Method and system for adjusting topological structure of material and electronic equipment

Abstract

The invention is suitable for the field of computer aided design and engineering, and provides a method and system for adjusting a material topological structure and electronic equipment, and the method comprises the steps: determining a structure optimization task based on an obtained to-be-optimized material topological structure and optimization demand information; based on a target analysis method, decomposing the structure optimization task to obtain a plurality of optimization sub-tasks; distributing the optimization sub-tasks to a plurality of target neural network processors, and enabling each target neural network processor to solve at least one optimization sub-task to obtain a local optimal solution corresponding to the optimization sub-task; a global optimal solution corresponding to the structure optimization task is obtained based on the local optimal solution, and the global optimal solution comprises a plurality of second manufacturing parameters and is used for determining a target manufacturing parameter at each target position; and adjusting the first manufacturing parameter at each target position in the to-be-optimized material topological structure into a corresponding target manufacturing parameter to obtain an optimized material topological structure. According to the scheme, the optimization efficiency of the material topological structure can be improved.

Description

Technical Field

[0001] The present application belongs to the field of computer-aided design and engineering, and in particular relates to a method, system and electronic equipment for adjusting the topological structure of a material. Background Art

[0002] In the field of industrial engineering design and manufacturing, material topology describes the distribution of materials in the design space and is used to guide structural design. By adjusting the material distribution within a given design domain, the material topology can be optimized to obtain a material topology with the best performance that meets user needs. However, as the design variables increase, the complexity of the solution increases, and the optimization efficiency of the material topology decreases. Summary of the invention

[0003] The embodiments of the present application provide a method, system and electronic device for adjusting the material topological structure to solve the problem of reduced optimization efficiency of the material topological structure in the prior art.

[0004] A first aspect of an embodiment of the present application provides a method for adjusting a material topological structure, comprising: Based on the obtained material topology structure to be optimized and the optimization requirement information, a structural optimization task is determined; the material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter; Based on the target analysis method, the structural optimization task is decomposed to obtain multiple optimization subtasks; Allocating the optimization subtasks to a plurality of target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks; Based on the local optimal solution, a global optimal solution corresponding to the structural optimization task is obtained, wherein the global optimal solution includes a plurality of second manufacturing parameters, and the second manufacturing parameters are used to compare with the structural manufacturing constraint parameters to determine the target manufacturing parameters at each target position; the target manufacturing parameters are the second manufacturing parameters or the structural manufacturing constraint parameters; The first manufacturing parameter at each target position in the material topological structure to be optimized is adjusted to the target manufacturing parameter corresponding to the target position to obtain an optimized material topological structure.

[0005] A second aspect of an embodiment of the present application provides a system for adjusting a material topological structure, including: A determination module, used to determine the structural optimization task based on the acquired material topology structure to be optimized and the optimization requirement information; the material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter; A decomposition module, used for decomposing the structural optimization task based on a target analysis method to obtain a plurality of optimization subtasks; An allocation and solving module is used to allocate the optimization subtasks to a plurality of target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks; A solution module, used for obtaining a global optimal solution corresponding to the structural optimization task based on a local optimal solution, wherein the global optimal solution includes a plurality of second manufacturing parameters, and the second manufacturing parameters are used for comparing with the structural manufacturing constraint parameters to determine a target manufacturing parameter at each of the target positions; the target manufacturing parameter is the second manufacturing parameter or the structural manufacturing constraint parameter; An adjustment module is used to adjust the first manufacturing parameter at each target position in the material topology structure to be optimized to the target manufacturing parameter corresponding to the target position, so as to obtain an optimized material topology structure.

[0006] A third aspect of an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0007] A fourth aspect of an embodiment of the present application provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0008] A fifth aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0009] As can be seen from the above, the present application uses a target analysis method to decompose the determined structural optimization task into multiple optimization subtasks. After the task is decomposed, the solution complexity of each optimization subtask is greatly reduced. Afterwards, multiple target neural network processors are used to solve the optimization subtasks respectively to obtain the local optimal solution corresponding to each optimization subtask. This step is achieved through parallel solution through the target neural network processor to improve the solution efficiency. The global optimal solution corresponding to the structural optimization task is then determined through the local optimal solution, and the second manufacturing parameter in the global optimal solution is compared with the corresponding structural manufacturing constraint parameter to determine the target manufacturing parameter, and the first manufacturing parameter in the material topology to be optimized is adjusted to the target manufacturing parameter to achieve the adjustment of the material topology and obtain the optimized material topology. The reduction in solution complexity and the improvement in solution efficiency jointly improve the optimization efficiency of the material topology. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 It is a flow chart of a method for adjusting a material topological structure provided in an embodiment of the present application; Figure 2 It is a task linear conversion flowchart based on MMA provided in an embodiment of the present application; Figure 3 It is a structural diagram of a material topological structure adjustment system provided in an embodiment of the present application; Figure 4 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0013] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0014] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0015] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0017] In a specific implementation, the terminal described in the embodiments of the present application includes, but is not limited to, other portable devices such as mobile phones, laptop computers, or tablet computers with touch-sensitive surfaces (e.g., touch screen displays and / or touch pads). It should also be understood that in some embodiments, the device is not a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., touch screen displays and / or touch pads).

[0018] In the following discussion, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse and / or joystick.

[0019] The terminal supports various applications, such as one or more of the following: a drawing application, a presentation application, a word processing application, a website creation application, a disk burning application, a spreadsheet application, a game application, a telephone application, a video conferencing application, an email application, an instant messaging application, a workout support application, a photo management application, a digital camera application, a digital camcorder application, a web browsing application, a digital music player application, and / or a digital video player application.

[0020] Various applications that can be executed on the terminal can use at least one common physical user interface device such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and corresponding information displayed on the terminal can be adjusted and / or changed between applications and / or within corresponding applications. In this way, the common physical architecture of the terminal (e.g., the touch-sensitive surface) can support various applications with user interfaces that are intuitive and transparent to the user.

[0021] It should be understood that the size of the serial numbers of the steps in this embodiment does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0022] With the development of materials science, computer technology and optimization algorithms, material topology optimization has gradually been applied in the field of engineering design and manufacturing. For example, by adjusting the structural weight of aircraft to improve fuel efficiency, by reducing the weight of automobile parts to improve the safety and driving performance of automobiles, and so on. Material topology optimization refers to finding the optimal material distribution within a given design domain to meet the user's specific performance requirements. That is, under the constraints of specific volume, mass or other specific performance requirements, the material distribution of the material topology is adjusted to obtain a material topology that meets the user's specific performance requirements and has the best performance.

[0023] When adjusting the material topology to achieve its optimization, the increase in design complexity such as the increase in design variables will increase the complexity and difficulty of the solution, making it difficult to complete the solution efficiently and achieve rapid optimization and adjustment of the material topology.

[0024] In this regard, the present application provides a method, system and electronic device for adjusting the material topological structure to improve the optimization efficiency of the material topological structure.

[0025] In order to illustrate the technical solution described in this application, a specific embodiment is provided below for illustration.

[0026] See also Figure 1 , Figure 1 1 is a flow chart of a method for adjusting the material topology provided in an embodiment of the present application. Figure 1 As shown, a method for adjusting the topological structure of a material comprises the following steps: Step 101, determining a structural optimization task based on the acquired topological structure of the material to be optimized and the optimization requirement information.

[0027] The material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter.

[0028] The material topology to be optimized is the material topology that needs to be optimized given by the user. The optimization requirement information is the limited configuration information corresponding to the adjustment and optimization of the material topology to be optimized, input by the user, including the design domain, design variables, objective function and constraints, etc. The structural optimization task is the task that needs to be solved, and its purpose is to make the target performance of the material topology structure reach the optimal under given conditions. The structural manufacturing constraint parameters are the constraint parameters to be followed in the actual production process to realize the production and manufacturing of the material topology structure, usually the parameter range. The target position is the position with manufacturing parameters, which is located in the design domain. The first manufacturing parameter is the original manufacturing parameter of the material topology to be optimized.

[0029] Based on the topological structure of the material to be optimized and the optimization requirement information, the structural optimization task is determined, the optimization scope is clarified, and the computing resources are avoided from being wasted in irrelevant areas, thereby improving the solution efficiency and thus the optimization efficiency.

[0030] In some embodiments, an interactive interface is provided to facilitate the user to import the material topology to be optimized and input the optimization requirement information through the interactive interface. In addition, the user can also view and obtain the adjusted optimized material topology and other information through the interactive interface.

[0031] In some embodiments, the structural optimization task is determined based on the acquired material topology structure to be optimized and the optimization requirement information, including: extracting the optimization configuration parameters in the optimization requirement information; the optimization configuration parameters include the design domain, design variables, objective function and constraints; taking the material topology structure to be optimized as the optimization object, and obtaining the optimization task including the optimization object and the optimization configuration parameters.

[0032] The design domain is the physical area in the material topology that allows the adjustment of manufacturing parameters, such as the wings of an aircraft, the hood of a car, etc. The design variables are the physical quantities that can be adjusted within the design domain, such as the cross-sectional dimensions of the connecting rod, etc. The objective function is an expression that quantifies the performance of the structure. The constraints are the adjustable range of the design variables, such as stress constraints, mass constraints, material properties, etc.

[0033] The topological structure of the material to be optimized is the optimization object in the structural optimization task, and the optimization configuration parameters composed of the design domain, design variables, objective function and constraints are the optimization guidance information in the structural optimization task.

[0034] Through the extraction operation, the optimization configuration parameters in the structural optimization task are clarified, which facilitates the accurate and effective implementation of structural optimization.

[0035] In some embodiments, when the task scenario complexity of the structural optimization task is greater than or equal to the set scenario complexity, the neural network processing method is determined as the target analysis method; or, when the task scenario complexity of the structural optimization task is less than the set scenario complexity, the finite element analysis method is determined as the target analysis method; or, when a configuration analysis method is configured in the optimization requirement configuration information, the configuration analysis method is determined as the target analysis method, and the configuration analysis method is the neural network processing method or the finite element analysis method.

[0036] The target analysis method is used to analyze and decompose the structural optimization task to simplify the complex task. The target analysis method is a neural network processing method or a finite element analysis (FEA). The neural network processing method can be a physics-informed neural network (PINN) method.

[0037] Determining the appropriate target analysis method to achieve task decomposition will help optimize solution resource allocation and improve optimization efficiency.

[0038] In some embodiments, the task scenario complexity is determined by multi-dimensional factors in the structural optimization task, such as physical field coupling, geometric dimension, number of design variables, etc. The target analysis method is determined based on the comparison between the task scenario complexity and the set scenario complexity. For example, when encountering multi-scale physical fields, such as acoustic-solid coupling, fluid-solid thermal coupling, etc., a neural network processing method, such as the PINN method, is selected.

[0039] In some embodiments, when the task scenario complexity of the structure optimization task is less than the set scenario complexity, the neural network processing method can also be determined as the target analysis method.

[0040] In some embodiments, when the user inputs optimization requirement information, the user can set the analysis method. If a configuration analysis method is configured in the optimization requirement information input by the user, the configuration analysis method can be used as the target analysis method to meet the user's needs.

[0041] In some embodiments, the target analysis method is determined based on the user's solution accuracy.

[0042] In some embodiments, the target analysis method may also be a machine learning method or other hybrid methods, such as a support vector machine method.

[0043] Step 102: based on the target analysis method, decompose the structural optimization task to obtain a plurality of optimization subtasks.

[0044] According to the target analysis method, the structural optimization task is decomposed into multiple optimization subtasks, which simplifies the complex tasks and reduces the difficulty of solving them.

[0045] In some embodiments, the target-based analysis method decomposes the structural optimization task to obtain multiple optimization subtasks, including: when the target analysis method is the finite element analysis method, converting the topological structure of the material to be optimized into a finite element model; meshing the finite element model to obtain multiple mesh units; extracting the unit design domain and unit design variables that match the unit structure parameters of each mesh unit from the design domain and the design variables of the optimization configuration parameters; constructing the unit constraint conditions and the unit objective function of the mesh unit according to the unit boundary properties and the unit structure parameters of the mesh unit and the constraint conditions and the objective function; taking each mesh unit as an optimization sub-object, and taking the unit design domain, the unit design variables, the unit constraint conditions and the unit objective function of the mesh unit as the unit optimization configuration parameters of the mesh unit, to obtain multiple optimization subtasks containing the optimization sub-objects and the unit optimization configuration parameters.

[0046] When the target analysis method is finite element analysis, the topological structure of the material to be optimized is first converted into a finite element model, which has the load and boundary conditions of the topological structure of the material to be optimized. Then, the finite element model is divided into multiple grid units through meshing, and the complex structure is discretized into finite element grids to realize the decomposition of the optimization object in the structural optimization task.

[0047] The unit structure parameters are manufacturing parameters contained in the grid unit. The unit structure parameters of the grid unit can be combined to extract information containing these manufacturing parameters from the design domain and design variables of the optimization configuration parameters as the unit design domain and the unit design variables.

[0048] The cell boundary attribute is the cell boundary quantity of the mesh cell, such as the position of the mesh cell in the entire material topology. According to the cell boundary attributes and cell structure parameters of the mesh cell itself and the constraints and objective functions in the optimization configuration information, the cell constraints and cell objective functions corresponding to each mesh cell during optimization are constructed respectively.

[0049] In some embodiments, an interpolation method is used to obtain the unit optimization configuration parameters of each grid unit, thereby achieving decomposition of the optimization configuration information.

[0050] When the target analysis method is the finite element analysis method, multiple grid units and the unit optimization configuration parameters of each grid unit are obtained through decomposition, that is, multiple optimization subtasks including optimization sub-objects and unit optimization configuration parameters are obtained, so as to realize the refined configuration of local parameters, reduce the difficulty of task processing, and facilitate the analysis of each finite element.

[0051] In some embodiments, the user specifies the target analysis method as finite element analysis and sets the number of grids, and the finite element model is divided according to the number of grids.

[0052] In some embodiments, in order to ensure efficient solution of the local optimal solution corresponding to each grid unit, it is necessary to make each grid unit have approximately the same amount of calculation when dividing the grid.

[0053] In some embodiments, the grid units are connected by boundaries, the boundaries are the dividing lines of the grid units, the boundary nodes are the nodes located at the dividing lines, i.e., data, and the boundary nodes are shared by at least two grid units. The grid division is realized by Metis data processing software, the number of boundary nodes is reduced as much as possible, the communication complexity and overhead are reduced, and the solution efficiency is improved.

[0054] In some embodiments, the target-based analysis method decomposes the structural optimization task to obtain multiple optimization sub-tasks, including: when the target analysis method is a neural network processing method, the material topology structure to be optimized and the optimization configuration parameters are input into a first neural network model to obtain a second neural network model; the optimizable parameter set of each neuron in the second neural network model is used as the optimization sub-task to obtain multiple optimization sub-tasks.

[0055] In the case where the target analysis method is a neural network processing method, the topological structure of the material to be optimized and the optimization configuration parameters are input into the first neural network model, and the second neural network model is obtained accordingly, and then the optimizable parameter set of each neuron in the second neural network model is used as an optimization subtask to obtain a local optimization task at the neuron level, that is, to obtain a plurality of the optimization subtasks. Among them, the optimizable parameter set is a weight matrix, a bias term, an activation function, etc.

[0056] Task decomposition is achieved through a neural network model, which converts the high-dimensional material topology to be optimized into multiple low-dimensional optimization subtasks, reducing the difficulty of solving the task.

[0057] Step 103, assigning the optimization subtasks to multiple target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks.

[0058] After obtaining multiple optimization subtasks, the optimization subtasks are assigned to multiple target neural network processors. The optimization subtasks are solved by multiple target neural network processors to achieve parallel solving and improve solving efficiency. The local optimal solutions of multiple optimization subtasks can be quickly obtained, further improving the optimization efficiency.

[0059] In some embodiments, allocating the optimization subtasks to multiple target neural network processors includes: determining the load states of multiple neural network processors and the processing resource requirements of the multiple optimization subtasks; and, based on the load states and the processing resource requirements, allocating the target neural network processor whose load state meets the processing resource requirements to each optimization subtask from the multiple neural network processors.

[0060] Determine the load status of multiple neural network processors (Neural Processing Unit, NPU), and the processing resource requirements of multiple optimization subtasks when solving the task, so as to determine the target neural network processor from the multiple neural network processors according to the load status and the processing resource requirements, so as to achieve the purpose of matching the optimization subtask to the target neural network processor for solving it.

[0061] In some embodiments, optimization subtasks can be sequentially allocated to multiple target neural network processors through a polling allocation method to achieve efficient utilization of solution resources.

[0062] During the process of solving the optimization subtasks, if it is monitored that some target neural network processors are in an idle state for solving, the optimization subtasks can be flexibly scheduled to allocate optimization subtasks to the target neural network processors in the idle state for solving, thereby improving the solving efficiency.

[0063] Taking the finite element analysis method as an example, the stiffness matrix or mass matrix of each mesh element is solved.

[0064] When solving the problem, the solution code required by the neural network processor for solving the task, such as the finite element analysis code, is scheduled according to the objective function.

[0065] In order to ensure the stability of the solution, suppress the influence of noise, and obtain a smoother solution result, the sensitivity of the design variables is analyzed, and the solution result is smoothed by a processing method such as a motion range-based or weighted average method. Taking the finite element analysis method as an example, the sensitivity of each grid cell and the sensitivity of its neighboring grid cells need to be considered to improve the stability of the solution. The retrieval of neighboring grid cells can be performed in parallel on the neural network processor through matrix operations, so that the retrieval can be completed quickly and accurately.

[0066] Among them, sensitivity analysis is to calculate the gradient of each design variable under the objective function and constraint conditions, so as to facilitate numerical adjustment during iteration.

[0067] In some embodiments, gradient calculation is performed through the Compute Unified Device Architecture (CUDA) to achieve efficient processing of large amounts of parallel data.

[0068] When solving each optimization subtask, iterative solutions are performed based on the parameters after sensitivity analysis, and the convergence is judged. The solution that meets the convergence conditions of the optimization subtask is taken as the local optimal solution of the optimization subtask. If it does not meet the conditions, the iteration is continued, and the current solution result can be used as the initial parameter of the next iteration to improve the iteration efficiency.

[0069] In some embodiments, in order to improve the solution efficiency, reduce the solution difficulty, and obtain a converged local optimal solution, each of the target neural network processors is controlled to perform linear analysis processing on the assigned optimization subtask; wherein, when the optimization subtask is a nonlinear task, the target neural network processor is controlled to convert the nonlinear task into a linear task and then perform the linear analysis processing on the optimization subtask.

[0070] In some embodiments, a nonlinear task is converted into a linear task using a moving asymptotes method (MMA), a convex envelope method, a penalty function linearization method, or the like.

[0071] like Figure 2 As shown, Figure 2 This is a task linear conversion flow chart based on MMA provided in an embodiment of the present application. When MMA is used to implement linear conversion, an approximate linear equation is first constructed based on a nonlinear task, and then the variable increment of the approximate linear equation is solved, and it is determined whether the variable increment converges in the approximate linear equation. If not, the variable increment of the approximate linear equation is solved iteratively until the variable increment that converges in the approximate linear equation is obtained. The variables in the nonlinear task are adjusted to the converged variable increment, and it is further determined whether the converged variable increment converges in the nonlinear task. If it converges, the nonlinear task containing the converged variable increment is used as the converted linear task. If not, the approximate linear equation is adjusted until the converged variable increment converges in the nonlinear task. MMA continuously promotes the convergence of the variable increment by constructing an approximate linear equation and iteratively adjusting it, and obtains the corresponding linear task, thereby reducing the difficulty of solving the task.

[0072] When performing linear conversion, parallel processing can also be achieved through a neural network processor to speed up processing efficiency.

[0073] In some embodiments, the central processing unit (CPU) is used to implement data input and output (such as output of material topological structure, output of iterative data, output of physical field quantity, etc.), determination of structural optimization tasks (such as parsing optimization demand information, clarifying design domain and non-design domain), determination of target analysis method, determination of solution method (such as determining whether to use analytical method or numerical method when solving according to the processing resource requirements of the task and the number of grid units), decomposition of structural optimization tasks, allocation of optimization subtasks, and adjustment of manufacturing parameters of the topological structure to be optimized. Solution operation and convergence judgment are completed by the neural network processor.

[0074] Among them, the non-design domain participates in solving the optimal solution, and the design domain participates in solving the optimal solution, sensitivity analysis, manufacturing parameter adjustment, etc.

[0075] Heterogeneous parallel processing is achieved through the central processing unit and multiple target neural network processors. Heterogeneous parallel processing combines different types of processors in the hardware architecture, using the strengths of different types of processors to effectively handle complex tasks, rationally planning the functions of each processor, avoiding resource competition between the central processing unit and the neural network processor during processing, and improving processing performance and efficiency. In structural optimization, heterogeneous parallel processing can improve the accuracy and efficiency of sensitivity analysis, improve the efficiency of solving optimization subtasks, reduce optimization time, and improve optimization accuracy.

[0076] The central processing unit and multiple neural network processors have good scalability. Through the combination of the two, they can adapt to different hardware environments and optimization requirements.

[0077] In some embodiments, the control logic and corresponding data interface of the central processing unit are constructed using programming languages ​​such as C++, and task scheduling and management are completed through the Message Passing Interface (MPI) and Open Multi-Processing (OpenMP) methods.

[0078] Neural network processors can efficiently handle matrix operations, convolution calculations and other computationally intensive tasks. They are more advantageous in parallel solutions, which can significantly shorten the task solution time and reduce the solution cost.

[0079] In some embodiments, parallel programming processing of a neural network processor is implemented through Mindspore, and parallel computing is implemented on multiple neural network processors, so as to utilize the parallel processing capabilities of the neural network processor and accelerate the solution process.

[0080] In some embodiments, an intermediate layer is set between the central processing unit and multiple neural network processors, and the central processing unit monitors the load status of the neural network processor through the intermediate layer to optimize the flexible allocation of subtasks.

[0081] In some embodiments, shared memory is set to store shared data, communication data, etc. of the optimization subtasks, for example, to store boundary node data and NVIDIA Collective Communication Library (NCCL) data, so as to avoid the impact of fragmented storage on optimization performance and improve the data access efficiency and data storage efficiency of the neural network processor.

[0082] In some embodiments, during the solution process of the finite element analysis method, data is stored in the form of an array or a sparse matrix.

[0083] Step 104, obtaining a global optimal solution corresponding to the structural optimization task based on the local optimal solution, wherein the global optimal solution includes a plurality of second manufacturing parameters, and the second manufacturing parameters are used to compare with the structural manufacturing constraint parameters to determine the target manufacturing parameters at each target position.

[0084] The target manufacturing parameter is the second manufacturing parameter or the structural manufacturing constraint parameter.

[0085] After local parallel solving, we focus on the global task, i.e. the structural optimization task, and further obtain the global optimal solution corresponding to the structural optimization task through the local optimal solution, and further obtain the target manufacturing parameters for achieving optimization adjustment through comparison. While optimizing, we ensure the feasibility of the optimization scheme, avoid re-solving due to the inability to implement the optimization scheme, and improve the optimization efficiency.

[0086] In some embodiments, multiple local optimal solutions are taken as solution results to determine whether the structural optimization task meets its convergence conditions under this set of local optimal solutions. If so, this set of local optimal solutions is taken as the global optimal solution. If not, fine-tuning is performed based on the local optimal solutions to obtain the global optimal solution corresponding to the structural optimization task.

[0087] Taking the finite element analysis method as an example, the unit matrix is ​​assembled into the global matrix, and boundary conditions are added to obtain the global optimal solution by solving the linear equations.

[0088] The global optimal solution contains multiple second manufacturing parameters. The second manufacturing parameters at the same position are compared with the structural manufacturing constraint parameters. If the second manufacturing parameter belongs to the structural manufacturing constraint parameter, the second manufacturing parameter is determined as the target manufacturing parameter. If the second manufacturing parameter does not belong to the structural manufacturing constraint parameter, the parameter closest to the second manufacturing parameter in the structural manufacturing constraint parameter is determined as the target manufacturing parameter, thereby determining the target manufacturing parameter at each target position. Combine the optimization with the actual production and manufacturing scenario to ensure the manufacturability of the material topological structure. The comparison and determination process can be completed in parallel in the neural network processor to improve the efficiency of the comparison and determination.

[0089] Step 105 , adjusting the first manufacturing parameter at each target position in the material topology structure to be optimized to the target manufacturing parameter corresponding to the target position, to obtain an optimized material topology structure.

[0090] The first manufacturing parameter at each target position in the material topology to be optimized is adjusted to the target manufacturing parameter corresponding to the target position, so as to realize the transformation of the solution data to the physical structure and obtain the optimized material topology, which is convenient for guiding users in production and manufacturing.

[0091] The central processing unit performs post-processing on the optimized material topology structure and outputs a visualized optimized material topology structure for users to review.

[0092] In some embodiments, an analysis report including manufacturing parameters before and after adjustment, such as stress distribution, mode, etc., may also be output together so that the user can understand the final optimization effect.

[0093] In an embodiment of the present application, a target analysis method is used to decompose the determined structural optimization task into multiple optimization subtasks. After the task is decomposed, the solution complexity of each optimization subtask is greatly reduced. Afterwards, multiple target neural network processors are used to solve the optimization subtasks respectively to obtain the local optimal solution corresponding to each optimization subtask. This step is solved in parallel by the target neural network processor to improve the solution efficiency. The global optimal solution corresponding to the structural optimization task is then determined by the local optimal solution, and the second manufacturing parameter in the global optimal solution is compared with the corresponding structural manufacturing constraint parameter to determine the target manufacturing parameter, and the first manufacturing parameter in the material topology to be optimized is adjusted to the target manufacturing parameter to achieve the adjustment of the material topology and obtain the optimized material topology. The reduction in solution complexity and the improvement in solution efficiency jointly improve the optimization efficiency of the material topology.

[0094] See also Figure 3 , Figure 3 This is a structural diagram of a material topological structure adjustment system provided in an embodiment of the present application. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0095] The material topology adjustment system 300 includes: a determination module 301 , a decomposition module 302 , a distribution solution module 303 , a solution module 304 , and an adjustment module 305 .

[0096] The determination module 301 is used to determine the structural optimization task based on the acquired material topology structure to be optimized and the optimization requirement information; the material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter.

[0097] The decomposition module 302 is used to decompose the structural optimization task based on the target analysis method to obtain multiple optimization subtasks.

[0098] The allocation and solving module 303 is used to allocate the optimization subtasks to multiple target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks.

[0099] The solution module 304 is used to obtain a global optimal solution corresponding to the structural optimization task based on the local optimal solution, wherein the global optimal solution includes multiple second manufacturing parameters, and the second manufacturing parameters are used to compare with the structural manufacturing constraint parameters to determine the target manufacturing parameters at each of the target positions; the target manufacturing parameters are the second manufacturing parameters or the structural manufacturing constraint parameters.

[0100] The adjustment module 305 is used to adjust the first manufacturing parameter at each target position in the material topology structure to be optimized to the target manufacturing parameter corresponding to the target position, so as to obtain an optimized material topology structure.

[0101] In some embodiments, the determining module is specifically used to: Extracting optimization configuration parameters from the optimization requirement information; the optimization configuration parameters include design domain, design variables, objective function and constraint conditions; The material topology structure to be optimized is taken as the optimization object, and the optimization task including the optimization object and the optimization configuration parameters is obtained.

[0102] In some embodiments, the decomposition module is used to: In the case where the target analysis method is a finite element analysis method, converting the topological structure of the material to be optimized into a finite element model; Meshing the finite element model to obtain a plurality of mesh units; Extracting a unit design domain and a unit design variable that match the unit structure parameters of each of the grid units from the design domain and the design variables of the optimization configuration parameters respectively; According to the unit boundary attributes of the grid unit, the unit structure parameters, the constraint conditions and the objective function, respectively construct the unit constraint conditions and the unit objective function of the grid unit; Each of the grid cells is taken as an optimization sub-object, and the cell design domain, the cell design variables, the cell constraints and the cell objective function of the grid cell are taken as cell optimization configuration parameters of the grid cell, so as to obtain a plurality of the optimization sub-tasks including the optimization sub-objects and the cell optimization configuration parameters.

[0103] In some embodiments, the decomposition module is further configured to: In the case where the target analysis method is a neural network processing method, the topological structure of the material to be optimized and the optimization configuration parameters are input into a first neural network model to obtain a second neural network model; The optimizable parameter set of each neuron in the second neural network model is used as the optimization subtask to obtain a plurality of the optimization subtasks.

[0104] In some embodiments, the system further comprises a target analysis method determination module, which is used to: In the case where the task scenario complexity of the structural optimization task is greater than or equal to the set scenario complexity, the neural network processing method is determined as the target analysis method; or, In the case where the task scenario complexity of the structural optimization task is less than the set scenario complexity, the finite element analysis method is determined as the target analysis method; or, In the case where a configuration analysis method is configured in the optimization requirement configuration information, the configuration analysis method is determined as the target analysis method, and the configuration analysis method is the neural network processing method or the finite element analysis method.

[0105] In some embodiments, the allocation module is specifically used to: Determining the load status of a plurality of neural network processors and the processing resource requirements of a plurality of the optimization subtasks; According to the load status and the processing resource requirement, the target neural network processor whose load status meets the processing resource requirement is allocated to each of the optimization subtasks from the plurality of neural network processors.

[0106] In some embodiments, the system further comprises a linear analysis processing module for: Controlling each of the target neural network processors to perform linear analysis processing on the assigned optimization subtask; Wherein, when the optimization subtask is a nonlinear task, the target neural network processor is controlled to convert the nonlinear task into a linear task and then perform the linear analysis processing on the optimization subtask.

[0107] The material topology adjustment system provided in the embodiment of the present application can implement each process of the embodiment of the above-mentioned material topology adjustment method and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0108] Figure 4 4 is a structural diagram of an electronic device provided in an embodiment of the present application. As shown in the figure, the electronic device 4 of the embodiment includes: at least one processor 40 ( Figure 4 Only one is shown in the figure), a memory 41, and a computer program 42 stored in the memory 41 and executable on the at least one processor 40, wherein the processor 40 implements the steps of any of the above-mentioned method embodiments when executing the computer program 42.

[0109] The electronic device 4 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will appreciate that Figure 4 It is only an example of the electronic device 4 and does not constitute a limitation of the electronic device 4. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0110] The processor 40 may be a CPU, or other general-purpose processors, 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, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0111] The memory 41 may be an internal storage unit of the electronic device 4, such as a hard disk or memory of the electronic device 4. The memory 41 may also be an external storage device of the electronic device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 4. Further, the memory 41 may also include both an internal storage unit and an external storage device of the electronic device 4. The memory 41 is used to store the computer program and other programs and data required by the electronic device. The memory 41 may also be used to temporarily store data that has been output or is to be output.

[0112] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0113] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0114] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0115] In the embodiments provided in the present application, it should be understood that the disclosed systems / electronic devices and methods can be implemented in other ways. For example, the system / electronic device embodiments described above are merely schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the system or unit can be electrical, mechanical or other forms.

[0116] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0118] If the integrated module / unit is implemented in the form of 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 present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0119] The present application implements all or part of the processes in the above-mentioned embodiment methods, and may also be implemented through a computer program product. When the computer program product runs on an electronic device, the electronic device can implement the steps in the above-mentioned method embodiments when the computer program product is executed.

[0120] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for adjusting the topological structure of a material, characterized in that: include: Determine the structural optimization task based on the obtained topological structure of the material to be optimized and the optimization requirement information; The material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter; Based on the target analysis method, the structural optimization task is decomposed to obtain multiple optimization subtasks; Allocating the optimization subtasks to a plurality of target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks; Based on the local optimal solution, a global optimal solution corresponding to the structural optimization task is obtained, wherein the global optimal solution includes a plurality of second manufacturing parameters, and the second manufacturing parameters are used to compare with the structural manufacturing constraint parameters to determine the target manufacturing parameters at each target position; the target manufacturing parameters are the second manufacturing parameters or the structural manufacturing constraint parameters; The first manufacturing parameter at each target position in the material topological structure to be optimized is adjusted to the target manufacturing parameter corresponding to the target position to obtain an optimized material topological structure.

2. The method according to claim 1, characterized in that The step of determining the structural optimization task based on the obtained topological structure of the material to be optimized and the optimization requirement information includes: Extracting optimization configuration parameters from the optimization requirement information; the optimization configuration parameters include design domain, design variables, objective function and constraint conditions; The material topology structure to be optimized is taken as the optimization object, and the optimization task including the optimization object and the optimization configuration parameters is obtained.

3. The method according to claim 2, characterized in that The target-based analysis method decomposes the structural optimization task to obtain multiple optimization subtasks, including: In the case where the target analysis method is a finite element analysis method, converting the topological structure of the material to be optimized into a finite element model; Meshing the finite element model to obtain a plurality of mesh units; Extracting a unit design domain and a unit design variable that match the unit structure parameters of each of the grid units from the design domain and the design variables of the optimization configuration parameters respectively; According to the unit boundary attributes of the grid unit, the unit structure parameters, the constraint conditions and the objective function, respectively construct the unit constraint conditions and the unit objective function of the grid unit; Each of the grid cells is taken as an optimization sub-object, and the cell design domain, the cell design variables, the cell constraints and the cell objective function of the grid cell are taken as cell optimization configuration parameters of the grid cell, so as to obtain a plurality of the optimization sub-tasks including the optimization sub-objects and the cell optimization configuration parameters.

4. The method according to claim 2, characterized in that: The target-based analysis method decomposes the structural optimization task to obtain multiple optimization subtasks, including: In the case where the target analysis method is a neural network processing method, the topological structure of the material to be optimized and the optimization configuration parameters are input into a first neural network model to obtain a second neural network model; The optimizable parameter set of each neuron in the second neural network model is used as the optimization subtask to obtain a plurality of the optimization subtasks.

5. The method according to claim 1, characterized in that The method further comprises: In the case where the task scenario complexity of the structural optimization task is greater than or equal to the set scenario complexity, the neural network processing method is determined as the target analysis method; or, In the case where the task scenario complexity of the structural optimization task is less than the set scenario complexity, the finite element analysis method is determined as the target analysis method; or, In the case where a configuration analysis method is configured in the optimization requirement configuration information, the configuration analysis method is determined as the target analysis method, and the configuration analysis method is the neural network processing method or the finite element analysis method.

6. The method according to claim 1, characterized in that The step of allocating the optimization subtasks to a plurality of target neural network processors comprises: Determining the load status of a plurality of neural network processors and the processing resource requirements of a plurality of the optimization subtasks; According to the load status and the processing resource requirement, the target neural network processor whose load status meets the processing resource requirement is allocated to each of the optimization subtasks from the plurality of neural network processors.

7. The method according to claim 1, characterized in that The method further comprises: Controlling each of the target neural network processors to perform linear analysis processing on the assigned optimization subtask; Wherein, when the optimization subtask is a nonlinear task, the target neural network processor is controlled to convert the nonlinear task into a linear task and then perform the linear analysis processing on the optimization subtask.

8. A system for adjusting the topological structure of a material, characterized in that: include: A determination module is used to determine the structural optimization task based on the obtained topological structure of the material to be optimized and the optimization requirement information; The material topology structure to be optimized corresponds to a structural manufacturing constraint parameter, and each target position of the material topology structure to be optimized corresponds to a first manufacturing parameter; A decomposition module, used for decomposing the structural optimization task based on a target analysis method to obtain a plurality of optimization subtasks; An allocation and solving module is used to allocate the optimization subtasks to a plurality of target neural network processors, so that each of the target neural network processors solves at least one of the optimization subtasks to obtain a local optimal solution corresponding to each of the optimization subtasks; A solution module, used for obtaining a global optimal solution corresponding to the structural optimization task based on a local optimal solution, wherein the global optimal solution includes a plurality of second manufacturing parameters, and the second manufacturing parameters are used for comparing with the structural manufacturing constraint parameters to determine a target manufacturing parameter at each of the target positions; the target manufacturing parameter is the second manufacturing parameter or the structural manufacturing constraint parameter; An adjustment module is used to adjust the first manufacturing parameter at each target position in the material topology structure to be optimized to the target manufacturing parameter corresponding to the target position, so as to obtain an optimized material topology structure.

9. An electronic device, characterized in that: The electronic device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the electronic device implements the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The invention comprises a computer program which, when executed, causes the method according to any one of claims 1 to 7 to be performed.

Images