Structural topology optimization method and device, equipment and storage medium

Abstract

The invention discloses a structure topological optimization method and device, equipment and a storage medium, and the method comprises the steps: firstly obtaining topological optimization problem data of a target structural member, and inputting the topological optimization problem data into a decision model to obtain a recommended optimization strategy; and then, starting a CAE solver by taking the strategy as a parameter to carry out iterative calculation. In the calculation process, state information representing the optimization process is fed back to the decision model in real time for evaluation. And if the model judges that the strategy needs to be adjusted, updating solving parameters according to an output instruction and continuing iteration until a final optimization result is obtained. The decision model is based on historical case data and can learn complex mapping and adjustment strategies between optimization parameters and results, so that manual intervention is remarkably reduced, and optimization efficiency and result quality are improved. The method can be widely applied to the technical field of industrial design.

Description

Technical Field

[0001] This application relates to the field of industrial design technology, and in particular to a method, apparatus, device and storage medium for structural topology optimization. Background Technology

[0002] In the manufacturing of high-end equipment such as automobiles, aerospace, and construction machinery, structural lightweighting is a key approach to improving product performance (such as vehicle range and spacecraft carrying capacity) and achieving energy conservation and emission reduction. Topology optimization, as an advanced "inside-out" structural design method, can find the optimal material distribution scheme under given design space, load conditions, and performance constraints, thereby achieving the best balance between lightweighting and mechanical performance. In this process, computer-aided engineering (CAE) software plays a core role. Among them, Altair HyperMesh, as a mainstream pre-processing and optimization tool in the industry, is widely used in the entire topology optimization process, including geometry cleaning, mesh generation, load and boundary condition setting, optimization parameter definition, and result post-processing, and has become an indispensable part of modern structural design.

[0003] In related technologies, traditional topology optimization processes based on commercial tools such as HyperMesh still face significant challenges. First, this process heavily relies on the personal experience of engineers. Key decision-making steps such as setting optimization objectives, selecting constraints, and adjusting mesh density all require manual intervention, leading to deviations in optimization results due to different engineers' operations, making it difficult to guarantee the consistency and optimality of the solution. Second, the optimization process typically requires multiple manual adjustments to parameters (such as the number of optimization steps, penalty factor, and filter radius) and repeated finite element calculations, resulting in lengthy iteration cycles that can last for days or even weeks, severely restricting design efficiency. Furthermore, in complex engineering scenarios with multiple loads and constraints, traditional optimization algorithms are prone to convergence difficulties or produce unreasonable results such as checkerboard patterns and mesh dependencies, making it difficult to efficiently and robustly output optimal topologies that combine excellent mechanical performance and engineering practicality.

[0004] In summary, the problems with the relevant technologies urgently need to be addressed. Summary of the Invention

[0005] The purpose of this application is to at least partially solve one of the technical problems existing in the related art.

[0006] Therefore, one objective of the embodiments of this application is to provide a structural topology optimization method, apparatus, device, and storage medium.

[0007] To achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of this application include: On one hand, embodiments of this application provide a structural topology optimization method, the method comprising: Obtain the topology optimization problem data corresponding to the target structural component; The topology optimization problem data is input into a trained decision model, and the decision model makes predictions to obtain the output optimization strategy. Using the optimization strategy as the setting parameters, start the CAE solver to perform topology optimization iterative calculations; During the iterative calculation process, the state information representing the optimization process is input into the decision model, which evaluates the data and determines whether the optimization strategy needs to be adjusted. If the optimization strategy needs to be adjusted, the setting parameters are updated according to the adjustment instructions output by the decision model, and the iterative calculation process continues until the optimization strategy no longer needs to be adjusted, thus obtaining the structural topology optimization result corresponding to the target structural component.

[0008] In addition, the structural topology optimization method according to the above embodiments of this application may also have the following additional technical features: Furthermore, in one embodiment of this application, the decision model is trained through the following steps: Obtain a historical topology optimization case dataset and standardize the historical topology optimization case dataset to obtain a training dataset; Based on the training dataset, the decision model is trained using a hybrid strategy that combines supervised learning and reinforcement learning to obtain a well-trained decision model.

[0009] Furthermore, in one embodiment of this application, the step of obtaining a historical topology optimization case dataset and standardizing the historical topology optimization case dataset to obtain a training dataset includes: Obtain the mesh model, working parameters, and topology result labels corresponding to the sample structural components; The mesh model is subjected to size unification processing, and the working condition parameters are encoded to obtain the training dataset.

[0010] Furthermore, in one embodiment of this application, training the decision model using a hybrid strategy combining supervised learning and reinforcement learning based on the training dataset includes: During the supervised learning training phase, the working condition parameters in the training dataset are used as inputs, and the corresponding topology result labels are used as supervision signals to train the decision model to learn the mapping relationship from the optimization problem input to the topology result. Once the supervised learning training phase is completed, the decision model is placed as an agent in a simulation optimization environment for reinforcement learning training.

[0011] Furthermore, in one embodiment of this application, the step of placing the decision model as an intelligent agent in a simulated optimization environment for reinforcement learning training includes: The intelligent agent adjusts its actions by outputting parameters based on the current optimized state information; The parameter adjustment action is executed in the simulation optimization environment, and the new state and reward signal after execution are fed back. The reward signal is calculated based on the engineering performance index of the topology result obtained by the parameter adjustment action. Based on the reward signal, the policy of the decision model is updated using a reinforcement learning algorithm to maximize the cumulative reward.

[0012] Furthermore, in one embodiment of this application, the decision model is constructed using a graph convolutional network; the input of the graph convolutional network is graph structure data constructed based on the mesh model, wherein the nodes in the graph structure data are mesh cells and contain material properties and position coordinate features, and the edges in the graph structure data are the connection relationships between the mesh cells.

[0013] Furthermore, in one embodiment of this application, the step of starting the CAE solver to perform topology optimization iterative calculation using the optimization strategy as a setting parameter includes: The optimization strategy output by the decision model is written into the optimization configuration card of the HyperMesh software through a preset application interface. The HyperMesh software's built-in OptiStruct solver is invoked to perform topology optimization iterative calculations.

[0014] On the other hand, embodiments of this application provide a structural topology optimization device, the device comprising: The acquisition unit is used to acquire the topology optimization problem data corresponding to the target structural component; The prediction unit is used to input the topology optimization problem data into the trained decision model, and make predictions through the decision model to obtain the output optimization strategy. The computing unit is used to start the CAE solver to perform topology optimization iterative calculations using the optimization strategy as the setting parameters. An evaluation unit is used to input state information representing the optimization process into the decision model during the iterative calculation process, and the decision model evaluates the information to determine whether the optimization strategy needs to be adjusted. The adjustment unit is used to update the setting parameters according to the adjustment instructions output by the decision model if the optimization strategy needs to be adjusted, and continue the iterative calculation process until the optimization strategy no longer needs to be adjusted, thereby obtaining the structural topology optimization result corresponding to the target structural component.

[0015] On the other hand, embodiments of this application provide an electronic device, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above-described structural topology optimization method.

[0016] On the other hand, embodiments of this application also provide a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the aforementioned structural topology optimization method.

[0017] On the other hand, embodiments of this application also provide a computer program product, which includes a computer program stored in a computer-readable storage medium. The processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform a structural topology optimization method as described above.

[0018] The advantages and beneficial effects of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application: This application discloses a structural topology optimization method, apparatus, device, and storage medium, aiming to improve the problems of traditional methods relying on human experience, long iteration cycles, and poor adaptability to complex working conditions. The core of this application lies in introducing a pre-trained AI decision-making model to achieve intelligent optimization. Specifically, firstly, topology optimization problem data of the target structural component is acquired and input into the decision-making model to obtain a recommended optimization strategy. Subsequently, a CAE solver is started using this strategy as parameters for iterative calculation. During the calculation process, the state information representing the optimization process is fed back to the decision-making model in real time for evaluation. If the model determines that the strategy needs adjustment, it updates the solution parameters according to its output instructions and continues iterating until the final optimization result is obtained. This decision-making model, based on historical case data, can learn the complex mapping between optimization parameters and results, as well as adjustment strategies, thereby significantly reducing human intervention and improving optimization efficiency and result quality. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of this application or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions in this application. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0020] Figure 1 This is a schematic diagram illustrating the implementation environment of a structural topology optimization method provided in this application embodiment; Figure 2 This is a flowchart illustrating a structural topology optimization method provided in an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0022] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0024] In the manufacturing of high-end equipment such as automobiles, aerospace, and construction machinery, structural lightweighting is a key approach to improving product performance (such as vehicle range and spacecraft carrying capacity) and achieving energy conservation and emission reduction. Topology optimization, as an advanced "inside-out" structural design method, can find the optimal material distribution scheme under given design space, load conditions, and performance constraints, thereby achieving the best balance between lightweighting and mechanical performance. In this process, computer-aided engineering (CAE) software plays a core role. Among them, Altair HyperMesh, as a mainstream pre-processing and optimization tool in the industry, is widely used in the entire topology optimization process, including geometry cleaning, mesh generation, load and boundary condition setting, optimization parameter definition, and result post-processing, and has become an indispensable part of modern structural design.

[0025] In related technologies, traditional topology optimization processes based on commercial tools such as HyperMesh still face significant challenges. First, this process heavily relies on the personal experience of engineers. Key decision-making steps such as setting optimization objectives, selecting constraints, and adjusting mesh density all require manual intervention, leading to deviations in optimization results due to different engineers' operations, making it difficult to guarantee the consistency and optimality of the solution. Second, the optimization process typically requires multiple manual adjustments to parameters (such as the number of optimization steps, penalty factor, and filter radius) and repeated finite element calculations, resulting in lengthy iteration cycles that can last for days or even weeks, severely restricting design efficiency. Furthermore, in complex engineering scenarios with multiple loads and constraints, traditional optimization algorithms are prone to convergence difficulties or produce unreasonable results such as checkerboard patterns and mesh dependencies, making it difficult to efficiently and robustly output optimal topologies that combine excellent mechanical performance and engineering practicality.

[0026] In view of this, this application provides a structural topology optimization method, apparatus, device, and storage medium, aiming to improve the problems of traditional methods relying on human experience, long iteration cycles, and poor adaptability to complex working conditions. The core of this application lies in introducing a pre-trained AI decision-making model to achieve intelligent optimization. Specifically, firstly, topology optimization problem data of the target structural component is acquired and input into the decision-making model to obtain a recommended optimization strategy. Subsequently, a CAE solver is started using this strategy as parameters for iterative calculation. During the calculation process, the state information representing the optimization process is fed back to the decision-making model in real time for evaluation. If the model determines that the strategy needs to be adjusted, the solution parameters are updated according to its output instructions, and iteration continues until the final optimization result is obtained. This decision-making model, based on historical case data, can learn the complex mapping between optimization parameters and results, as well as adjustment strategies, thereby significantly reducing human intervention and improving optimization efficiency and result quality.

[0027] Please refer to Figure 1 , Figure 1 This diagram illustrates the implementation environment of a structural topology optimization method provided in this embodiment. In this implementation environment, the main hardware and software components involved include a terminal device 110 and a backend server 120. The terminal device 110 and the backend server 120 are connected by a communication link.

[0028] Specifically, the structural topology optimization method provided in this application embodiment can be executed independently on the terminal device 110 side, or based on data interaction between the terminal device 110 and the backend server 120. The terminal device 110 can be an in-vehicle terminal, such as the central control unit of a vehicle; the backend server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0029] The terminal device 110 and the backend server 120 can establish a communication connection via a wireless network or a wired network. This wireless or wired network uses standard communication technologies and / or protocols. The network can be the Internet or any other network, including but not limited to any combination of Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), mobile, wired or wireless networks, private networks, or virtual private networks.

[0030] Of course, this is understandable. Figure 1 The implementation environment described in this application is only one of the optional application scenarios for the structural topology optimization method provided in this embodiment. The actual application is not fixed. Figure 1 The software and hardware environment shown.

[0031] Below, in conjunction with the aforementioned description of the implementation environment, a structural topology optimization method provided in the embodiments of this application will be introduced and explained.

[0032] Please refer to Figure 2 , Figure 2 This is a schematic diagram of a structural topology optimization method provided in an embodiment of this application. The structural topology optimization method includes, but is not limited to: Step 210: Obtain the topology optimization problem data corresponding to the target structural component; Step 220: Input the topology optimization problem data into the trained decision model, and use the decision model to make predictions to obtain the output optimization strategy; Step 230: Using the optimization strategy as the setting parameters, start the CAE solver to perform topology optimization iterative calculation; Step 240: During the iterative calculation, the state information representing the optimization process is input into the decision model, which evaluates the data and determines whether the optimization strategy needs to be adjusted. Step 250: If the optimization strategy needs to be adjusted, the setting parameters are updated according to the adjustment instructions output by the decision model, and the iterative calculation process continues until the optimization strategy no longer needs to be adjusted, and the structural topology optimization result corresponding to the target structural component is obtained.

[0033] This application provides a structural topology optimization method, aiming to automate and intelligentize the topology optimization process, thereby overcoming the shortcomings of traditional methods such as excessive reliance on human experience, long iteration cycles, and poor adaptability under complex working conditions. Figure 2 As shown, the core of this application lies in constructing an intelligent closed-loop optimization process: This process begins by acquiring topology optimization problem data for the target structural component, and then inputs the data into a pre-trained decision model. This decision model, as the intelligent core, can predict and output an optimization strategy for the current problem. This strategy is used as initial parameters to start the CAE solver for iterative topology optimization calculations. During the iterative calculation process, the system continuously feeds back state information representing the optimization process to the decision model for evaluation. Based on this, the decision model determines whether the current optimization strategy needs adjustment. If adjustment is needed, the solution parameters are updated according to the instructions output by the model, and the iterative process continues. This "execution-monitoring-evaluation-adjustment" cycle continues until the decision model determines that no further adjustment is needed, at which point the final structural topology optimization result is output.

[0034] The technical solution provided in this application deeply integrates AI decision-making capabilities into the traditional CAE optimization process, significantly reducing human intervention. This not only greatly shortens the optimization cycle but also effectively handles complex working conditions, ensuring the reliability and engineering applicability of the optimization results.

[0035] The structural topology optimization method provided in the embodiments of this application will be described and explained in detail below.

[0036] The core of this application's embodiments lies in constructing an automated closed-loop optimization system with an artificial intelligence decision-making model as its intelligent core. This method transforms the traditional iterative process, reliant on engineer experience and based on trial and error, into a data-driven, intelligent process capable of real-time perception and autonomous decision-making. For example... Figure 2As shown, the implementation of this method begins with acquiring the topology optimization problem data corresponding to the target structural component. This topology optimization problem data is a comprehensive input; it can be a complete set of information defining the entire optimization problem. Specifically, the topology optimization problem data can include the design space of the target structural component, i.e., the envelope region where the material is allowed to distribute, usually represented by a 3D CAD model; detailed loads and boundary conditions, such as the point of application, magnitude, and direction of forces, as well as fixed constraint locations; clear optimization objectives, such as maximizing stiffness (minimizing flexibility) or maximizing the natural frequency of a specific order; and constraints that need to be satisfied, such as volume fraction constraints, i.e., requiring the material used in the final structure to not exceed a certain percentage of the original design space. Acquiring this topology optimization problem data is the foundation for initiating the entire intelligent optimization process.

[0037] In step 220, these fully defined topology optimization problem data can be input into a pre-trained decision model. This decision model is the intelligent decision-making hub of this application, and it is trained using a hybrid strategy combining supervised learning and reinforcement learning, based on a large amount of historical topology optimization case data. The supervised learning stage aims to allow the model to learn and imitate the successful experiences of human experts. Its training data consists of "input-output" pairs: the "input" is the historical operating condition parameters, and the "output" is the topology result or parameter setting that has been verified as successful under that operating condition. Through this stage, the model can quickly establish a mapping relationship from problem description to preliminary solution. The reinforcement learning stage goes a step further, placing the trained model in a simulated optimization environment, allowing it to interact with the environment as an intelligent agent. The agent selects actions (such as adjusting the penalty factor) based on the current optimization state (e.g., density field, objective function change curve). After the environment (i.e., the simulated CAE solver) executes the action, it feeds back a new state and reward signal. This reward signal can be calculated based on the engineering performance of the optimization result (e.g., stiffness-to-weight ratio). By maximizing cumulative rewards, the model learns superior decision-making strategies that go beyond simple imitation, enabling it to make better judgments than historical experience in complex situations. Therefore, when topology optimization problem data is input, the model can perform comprehensive analysis and prediction, outputting a targeted optimization strategy with a high probability of success. This strategy may specifically include key parameter settings in the topology optimization algorithm, such as penalty factors to drive intermediate density units to converge to 0 or 1, filtering radii to control the minimum structure size and eliminate checkerboard patterns, and suggested optimization iteration steps.

[0038] In step 230, the optimization strategy recommended by the decision model is used as initial setup parameters and directly passed to the CAE solver (e.g., the OptiStruct solver integrated in HyperMesh), initiating iterative topology optimization calculations. This step can be automated through a developed application programming interface (API), such as using a Python script to call HyperMesh's TCL / TK interface to automatically write the recommended parameters into the corresponding cards (e.g., DOPT, DCON, etc.) in the solver's input file (.fem file), eliminating the need for manual input by engineers. This not only avoids human error but also significantly improves setup efficiency. The solver will then begin iterative calculations based on these parameters, gradually updating the pseudo-density of each element within the design region to guide the material towards the optimal layout.

[0039] During the iterative calculations of the solver, this application implements real-time process monitoring and dynamic adjustment, which is a key difference from traditional static optimization processes. The system periodically or at key nodes collects state information representing the optimization process. This state information is a "snapshot" of the optimization process, mainly including but not limited to: the pseudo-density distribution of all elements in the current iteration step, which reflects the intermediate form of the structure; the change history of the objective function (such as overall compliance) to determine the convergence trend; and the satisfaction status of constraints (such as volume fraction), etc. This real-time state information is immediately input back into the decision model for evaluation. The decision model can diagnose the current optimization process, determining whether it is healthy and efficient, such as whether convergence is slow or whether oscillations are likely to occur. Based on this diagnosis, the model decides whether the optimization strategy needs to be adjusted to guide the optimization in a better direction.

[0040] In step 250, if the decision model deems it necessary to adjust the strategy after evaluation, it will output specific adjustment instructions, such as "increase the penalty factor from 3.0 to 3.2" or "reduce the filter radius by 5%". Upon receiving the instructions, the system will automatically update the corresponding settings parameters of the CAE solver. Subsequently, the solver continues iterative calculations based on the new parameters, while the system continues to monitor and prepare for the next evaluation. This intelligent closed loop of "execution-monitoring-evaluation-adjustment" will continue to run until the decision model determines, based on state information, that the optimization has stabilized or met performance requirements, meaning no further adjustment of the optimization strategy is needed. At this point, the loop terminates, and the CAE solver outputs the final structural topology optimization result, namely a lightweight topology with optimal material distribution that satisfies all engineering constraints. Furthermore, in practical engineering applications, the input of this optimization task, the strategy sequence given by the decision model, and the final verified results can be automatically collected and archived by the system as new training samples for periodic incremental learning or fine-tuning of the decision model, thereby enabling the entire system to continuously evolve and adapt to new scenarios.

[0041] It is understood that the structural topology optimization method provided in this application obtains topology optimization problem data of the target structural component and inputs it into a decision model to obtain a recommended optimization strategy. Subsequently, a CAE solver is started using this strategy as parameters for iterative calculation. During the calculation process, the status information representing the optimization process is fed back to the decision model in real time for evaluation. If the model determines that the strategy needs to be adjusted, it updates the solution parameters according to its output instructions and continues iterating until the final optimization result is obtained. This decision model, based on historical case data, can learn the complex mapping between optimization parameters and results, as well as adjustment strategies, thereby significantly reducing manual intervention and improving optimization efficiency and result quality.

[0042] Specifically, in some embodiments, the decision model is trained through the following steps: Obtain a historical topology optimization case dataset and standardize the historical topology optimization case dataset to obtain a training dataset; Based on the training dataset, the decision model is trained using a hybrid strategy that combines supervised learning and reinforcement learning to obtain a well-trained decision model.

[0043] Specifically, in some embodiments, the step of obtaining a historical topology optimization case dataset and standardizing the historical topology optimization case dataset to obtain a training dataset includes: Obtain the mesh model, working parameters, and topology result labels corresponding to the sample structural components; The mesh model is subjected to size unification processing, and the working condition parameters are encoded to obtain the training dataset.

[0044] The step of training the decision model using a hybrid strategy combining supervised learning and reinforcement learning based on the training dataset includes: During the supervised learning training phase, the working condition parameters in the training dataset are used as inputs, and the corresponding topology result labels are used as supervision signals to train the decision model to learn the mapping relationship from the optimization problem input to the topology result. Once the supervised learning training phase is completed, the decision model is placed as an agent in a simulation optimization environment for reinforcement learning training.

[0045] Specifically, in some embodiments, placing the decision model as an agent in a simulated optimization environment for reinforcement learning training includes: The intelligent agent adjusts its actions by outputting parameters based on the current optimized state information; The parameter adjustment action is executed in the simulation optimization environment, and the new state and reward signal after execution are fed back. The reward signal is calculated based on the engineering performance index of the topology result obtained by the parameter adjustment action. Based on the reward signal, the policy of the decision model is updated using a reinforcement learning algorithm to maximize the cumulative reward.

[0046] This application also provides a training method for a decision model used in structural topology optimization. The ultimate goal of this method is to obtain an intelligent model that can understand optimization problems, formulate optimization strategies, and dynamically adjust them. Its core lies in employing a phased, hybrid training strategy that combines supervised learning and reinforcement learning.

[0047] The training process of the decision-making model will be described and explained in detail below.

[0048] The first step is the preparation of training data. This method is based on acquiring a high-quality dataset of historical topology optimization cases. These cases originate from successful past engineering projects, such as optimization projects for automotive chassis and aircraft bracket components. Each case is a complete data unit, containing not only the starting point of the optimization problem—the geometric model of the design space and its discretized finite element mesh information—but also all the key parameters defining the optimization problem, such as applied load conditions, set boundary constraints, explicit optimization objectives (e.g., minimizing compliance), and constraints that must be followed (e.g., upper limits on volume fraction). Furthermore, each case includes optimization results calculated by a CAE solver and ultimately verified as successful by engineers; this is typically represented by the final pseudo-density distribution map of each element.

[0049] It should be noted that, in this embodiment of the application, since these historical data may come from different projects and be created by different engineers, their formats, mesh sizes, and parameter description methods may differ. Therefore, standardized preprocessing is generally required before use. This process may include unifying the mesh to the same cell type and size, encoding non-numerical working condition parameters (such as load application surfaces) into standard numerical vectors, and normalizing the resulting data. After this series of operations, the original, heterogeneous historical data is transformed into a standardized dataset that can be used for model training.

[0050] Next, we enter the core stage of model training, which involves training the decision model using a hybrid strategy. In this embodiment, the hybrid strategy is not simply a parallel application of two learning methods, but rather an ordered and interconnected process. The first stage is supervised learning. In this stage, the "optimization problem parameters" in the training dataset serve as input features, while the corresponding "successful optimization results" serve as labels or supervision signals. The learning objective of the model is to establish a mapping relationship from input to output, that is, to learn to predict an optimization scheme or parameter setting that closely approximates historical successful cases when given an optimization problem. This is equivalent to allowing the model to quickly learn and imitate the successful experiences of human experts, providing it with a high-performance initial starting point, avoiding the huge cost of random exploration from scratch, and ensuring the basic rationality of the model's behavior.

[0051] After supervised learning equips the model with basic imitation capabilities, training enters the second stage—reinforcement learning. In this stage, the pre-trained model is placed as an agent within a simulated topology optimization environment. The agent continuously interacts with the environment: it observes the current state of the optimization process (e.g., the current density field, objective function value) and then selects an action based on its internal policy (e.g., fine-tuning the penalty factor). After the environment (i.e., the simulated CAE solution process) executes this action, it provides a new state and a reward signal. This reward signal can be calculated from the engineering performance metrics of the topology result (e.g., the stiffness-to-weight ratio) and used to evaluate the quality of the decision action. The model's goal is no longer imitation, but rather learning how to maximize long-term cumulative rewards by trying different action sequences. This means it learns optimization itself, exploring how to adjust its policy to achieve better performance than existing solutions in the training data. This process enables the model to transcend simple imitation, facilitating the discovery of novel and efficient optimization strategies that surpass historical experience.

[0052] Through the hybrid training of the two stages mentioned above, the final decision model possesses both a reliable knowledge base from supervised learning and autonomous optimization and decision-making capabilities from reinforcement learning, thus enabling it to intelligently cope with various complex structural topology optimization problems.

[0053] Specifically, in some embodiments, the decision model is constructed using a graph convolutional network; the input of the graph convolutional network is graph structure data constructed based on the grid model, wherein the nodes in the graph structure data are grid cells and contain material properties and position coordinate features, and the edges in the graph structure data are the connection relationships between the grid cells.

[0054] In this application embodiment, a preferred implementation of the decision model is also provided, namely, using a Graph Convolutional Network (GCN / GNN) as the core architecture of the model. The fundamental reason for this choice is that the finite element mesh on which topology optimization depends is mathematically a graph structure, and the Graph Convolutional Network is a deep learning model specifically designed to process such non-Euclidean spatial data, which can better capture the topological and geometric relationships contained in the mesh data.

[0055] Specifically, in this embodiment, traditional mesh data is transformed into graph-structured data that can be directly processed by graph convolutional networks. This transformation process gives the data a new meaning that is more suitable for AI understanding. In graph-structured data, nodes are defined as each element in the finite element mesh. Each node is no longer an isolated data point, but carries rich feature information, including at least the material properties of the element (such as Young's modulus, Poisson's ratio, and density) and its position coordinates in three-dimensional space. Edges in graph-structured data are defined as the connection relationships between mesh elements. Specifically, if two elements share a face, edge, or vertex in the mesh (i.e., are adjacent to each other), an edge is established between their corresponding nodes. This connection relationship is crucial because it defines the path of stress wave transmission and load distribution in the physical world. Through edge connections, graph convolutional networks can simulate the transmission effects of physical fields in structures, such as how a force on one element affects its adjacent elements.

[0056] During operation, graph convolutional networks (GCNNs) perform multiple convolutional operations on the input graph structure data. At each layer, the network executes a message-passing mechanism: each node aggregates the feature information of its immediate neighbors and updates it in conjunction with its own features. Through the stacking of multiple layers, a node eventually gains access to information about nodes beyond its multiple connections. This process allows the model to learn the mechanical properties of the structure from a local to a global perspective. For example, the model can identify which regions' cells are interconnected to form an efficient force flow path, and which connection patterns might lead to stress concentration. It is this powerful perception of the structure's inherent connection patterns that enables GCNNs to understand topology optimization problems more naturally and accurately than traditional convolutional neural networks (CNNs) that treat grids as regular images, thus making more reasonable parameter recommendations and adjustment decisions, effectively improving the quality and reliability of optimization results.

[0057] This application embodiment also provides a structural topology optimization device, the device comprising: The acquisition unit is used to acquire the topology optimization problem data corresponding to the target structural component; The prediction unit is used to input the topology optimization problem data into the trained decision model, and make predictions through the decision model to obtain the output optimization strategy. The computing unit is used to start the CAE solver to perform topology optimization iterative calculations using the optimization strategy as the setting parameters. An evaluation unit is used to input state information representing the optimization process into the decision model during the iterative calculation process, and the decision model evaluates the information to determine whether the optimization strategy needs to be adjusted. The adjustment unit is used to update the setting parameters according to the adjustment instructions output by the decision model if the optimization strategy needs to be adjusted, and continue the iterative calculation process until the optimization strategy no longer needs to be adjusted, thereby obtaining the structural topology optimization result corresponding to the target structural component.

[0058] Reference Figure 3 This application provides an electronic device, including: At least one processor 310; At least one memory 320 is used to store at least one program; When at least one program is executed by at least one processor 310, the at least one processor 310 implements the above-described structural topology optimization method.

[0059] Similarly, the content of the above method embodiments is applicable to the embodiments of this electronic device. The specific functions implemented by the embodiments of this electronic device are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0060] This application embodiment also provides a computer-readable storage medium storing a program executable by a processor 310, which, when executed by the processor 310, is used to perform the above-described structural topology optimization method.

[0061] Similarly, the content of the above method embodiments is applicable to the present computer-readable storage medium embodiments. The specific functions implemented by the present computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0062] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. The processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the above-described structural topology optimization method.

[0063] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0064] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0065] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part 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 to cause a device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0066] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0067] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0068] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0069] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0070] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0071] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A structural topology optimization method, characterized in that, The method includes: Obtain the topology optimization problem data corresponding to the target structural component; The topology optimization problem data is input into a trained decision model, and the decision model makes predictions to obtain the output optimization strategy. Using the optimization strategy as the setting parameters, start the CAE solver to perform topology optimization iterative calculations; During the iterative calculation process, the state information representing the optimization process is input into the decision model, which evaluates the data and determines whether the optimization strategy needs to be adjusted. If the optimization strategy needs to be adjusted, the setting parameters are updated according to the adjustment instructions output by the decision model, and the iterative calculation process continues until the optimization strategy no longer needs to be adjusted, thus obtaining the structural topology optimization result corresponding to the target structural component.

2. The structural topology optimization method according to claim 1, characterized in that, The decision model is trained through the following steps: Obtain a historical topology optimization case dataset and standardize the historical topology optimization case dataset to obtain a training dataset; Based on the training dataset, the decision model is trained using a hybrid strategy that combines supervised learning and reinforcement learning to obtain a well-trained decision model.

3. The structural topology optimization method according to claim 2, characterized in that, The process of obtaining a historical topology optimization case dataset and standardizing the historical topology optimization case dataset to obtain a training dataset includes: Obtain the mesh model, working parameters, and topology result labels corresponding to the sample structural components; The mesh model is subjected to size unification processing, and the working condition parameters are encoded to obtain the training dataset.

4. The structural topology optimization method according to claim 3, characterized in that, The step of training the decision model using a hybrid strategy combining supervised learning and reinforcement learning based on the training dataset includes: During the supervised learning training phase, the working condition parameters in the training dataset are used as inputs, and the corresponding topology result labels are used as supervision signals to train the decision model to learn the mapping relationship from the optimization problem input to the topology result. Once the supervised learning training phase is completed, the decision model is placed as an agent in a simulation optimization environment for reinforcement learning training.

5. The structural topology optimization method according to claim 4, characterized in that, The step of placing the decision model as an agent in a simulated optimization environment for reinforcement learning training includes: The intelligent agent adjusts its actions by outputting parameters based on the current optimized state information; The parameter adjustment action is executed in the simulation optimization environment, and the new state and reward signal after execution are fed back. The reward signal is calculated based on the engineering performance index of the topology result obtained by the parameter adjustment action. Based on the reward signal, the policy of the decision model is updated using a reinforcement learning algorithm to maximize the cumulative reward.

6. A structural topology optimization method according to any one of claims 3-5, characterized in that, The decision model is constructed using a graph convolutional network; the input of the graph convolutional network is graph structure data constructed based on the grid model, where the nodes in the graph structure data are grid cells and contain material properties and location coordinate features, and the edges in the graph structure data are the connection relationships between the grid cells.

7. The structural topology optimization method according to claim 1, characterized in that, The step of starting the CAE solver for topology optimization iterative calculation using the optimization strategy as the setting parameter includes: The optimization strategy output by the decision model is written into the optimization configuration card of the HyperMesh software through a preset application interface. The HyperMesh software's built-in OptiStruct solver is invoked to perform topology optimization iterative calculations.

8. A structural topology optimization device, characterized in that, The device includes: The acquisition unit is used to acquire the topology optimization problem data corresponding to the target structural component; The prediction unit is used to input the topology optimization problem data into the trained decision model, and make predictions through the decision model to obtain the output optimization strategy. The computing unit is used to start the CAE solver to perform topology optimization iterative calculations using the optimization strategy as the setting parameters. An evaluation unit is used to input state information representing the optimization process into the decision model during the iterative calculation process, and the decision model evaluates the information to determine whether the optimization strategy needs to be adjusted. The adjustment unit is used to update the setting parameters according to the adjustment instructions output by the decision model if the optimization strategy needs to be adjusted, and continue the iterative calculation process until the optimization strategy no longer needs to be adjusted, thereby obtaining the structural topology optimization result corresponding to the target structural component.

9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a structural topology optimization method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to implement a structural topology optimization method as described in any one of claims 1-7.

Images