous preprocessing: Create a geometry instance and bind metadata such as user ID and project ID, and insert it into the database; trigger geometry preprocessing (broken surface repair, chamfer simplification) according to the solution type, generate a preprocessing script and submit an asynchronous job; improve system throughput and separate I / O-intensive tasks from core logic.

[0045] (3) Rendering file generation and status management: query the rendering status through the geometry ID, call the Ares service for the "to be rendered" file to generate a .zip format rendering file, and update the database storage path.

[0046] (4) Front-end interaction and parameter binding: Load the model rendering path and user input configuration. The user selects the load surface / fixed surface through the visual interface and enters the load / constraint parameters, and binds the geometric ID in real time.

[0047] (5) Job submission and solution preparation: parse user parameters to generate a JSON file, create a unique workspace path, initialize job metadata and submit it to the solver, and update the job status to queued.

[0048] The above solution realizes the standardization and automation of industrial simulation pre-processing processes through automatic conversion of multi-format files, asynchronous pre-processing scheduling, visual parameter binding and standardized job submission, solving problems such as insufficient compatibility, inefficient interaction, and chaotic task scheduling in existing technologies, and providing a reliable data foundation for meshing and stress analysis.

[0049] See also Figure 3 A flow chart of determining the output result of a model is shown. An embodiment of the present invention also provides an implementation method for determining the key area of ​​the model, determining the unit stress and the smooth stress based on node reconstruction according to the finite element calculation results, and determining the error estimation value based on the unit stress and the smooth stress through the error estimator, so as to determine the key area of ​​the model according to the error estimation value. When the error estimation value corresponding to any area of ​​the target input model is greater than the preset error threshold, it is determined that there is a stress mutation in the area, and the area is marked as the key area of ​​the model, so as to perform local encryption processing on the key area, and re-perform the finite element calculation to determine the output result. In one implementation method, the error analysis and allowable value judgment process of the finite element calculation results automatically performed in industrial simulation applications is as follows: the error estimator can be used for local encryption judgment, and the unit stress and smooth stress based on node reconstruction Compare and evaluate the relative error of the unit. The calculation formula is:

[0050]

[0051] Among them, Ω represents the area where the unit is located. The larger the error estimate η is, the larger the error in the area is, and the more local mesh encryption is needed. By counting the error indicators of all units in the entire model and sorting them accordingly, the unit areas with the largest errors are preferentially selected for local mesh encryption. Here, the allowable value will be judged. If the value is higher than the threshold set by the user, (The default setting is 5%, which can be adjusted according to the actual scenario of specific industrial simulation applications). If the stress in this area is significantly different from the actual stress in the real world, the area will be marked and the local mesh will be encrypted. If the stress is lower than the threshold set by the user, the area will be marked as a large stress mutation and the error is significant. , then the error estimation environment of the calculation results here is used to output the positive model quality results.

[0052] In one embodiment, the overall function is based on a low-code development platform to form a universal component suitable for industrial simulation applications, including functions such as allowable error ratio, grid encryption ratio, relative error convergence, etc. Specifically, in order to design the local grid encryption function into a component that can be used for universal model quality grid independence inspection in industrial simulation applications, specific parameter change properties are designed. The "allowable error ratio" is the error reference ratio for judging the results of two calculations, and it has been determined whether the two calculation results converge. The "grid encryption ratio" is a parameter that controls the change in grid density in each encryption operation, which is defined as the increase multiple of the number of units after each refinement relative to the original number. The grid encryption ratio is adjustable, and the default value is set to 1.5, which means that the number of units in the target area after encryption is 1.5 of the original, and the side length of the unit is also proportionally reduced, which is set here to be the original length. , where r is the encryption ratio, which ensures that each encryption can effectively improve the accuracy of the result without causing a sharp increase in computing resources.

[0053] The entire convergence process is an iterative optimization process. After each encryption, the key extreme value data in the current and previous analyses are recalculated and compared. Here, equivalent stress is used according to the actual operational requirements of industrial simulation applications. It can also be adjusted to node displacement data based on the actual focus. If the allowable error ratio is lower than a certain threshold (here, 5%, set as an adjustable parameter, which can be adjusted according to engineering needs), the result is considered to have converged. The relative error convergence judgment formula is as follows:

[0054]

[0055] in, Represents the relative stress change rate, used to determine whether it has converged, and the unit is dimensionless percentage; Represents the maximum equivalent stress value obtained after the kth encryption, in MPa; The maximum equivalent stress value obtained after the k-1th encryption, in MPa.

[0056] See also Figure 4 The following is a schematic diagram of a specific process of a model quality assessment and optimization method for industrial simulation applications. This embodiment of the present invention also provides an implementation method for implementing intelligent processing of the entire process from model import to simulation readiness through a gradient-driven data organization module, a multi-dimensional quality assessment module, and an intelligent optimization processing module. For details, see (A) to (C) below:

[0057] (A) Utilizing the data gradient dynamic partitioning algorithm and the view cone constraint resolution decision algorithm in the gradient-driven data organization module, the resolution of the model rendering is dynamically adjusted, the model output result is intelligently rendered, and the optimized model output result is determined. In one embodiment, the patch visibility judgment algorithm in the gradient-driven data organization module can also be utilized to eliminate invisible patches facing away from the viewpoint during intelligent rendering. For details, see (1) to (3) below:

[0058] (1) Data gradient-aware dynamic partitioning algorithm: An octree data structure is constructed, and the subdivision strategy is dynamically determined by calculating the data gradient (geometric curvature, flow field velocity gradient, etc.). For the fluid velocity field data, the vortex center area is subdivided to the 8th layer, and the smooth flow area is only subdivided to the 3rd layer, reducing the overall computational complexity by 60%. A unique dynamic partitioning algorithm is used here to break through the limitations of traditional fixed resolution processing methods. The specific details of the algorithm are as follows:

[0059] 1. Initialize the octree data structure and treat the entire model area as the root node.

[0060] 2. For each node, calculate its internal geometric curvature and flow field velocity gradient.

[0061] The geometric curvature calculation formula is: Among them, v is the tangent vector and a is the normal vector.

[0062] The calculation formula for the flow field velocity gradient is:

[0063] Among them, Ux, Uy, and Uz are the components of the velocity in the x, y, and z directions respectively.

[0064] 3. Determine the subdivision strategy based on the gradient value: If the geometric curvature or flow field velocity gradient is greater than a set threshold (e.g., 0.8), the node is subdivided into 8 child nodes, and steps (2)-(3) are recursively repeated until the maximum subdivision depth (e.g., the 8th layer) is reached. If the gradient value is less than the set threshold, the subdivision is stopped and the node is considered a leaf node.

[0065] (2) View cone constraint resolution decision model: Dynamically adjust the resolution in real time according to the camera position and viewing angle, achieving "near-view high definition - distant-view simplified" intelligent rendering, improving rendering flexibility and efficiency. This model makes intelligent resolution decisions based on the positional relationship between the camera and the model, innovatively solving the problem of fixed resolution in traditional rendering. The specific details of the algorithm are as follows:

[0066] 1. Calculate the distance d from the camera to the model area:

[0067]

[0068] in, etc. represent the coordinate components of the camera in the model coordinate system, Etc. represents the coordinate components of the center of the model bounding box.

[0069] 2. Calculate the resolution adjustment coefficient α based on the distance d and the viewing angle θ:

[0070]

[0071] Where k is an adjustment coefficient (such as 0.1) that controls the steepness of the resolution change.

[0072] 3. For each node in the model, determine its level of detail (LOD) based on its relative position to the camera: if the node is close to the camera (e.g., the distance is less than ), use high resolution (such as subdividing to the 8th layer). If the node is far away from the camera (such as the distance is greater than ), a lower resolution is used (e.g., subdivided to level 3). Nodes at intermediate distances determine the resolution based on linear interpolation of α.

[0073] 4. Return the resolution of each part of the model.

[0074] (3) Optimize backlight culling based on the face visibility judgment algorithm: culling invisible faces based on the angle between the normal and the viewpoint direction, reducing invalid face rendering by about 50% and improving GPU utilization. This backlight culling mechanism is an optimization and innovation based on existing technology, effectively improving rendering efficiency. The specific details of the algorithm are as follows:

[0075] 1. Initialize the visible face set to an empty set.

[0076] 2. For each patch Fi: calculate the normal of the patch , calculate the angle between the normal and the viewpoint direction : , where the viewpoint direction is .

[0077] Among them, if If the angle is less than a set threshold (such as 90°), the face is considered visible and added to the visible face set.

[0078] 3. Return the visible face set.

[0079] (B) The illumination energy calculation algorithm, normal smoothness quantification algorithm, principal curvature calculation algorithm and mesh defect detection algorithm in the multi-dimensional quality assessment module are combined to perform multi-dimensional quality assessment on the surface point radiance, vertex normal smoothness, principal curvature, defect type and defect location in the optimized model output results. In one embodiment, a set of adjacent facets of each vertex can be obtained, and the feature-aware normal calculation algorithm in the multi-dimensional quality assessment module is used to analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet. Then, based on the facet area and facet normal, the weighted average of the vertex normal and the weighted average of the curvature are determined, and based on the weighted average of the vertex normal and the weighted average of the curvature, the target normal vector is determined to separate the sharp edge of the model from the normal. For details, see (1) to (5) below:

[0080] (1) Lighting energy calculation algorithm: Based on PBR lighting energy calculation, the Monte Carlo path tracing algorithm is used to estimate the surface point radiance, and backlight culling technology is combined to improve efficiency. High light overflow detection is based on the physical rendering equation. After optimization, the diffuse reflection uniformity, specular reflection signal-to-noise ratio, ambient light contrast and other indicators reach a better level. At the same time, the normal smoothness index is quantified to ensure the continuity of light perception. A multi-dimensional quality evaluation system is established to comprehensively evaluate the model quality. The algorithm is as follows:

[0081] 1. Initialize the radiance array is zero.

[0082] 2. For each surface point

[0083] generate Random sampling directions (like , for each sampling direction , calculate the incident light intensity , based on factors such as light source position, intensity, and attenuation: , where is the light source intensity, is the distance from the light source to the surface point.

[0084] Calculating Fresnel terms Normal distribution function Geometric masking function :

[0085]

[0086] Simplified formula, actually calculated based on material parameters:

[0087]

[0088] Beckmann distribution function, is the half-range vector, For roughness:

[0089]

[0090] Calculating specular reflectivity

[0091]

[0092] Calculate the diffuse component: ,in, is the diffuse reflectance, is the direction of incident light.

[0093] Calculating the specular component .

[0094] in, is the specular reflectivity, is the reflection direction, is the viewpoint direction.

[0095] Cumulative lighting contribution: .

[0096] 3. Return the surface point irradiance

[0097] (2) Normal smoothness quantification algorithm

[0098] 1. For each vertex, get the normal set of all faces adjacent to the vertex.

[0099] 2. Calculate the average deviation angle between the vertex normal and the adjacent face normal .

[0100]

[0101] in, is the vertex normal, is the number of adjacent faces.

[0102] 3. Returns a set of vertex normal smoothness indices.

[0103] (3) Principal curvature calculation algorithm: The curvature and mesh defect detection and simulation adaptation evaluation submodule uses the local covariance matrix to calculate the principal curvature, detect mesh defects such as overlapping surfaces, voids, and non-manifold edges, and perform adaptation evaluation for different simulation types. It innovatively combines defect detection with simulation adaptation to improve the accuracy of the model in simulation. The specific details of the algorithm are as follows:

[0104] 1. For each surface point : Get the point set within its neighborhood Constructing the local covariance matrix :

[0105]

[0106] in, is the average coordinate of the neighborhood points.

[0107] Compute the eigenvalues ​​of the covariance matrix and the corresponding eigenvector .

[0108] The formula for calculating the principal curvature is:

[0109]

[0110] 2. Return the principal curvature set (principal curvature set ).

[0111] (4) Mesh defect detection algorithm:

[0112] 1. Detect overlapping surfaces: Traverse all patches and calculate the distance between patches. If the distance between two patches is less than the set threshold (such as 0.001mm), it is determined to be an overlapping surface.

[0113] 2. Detect holes: Traverse all edges and count the number of connected faces. If the number of connected faces on an edge is less than 2, the edge is considered to be a hole boundary and the hole area is further determined.

[0114] 3. Detect non-manifold edges: Traverse all edges and check whether the number of faces connected by the edge exceeds 2. If it exceeds, it is determined to be a non-manifold edge.

[0115] 4. Return the defect type and location.

[0116] (5) Feature-aware normal calculation algorithm in the intelligent optimization processing module: Feature-aware normal calculation is performed in the light-sensing optimization part. In the curvature weighted average calculation, higher weights are given to faces with larger areas to eliminate the "pseudo-angular" phenomenon. Sharp edges are marked and normal calculations are separated during sharp edge detection. In the smooth area, the weighted average of adjacent face normals is performed to achieve accurate normal calculation and feature preservation. The algorithm is as follows:

[0117] 1. Get the set of faces adjacent to the vertex

[0118] 2. For each adjacent patch Calculate patch area , calculate the patch normal .

[0119] Calculate the weighted average of vertex normals (weighted by patch area):

[0120]

[0121] Calculate the weighted average curvature (weight is patch curvature): for each adjacent patch , calculate its curvature , the curvature weighted average is:

[0122]

[0123] Comprehensive normal direction, considering the sharp edge detection result: if the vertex is on the sharp edge (the angle between adjacent faces ), then separate the normal calculation and retain the original surface normal direction, otherwise, take and The average value of is taken as the final normal direction and normalized.

[0124] (C) Through the Laplace smoothing algorithm, normal mean filtering algorithm, shape-preserving smoothing algorithm and mesh subdivision algorithm in the intelligent optimization processing module, the model output results after quality assessment are smoothed, filtered and normal redirected to determine the target model output results. In one embodiment, the lighting parameter configuration: the lighting model is automatically configured based on the material library and normal roughness. Different materials correspond to different lighting models and the parameters are dynamically adjusted. The backlight brightness is automatically enhanced in high-temperature scenes. The intelligent configuration of lighting parameters is innovatively realized to adapt to the needs of different scenes; surface smoothing component: a multi-algorithm fusion strategy is adopted for different regional features. Laplace smoothing + normal mean filtering is used in the curvature flat area. Shape-preserving smoothing is used in the feature-sensitive area and the original surface normal is reset for the sharp edge vertex. Butterfly subdivision + normal redirection is used in the complex surface area. Local subdivision is implemented in the stress-sensitive area in combination with the finite element preprocessing results and the normal deviation threshold is tightened. The optimization algorithm is selected according to the characteristics of different regions to balance the model smoothness and feature retention. For details, see (1) to (4) below:

[0125] (1) Laplace smoothing algorithm:

[0126] 1. For each vertex :Get its neighboring vertex set { }, calculate the average position of the vertices in the field and update the vertex positions:

[0127]

[0128] 2. Return the smoothed vertex position set:

[0129]

[0130] Among them, α is the smoothing coefficient, which is usually taken as .

[0131] (2) Normal mean filtering algorithm:

[0132] 1. For each vertex Get the normal set of neighboring vertices , calculate the normal mean:

[0133]

[0134] Normalized normal mean:

[0135]

[0136] in, is the filtered vertex normal set.

[0137] 2. Return the filtered vertex normal set.

[0138] (3) Shape-preserving smoothing algorithm:

[0139] 1. For each vertex If the vertex is on a sharp edge, the original position and normal are retained. Otherwise, a shape-preserving smoothing iteration is performed to calculate the weighted average position of the neighboring vertices:

[0140]

[0141] Among them, the weight Calculated based on edge length or other geometric features.

[0142] 0 Update vertex position:

[0143]

[0144] Update the normal direction, taking into account the curvature weight:

[0145]

[0146] 2. Return the smoothed vertex position set.

[0147] (4) Butterfly segmentation algorithm:

[0148] 1. For the midpoint of each triangle edge, calculate the new vertex position, for the two vertices connected by the edge and , and the vertices of adjacent triangles 、 :

[0149]

[0150] 2. Update the mesh topology, insert new vertices and reconnect edges.

[0151] 3. Return the subdivided mesh data.

[0152] Through gradient-driven adaptive technology, feature-aware normal optimization, and intelligent scene matching, it breaks through the efficiency and accuracy bottlenecks of traditional solutions, providing high-precision and high-efficiency model quality assurance for industrial simulation, and is suitable for complex R&D scenarios such as aerospace, automotive manufacturing, and mechanical engineering.

[0153] In daily simulation application scenarios, in order to ensure the performance of the model (including various components) in subsequent finite element analysis (such as structural strength analysis, heat conduction analysis, etc.) and actual production, the model must be strictly quality-tested. Taking the straight-board model as an example, users can use the file processing module of industrial simulation applications to easily upload model files (such as .x_t format) and convert the format. It will automatically recognize the file extension and call the corresponding conversion interface to convert it to the stp format, thereby ensuring that the model can be smoothly imported into the simulation system.

[0154] At the same time, the resolution of the model will be adjusted in real time according to the camera's position and viewing angle to achieve intelligent rendering effects with high definition in the near field and simplified in the far field, significantly improving the flexibility and efficiency of model rendering in industrial simulation applications.

[0155] After the above optimization process, the model's performance in finite element simulation analysis will be significantly improved. The improved mesh quality not only makes the analysis results more accurate, but also greatly shortens the calculation time, effectively avoiding simulation errors and calculation failures caused by mesh defects. This provides more reliable and efficient simulation support for application design, which in turn helps to improve the quality performance of products in actual production.

[0156] In one embodiment, for the quality of model rendering, a case study is also conducted based on industrial simulation applications: in the rendering of the model appearance, taking the thermos cup model as an example, its appearance needs to be rendered with high precision to show the visual effects under different lighting environments. However, traditional rendering methods have many problems, such as high rendering computing power consumption, decreased frame rate, invalid backlight surface patches participating in the calculation, wasting GPU resources, and rough normal calculations. These all affect the authenticity and rendering efficiency of the light and shadow effects. The model is imported for lighting rendering analysis, and the importance of each part of the model is analyzed in real time. For the screen border area of ​​the straight-panel model, the priority of ray tracing is increased, and the system automatically allocates more computing resources for ray tracing calculations to ensure that the lighting, reflection, and shadow effects of the screen border area are accurately presented; for non-important areas, the priority of ray tracing is reduced to reduce the input of computing resources.

[0157] Enable the backlight culling mechanism based on the face visibility judgment based on the normal direction. When rendering the model, the angle between the normal of each face of the model and the viewpoint direction is calculated, and the invisible faces facing away from the viewpoint are eliminated, reducing the rendering of invalid faces, effectively improving the GPU utilization, and then optimizing the normal calculation of the model. In the smooth areas of the model, the normal is calculated by weighted average considering the face area and curvature weight to eliminate the "pseudo-angular" phenomenon; at sharp edges, such as the corners of the thermos cup model, the sharp edges are marked and the normal calculation is separated, retaining the original face normal direction, so that the light and shadow transition of the straight model is more natural and retains the geometric features.

[0158] After optimization, when rendering the thermos cup model, the light and shadow effects in the visually critical areas are more realistic and delicate, the rendering accuracy of non-important areas is appropriately reduced without affecting the overall visual effect, the rendering efficiency and frame rate are greatly improved, and the utilization of GPU resources is more reasonable, which effectively solves the problems existing in traditional rendering methods and provides a more efficient and accurate simulation basis for model rendering.

[0159] Regarding the model quality evaluation and optimization method for industrial simulation applications provided in the aforementioned embodiment, an embodiment of the present invention provides a model quality evaluation and optimization device for industrial simulation applications, see Figure 5 The schematic diagram of the structure of a model quality evaluation and optimization device for industrial simulation applications is shown, and the device includes the following parts:

[0160] The model preprocessing unit 502 obtains the geometric model to be processed and determines the target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model;

[0161] The local encryption unit 504 performs meshing and finite element analysis on the target input model to determine the finite element calculation results, and performs local error estimation and analysis on the finite element calculation results to locally refine the model mesh of the target input model, determine the key areas of the model, perform local encryption on the key areas, and determine the model output results;

[0162] The evaluation and optimization unit 506 optimizes the model output results through a hierarchical model quality evaluation and optimization system to determine the optimized target model output results, wherein the model quality evaluation and optimization system includes: a gradient-driven data organization module, a multi-dimensional quality evaluation module and an intelligent optimization processing module.

[0163] The model quality assessment and optimization device for the above-mentioned industrial simulation application provided in the embodiment of the present application can significantly improve the model quality.

[0164] In one embodiment, when performing the steps of determining the target input model by performing format detection processing, format conversion processing and asynchronous processing on the geometric model, the above-mentioned model preprocessing unit 502 is also used to: perform format detection processing on the geometric model, determine the file extension of the geometric model file, and when it is detected that the geometric model file corresponding to the file extension is a non-target format file, upload the geometric model file to the cloud for format conversion processing to convert the non-target format file into a target format file; if the file format conversion is successful, delete the temporary files generated during the format detection processing and format conversion processing, and initialize the database to perform asynchronous processing on the geometric model according to the solution type corresponding to the geometric model in the initialized database to determine the target input model.

[0165] In one embodiment, when performing local error estimation analysis on the finite element calculation results, locally refining the model mesh of the target input model, and determining the key areas of the model, the above-mentioned local encryption unit 504 is also used to: determine the unit stress and the smooth stress based on the node reconstruction according to the finite element calculation results, and determine the error estimation value based on the unit stress and the smooth stress through the error estimator, so as to determine the key areas of the model according to the error estimation value.

[0166] In one embodiment, when performing the step of determining the key areas of the model based on the error estimate value, the above-mentioned local encryption unit 504 is also used to: when the error estimate value corresponding to any area of ​​the target input model is greater than a preset error threshold, it is determined that there is a stress mutation in the area, and the area is marked as the key area of ​​the model, so as to perform local encryption processing on the key area, and re-perform finite element calculation to determine the output result.

[0167] In one embodiment, when optimizing the model output results through the model quality evaluation and optimization system of the hierarchical architecture and determining the optimized target model output results, the above-mentioned evaluation and optimization unit 506 is also used to: utilize the data gradient dynamic partitioning algorithm and the cone constraint resolution decision algorithm in the gradient-driven data organization module to dynamically adjust the resolution of the model rendering, perform intelligent rendering processing on the model output results, and determine the optimized model output results; combine the illumination energy calculation algorithm, normal smoothness quantization algorithm, principal curvature calculation algorithm and mesh defect detection algorithm in the multidimensional quality evaluation module, perform multidimensional quality evaluation on the surface point radiance, vertex normal smoothness, principal curvature, defect type and defect position in the optimized model output results, and smooth, filter and normal redirect the model output results after quality evaluation through the Laplace smoothing algorithm, normal mean filtering algorithm, conformal smoothing algorithm and mesh subdivision algorithm in the intelligent optimization processing module to determine the target model output results.

[0168] In one embodiment, before the step of determining the optimized model output result, the above-mentioned evaluation optimization unit 506 is also used to: utilize the patch visibility judgment algorithm in the gradient-driven data organization module to eliminate the invisible patches facing away from the viewpoint during intelligent rendering processing.

[0169] In one embodiment, before the step of smoothing, filtering and normal redirecting the model output results after quality assessment through the Laplace smoothing algorithm, normal mean filtering algorithm, conformal smoothing algorithm and mesh subdivision algorithm in the intelligent optimization processing module to determine the target model output results, the above-mentioned evaluation optimization unit 506 is also used to: obtain a set of adjacent facets of each vertex, and use the feature-aware normal calculation algorithm in the multi-dimensional quality assessment module to analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet; determine the weighted average of the vertex normals and the weighted average of the curvature based on the facet area and the facet normal, and determine the target normal vector based on the weighted average of the vertex normals and the weighted average of the curvature to separate the sharp edges of the model from the normal.

[0170] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those in the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference can be made to the corresponding content in the aforementioned method embodiment.

[0171] An embodiment of the present invention provides a server. Specifically, the server includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned embodiments.

[0172] Figure 6 A structural diagram of a server provided in an embodiment of the present invention, wherein the server 100 includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected via the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0173] Memory 61 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage. Communication between the system network element and at least one other network element is achieved through at least one communication interface 63 (which may be wired or wireless), and may utilize the Internet, a wide area network, a local area network, a metropolitan area network, or the like.

[0174] The bus 62 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, Figure 6 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0175] Among them, the memory 61 is used to store programs, and the processor 60 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0176] The processor 60 may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above method may be performed by hardware integrated logic circuits or software instructions within the processor 60. The processor 60 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in conjunction with the embodiments of the present invention may be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules may be located in storage media well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or the like. The storage medium is located in the memory 61 , and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0177] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can be referred to the previous method embodiment and will not be repeated here.

[0178] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0179] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A model quality assessment optimization method for industrial simulation applications, characterized in that: The method comprises: Acquire a geometric model to be processed, and determine a target input model by performing format detection processing, format conversion processing, and asynchronous processing on the geometric model; Performing meshing and finite element analysis on the target input model to determine finite element calculation results, and performing local error estimation and analysis on the finite element calculation results to locally refine the model mesh of the target input model, determine key areas of the model, perform local encryption on the key areas, and determine model output results; The model output result is optimized by a hierarchical model quality assessment and optimization system to determine the optimized target model output result, wherein the model quality assessment and optimization system includes: a gradient-driven data organization module, a multi-dimensional quality assessment module, and an intelligent optimization processing module; The gradient-driven data organization module is used to dynamically adjust the resolution of the model rendering to perform intelligent rendering processing on the model output results, remove the invisible faces in the model output results, and determine the optimized model output results. Among them, the multi-dimensional quality assessment module is used to obtain a set of facets adjacent to each vertex, and analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet, and then determine the weighted average of the vertex normal and the weighted average of the curvature based on the facet area and the facet normal, and then determine the target normal vector based on the weighted average of the vertex normal and the weighted average of the curvature, so as to separate the sharp edges of the model from the normal. Among them, the intelligent optimization processing module is used to configure corresponding lighting models for different material configurations, dynamically adjust lighting parameters, and adopt multi-algorithm fusion strategies for different regional features. Laplace smoothing and normal mean filtering are used in areas with gentle curvature, conformal smoothing is performed in feature-sensitive areas, and the original surface normals of sharp edge vertices are reset. Butterfly subdivision is performed in complex surface areas, and normal redirection is performed. Local subdivision is performed in stress-sensitive areas based on the finite element preprocessing results, and the normal deviation threshold is tightened to select optimization algorithms according to the characteristics of different regions, balance the model smoothness, and retain features.

2. The model quality assessment and optimization method for industrial simulation applications according to claim 1, characterized in that: The step of determining a target input model by performing format detection processing, format conversion processing, and asynchronous processing on the geometric model includes: performing format detection processing on the geometric model to determine a file extension of the geometric model file, and when detecting that the geometric model file corresponding to the file extension is a non-target format file, uploading the geometric model file to a cloud for format conversion processing to convert the non-target format file into a target format file; If the file format conversion is successful, the temporary files generated during the format detection and format conversion processes are deleted, and the database is initialized to perform asynchronous processing on the geometric model according to the solution type corresponding to the geometric model in the initialized database to determine the target input model.

3. The model quality assessment and optimization method for industrial simulation applications according to claim 1, characterized in that: The step of performing local error estimation and analysis processing on the finite element calculation results to locally refine the model mesh of the target input model and determine the key areas of the model includes: The unit stress and the smooth stress based on the node reconstruction are determined according to the finite element calculation results, and an error estimation value is determined based on the unit stress and the smooth stress by an error estimator to determine the key area of ​​the model according to the error estimation value.

4. The model quality assessment and optimization method for industrial simulation applications according to claim 3, characterized in that: The step of determining the key area of ​​the model according to the error estimate comprises: When the error estimate corresponding to any area of ​​the target input model is greater than the preset error threshold, it is determined that there is a stress mutation in the area, and the area is marked as the key area of ​​the model, so that the key area is locally encrypted, and the finite element calculation is re-performed to determine the output result.

5. The model quality assessment and optimization method for industrial simulation applications according to claim 1, characterized in that: The step of optimizing the model output result by the hierarchical model quality evaluation and optimization system to determine the optimized target model output result includes: Utilizing the data gradient dynamic partitioning algorithm and the view cone constraint resolution decision algorithm in the gradient-driven data organization module, the resolution of the model rendering is dynamically adjusted, the model output result is intelligently rendered, and the optimized model output result is determined; The illumination energy calculation algorithm, normal smoothness quantification algorithm, principal curvature calculation algorithm and mesh defect detection algorithm in the multidimensional quality assessment module are combined to perform multidimensional quality assessment on the surface point radiance, vertex normal smoothness, principal curvature, defect type and defect position in the optimized model output results. The Laplace smoothing algorithm, normal mean filtering algorithm, conformal smoothing algorithm and mesh subdivision algorithm in the intelligent optimization processing module are used to smooth, filter and perform normal redirection processing on the model output results after quality assessment to determine the target model output results.

6. The model quality assessment and optimization method for industrial simulation applications according to claim 5, characterized in that: Before the step of determining the output result of the optimized model, the method includes: By utilizing the patch visibility judgment algorithm in the gradient-driven data organization module, invisible patches facing away from the viewpoint are eliminated during intelligent rendering processing.

7. The model quality assessment and optimization method for industrial simulation applications according to claim 5, characterized in that: Before the step of smoothing, filtering, and normal redirecting the model output result after quality assessment using the Laplace smoothing algorithm, normal mean filtering algorithm, shape-preserving smoothing algorithm, and mesh subdivision algorithm in the intelligent optimization processing module to determine the target model output result, the method includes: Obtaining a set of facets adjacent to each vertex, and using the feature-aware normal calculation algorithm in the multi-dimensional quality assessment module to analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet; According to the patch area and the patch normal, a weighted average of vertex normals and a weighted average of curvature are determined, and according to the weighted average of vertex normals and the weighted average of curvature, a target normal vector is determined to separate the sharp edge of the model from the normal.

8. A model quality assessment and optimization device for industrial simulation applications, characterized in that: The device comprises: A model preprocessing unit is configured to obtain a geometric model to be processed, and determine a target input model by performing format detection processing, format conversion processing, and asynchronous processing on the geometric model; a local encryption unit, performing meshing and finite element analysis on the target input model, determining finite element calculation results, and performing local error estimation and analysis on the finite element calculation results to locally refine the model mesh of the target input model, determine key areas of the model, perform local encryption on the key areas, and determine model output results; An evaluation and optimization unit optimizes the model output results through a hierarchical model quality evaluation and optimization system to determine an optimized target model output result, wherein the model quality evaluation and optimization system includes: a gradient-driven data organization module, a multi-dimensional quality evaluation module, and an intelligent optimization processing module; The gradient-driven data organization module is used to dynamically adjust the resolution of the model rendering to perform intelligent rendering processing on the model output results, remove the invisible faces in the model output results, and determine the optimized model output results. Among them, the multi-dimensional quality assessment module is used to obtain a set of facets adjacent to each vertex, and analyze and calculate each adjacent facet to determine the facet area and facet normal of each adjacent facet, and then determine the weighted average of the vertex normal and the weighted average of the curvature based on the facet area and the facet normal, and then determine the target normal vector based on the weighted average of the vertex normal and the weighted average of the curvature, so as to separate the sharp edges of the model from the normal. Among them, the intelligent optimization processing module is used to configure corresponding lighting models for different material configurations, dynamically adjust lighting parameters, and adopt multi-algorithm fusion strategies for different regional features. Laplace smoothing and normal mean filtering are used in areas with gentle curvature, conformal smoothing is performed in feature-sensitive areas, and the original surface normals of sharp edge vertices are reset. Butterfly subdivision is performed in complex surface areas, and normal redirection is performed. Local subdivision is performed in stress-sensitive areas based on the finite element preprocessing results, and the normal deviation threshold is tightened to select optimization algorithms according to the characteristics of different regions, balance the model smoothness, and retain features.

9. A server, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images