Industrial equipment topology optimization method and system based on deep learning

Abstract

The invention provides an industrial equipment topology optimization method and system based on deep learning, and the method comprises the steps: building a to-be-optimized equipment model based on a finite element modeling algorithm, and the model equipment at least comprises model geometric parameters and load parameters of the model in a design domain; setting a displacement boundary condition for model optimization; constructing a convolutional neural network model, wherein the convolutional neural network model is used for processing the input multi-channel tensor to obtain a density distribution optimization result of the equipment model; based on a variable density method, according to model geometric parameters and load parameters of the equipment model and displacement boundary conditions, determining a multi-channel tensor input into the convolutional neural network; the multi-channel tensor is input into the trained convolutional neural network model, target density distribution of the equipment model is obtained, a topological optimization result of the equipment model is obtained in combination with the density parameters in the material set, and the method provided by the invention can improve the efficiency of topological optimization of the industrial equipment structure.

Description

Technical Field

[0001] The present invention belongs to the technical field of product design, and specifically relates to a method and system for industrial equipment topology optimization based on deep learning. Background Art

[0002] Equipment product design is an important task in modern engineering, and its performance and effect directly affect the quality, efficiency and safety of industrial production. However, traditional equipment product design methods are often limited by experience and trial and error, and have problems such as long time consumption, waste of resources and high cost. In order to solve these problems and promote innovation in equipment design, topology optimization design has received widespread attention and application as an advanced design method.

[0003] Topological optimization is an optimization design method for equipment manufacturing products, combining structural mechanics and optimization methods. By optimizing the spatial distribution of materials, structural configuration and component size, the optimal structural form can be obtained from a variety of structural design schemes to achieve weight reduction, cost reduction and performance improvement of aviation, automotive and other products.

[0004] The important role of topology optimization software in product design is mainly reflected in the improvement of product performance, design efficiency and the acceleration of product release. Therefore, developing topology optimization software for specific scenarios can not only help engineers quickly master its use, but also better promote the development of related industries.

[0005] Since topology optimization software is generally based on the finite element method, the computational efficiency of topology optimization is often difficult to meet the needs of actual engineering.

[0006] How to improve the efficiency of topology optimization in industrial equipment product design is a technical problem that needs to be solved urgently. Summary of the invention

[0007] In view of the above-mentioned problems in the prior art, the object of the present invention is to provide an industrial equipment topology optimization method and system based on deep learning, which can improve the topology optimization efficiency of industrial equipment in product design.

[0008] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0009] In one aspect, the present invention provides an industrial equipment topology optimization method based on deep learning, the method comprising:

[0010] Based on the finite element modeling algorithm, a device model to be optimized is constructed, wherein the model device at least includes model geometry parameters and load parameters of the model in the design domain;

[0011] Set displacement boundary conditions for model optimization;

[0012] Constructing a convolutional neural network model, the convolutional neural network model is used to process the input multi-channel tensor to obtain the density distribution optimization result of the device model, the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input the multi-channel tensor, the multiple convolution modules are connected in sequence, the current convolution module is used to extract features according to the output result of the previous convolution module to obtain the feature data of the multi-channel tensor in sequence, the multiple transposed convolution modules are connected in sequence and correspond to the multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module;

[0013] Based on the variable density method, according to the model geometry parameters and load parameters of the device model and the displacement boundary conditions, a multi-channel tensor input to the convolutional neural network is determined, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction;

[0014] Inputting the multi-channel tensor into a trained convolutional neural network model to obtain a target density distribution of the device model;

[0015] According to the target density distribution of the device model and in combination with the density parameters in the material set, a topology optimization result of the device model is obtained.

[0016] The convolution module includes a convolution layer, an activation function, and a maximum pooling layer; the input layer is connected to multiple convolution layers, each convolution layer is sequentially connected to an activation function and a maximum pooling layer, the convolution layer is used to extract the feature map of the input multi-channel tensor, and the activation function is used to perform nonlinear operations on the multi-channel tensor after convolution processing;

[0017] The transposed convolution module includes a transposed convolution layer, a Dropout layer and a splicing layer. The transposed convolution layer is used to upsample to increase the size of the feature map, the Dropout layer is used to randomly discard neurons through the Dropout mechanism, and the splicing layer is used to splice the features extracted from different layers together and send the splicing results to the output layer.

[0018] Furthermore, the variable density method is expressed by the following formula:

[0019]

[0020] F = K·U;

[0021]

[0022] 0 <Xmin≤x≤1;

[0023] Among them, U is the global displacement vector, F is the force vector, K is the global stiffness matrix, and u e and k 0 are the unit displacement vector and stiffness matrix respectively, x is the design variable vector, xmi n is the minimum relative density vector, N is the number of elements used to discretize the design domain, P is the penalty factor, V(x) and V0 are the material volume and design domain volume respectively, and f is the set volume fraction.

[0024] Furthermore, the multi-channel tensor is obtained by the following steps:

[0025] Based on the variable density method and the initial design variable vector of the material in the equipment model, the displacement vector of the model in the horizontal direction and the displacement vector in the vertical direction are determined;

[0026] Respectively converting the displacement vector in the horizontal direction and the displacement vector in the vertical direction into a displacement matrix in the horizontal direction and a displacement matrix in the vertical direction of the same size;

[0027] Extract the initial volume fraction of the design domain and expand it into an initial volume fraction matrix of the same size as the displacement matrix in the horizontal direction and the displacement matrix in the vertical direction;

[0028] The displacement matrix in the horizontal direction, the displacement matrix in the vertical direction and the initial volume fraction matrix are combined into the multi-channel tensor.

[0029] Furthermore, the method further comprises:

[0030] Normalizing the target density distribution of the device model;

[0031] The normalized device model is visualized according to the pre-set identifiers of different density values, wherein the identifiers include colors.

[0032] Furthermore, the convolutional neural network model is trained by the following steps:

[0033] Build an initial convolutional neural network model;

[0034] Acquire training data, wherein the training data includes multiple groups of optimized equipment models, and density distribution results and optimization parameters corresponding to each equipment module, wherein the optimization parameters include at least model geometry parameters, load parameters, and displacement boundary conditions;

[0035] Processing the optimization parameters to obtain a multi-channel tensor;

[0036] Taking the multi-channel tensor as input data and the density distribution result of the device model as the target output, the initial convolutional neural network model is trained using a preset loss function until the convergence condition is met, thereby obtaining a trained convolutional neural network model.

[0037] Furthermore, the convolutional neural network model is a U-net network model.

[0038] On the other hand, the present invention also provides an industrial equipment topology optimization system based on deep learning, the system comprising:

[0039] A device model building module, used to build a device model to be optimized based on a finite element modeling algorithm, wherein the model device at least includes model geometry parameters and load parameters of the model in the design domain;

[0040] Condition setting module, used to set displacement boundary conditions for model optimization;

[0041] A convolutional neural network model construction module is used to construct a convolutional neural network model, wherein the convolutional neural network model is used to process an input multi-channel tensor to obtain a density distribution optimization result of a device model, and the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input a multi-channel tensor, and multiple convolution modules are connected in sequence, and the current convolution module is used to extract features according to the output result of the previous convolution module to obtain feature data of the multi-channel tensor in sequence, and multiple transposed convolution modules are connected in sequence and correspond to multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module;

[0042] An input module, for determining a multi-channel tensor input to a convolutional neural network based on a variable density method according to model geometry parameters and load parameters of the device model and the displacement boundary conditions, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction;

[0043] A target density distribution calculation module, used for inputting the multi-channel tensor into the trained convolutional neural network model to obtain the target density distribution of the device model;

[0044] The determination module is used to obtain a topology optimization result of the device model according to the target density distribution of the device model and the density parameters in the material set.

[0045] On the other hand, the present invention also provides an industrial equipment topology optimization system, the system comprising:

[0046] A modeling module, for constructing a geometric model of the industrial equipment to be optimized according to a finite element modeling algorithm, including at least the following controls for user input: geometric parameters, load parameters, and displacement boundary conditions;

[0047] An optimization module, used to perform topology optimization processing according to the industrial equipment topology optimization method based on deep learning described above, to obtain a target density distribution of the industrial equipment to be optimized;

[0048] A visualization module is used to visualize the industrial equipment according to the target density distribution in combination with a preset identification matching rule.

[0049] On the other hand, the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the industrial equipment topology optimization method based on deep learning as described above.

[0050] By adopting the above technical scheme, the present invention discloses a method for topology optimization of industrial equipment based on deep learning, the method comprising: constructing a device model to be optimized based on a finite element modeling algorithm, the model device at least comprising model geometric parameters and load parameters of the model in the design domain; setting displacement boundary conditions for model optimization; constructing a convolutional neural network model, the convolutional neural network model being used to process the input multi-channel tensor to obtain a density distribution optimization result of the device model, the convolutional neural network model comprising an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer being used to input a multi-channel tensor, the multiple convolution modules being connected in sequence, the current convolution module being used to perform feature extraction according to the output result of the previous convolution module to obtain feature data of the multi-channel tensor in sequence, the multiple transposed convolution modules being connected in sequence and corresponding one-to-one to the multiple convolution modules, wherein the current transposed convolution module The block performs splicing processing on the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the equipment model according to the processing result of the transposed convolution module; based on the variable density method, according to the model geometric parameters and load parameters of the equipment model, and the displacement boundary conditions, the multi-channel tensor input to the convolutional neural network is determined, and the multi-channel tensor at least includes a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction; the multi-channel tensor is input into the trained convolutional neural network model to obtain the target density distribution of the equipment model; according to the target density distribution of the equipment model, combined with the density parameters in the material set, the topology optimization result of the equipment model is obtained. The method provided by the present invention can improve the efficiency of topology optimization of industrial equipment structures.

[0051] In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A schematic diagram of the steps of a method for topology optimization of industrial equipment based on deep learning provided by an embodiment of the present invention is shown;

[0054] Figure 2A schematic diagram of the network structure of a convolutional neural network model in an embodiment of this specification is shown;

[0055] Figure 3 A schematic diagram of the visualization effect in the embodiment of this specification is shown;

[0056] Figure 4 A schematic diagram of the structure of an industrial equipment topology optimization system based on deep learning provided in an embodiment of this specification is shown;

[0057] Figure 5 A schematic diagram of the structure of an industrial equipment topology optimization system based on deep learning provided in an embodiment of this specification is shown;

[0058] Figure 6 The working process of the built-in software of the industrial equipment topology optimization system in the embodiment of this specification is shown;

[0059] Figure 7 A schematic diagram of a finite element modeling operation page in an embodiment of this specification is shown;

[0060] Figure 8 A schematic diagram of a topology optimization parameter input operation page in an embodiment of this specification is shown;

[0061] Fig. 9 A schematic diagram of the structure of the system provided in the embodiment of this specification is shown.

[0062] Description of the accompanying symbols:

[0063] 410. Equipment model building module; 420. Condition setting module; 430. Convolutional neural network model building module; 440. Input module; 450. Target density distribution calculation module; 460. Determination module; 510. Modeling module; 520. Optimization module; 530. Visualization module. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0066] In the prior art, topology optimization is an optimization design method for equipment manufacturing products, combining structural mechanics and optimization methods. By optimizing the spatial distribution of materials, structural configuration and component size, the optimal structural form can be obtained from a variety of structural design schemes to achieve weight reduction, cost reduction and performance improvement of aviation, automotive and other products.

[0067] The important role of topology optimization software in product design is mainly reflected in the improvement of product performance, design efficiency and the acceleration of product release. Therefore, developing topology optimization software for specific scenarios can not only help engineers quickly master its use, but also better promote the development of related industries.

[0068] Since topology optimization software is generally based solely on finite element methods, the computational efficiency of topology optimization is often difficult to meet the needs of actual engineering.

[0069] In order to solve the above problems, an embodiment of the present invention provides an industrial equipment topology optimization method based on deep learning, which can improve the topology optimization efficiency of industrial equipment in product design. Figure 1 It is a schematic diagram of the steps of a method for topology optimization of industrial equipment based on deep learning provided by an embodiment of the present invention. This specification provides the method operation steps described in the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or device product is executed, it can be executed in the order of the method shown in the embodiment or the accompanying drawings or in parallel. Specifically, Figure 1 As shown, the method may include:

[0070] S101: constructing a device model to be optimized based on a finite element modeling algorithm, wherein the model device at least includes model geometry parameters and load parameters of the model in a design domain;

[0071] S102: Setting displacement boundary conditions for model optimization;

[0072] S103: Construct a convolutional neural network model, the convolutional neural network model is used to process the input multi-channel tensor to obtain the density distribution optimization result of the device model, the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input the multi-channel tensor, the multiple convolution modules are connected in sequence, the current convolution module is used to extract features according to the output result of the previous convolution module to obtain the feature data of the multi-channel tensor in sequence, the multiple transposed convolution modules are connected in sequence and correspond to the multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module;

[0073] S104: Based on a variable density method, according to model geometry parameters and load parameters of the device model and the displacement boundary conditions, determine a multi-channel tensor input to a convolutional neural network, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction;

[0074] S105: Inputting the multi-channel tensor into the trained convolutional neural network model to obtain the target density distribution of the device model;

[0075] S106: Obtaining a topology optimization result of the device model according to the target density distribution of the device model and in combination with density parameters in a material set.

[0076] It can be understood that: the present invention can obtain the optimal topological structure by utilizing deep learning algorithms, such as convolutional neural network models, to predict and process the topological optimization of industrial equipment in the product design process. Compared with the traditional SIMP method based on isotropic solid materials, the prediction results based on the deep learning model have more confirmed density values ​​and more intuitive display of the topological structure boundaries.

[0077] Furthermore, the convolution module includes a convolution layer, an activation function, and a maximum pooling layer; the input layer is connected to multiple convolution layers, each convolution layer is sequentially connected to an activation function and a maximum pooling layer, the convolution layer is used to extract a feature map of the input multi-channel tensor, and the activation function is used to perform nonlinear operations on the multi-channel tensor after convolution processing;

[0078] The transposed convolution module includes a transposed convolution layer, a Dropout layer and a splicing layer. The transposed convolution layer is used to upsample to increase the size of the feature map, the Dropout layer is used to randomly discard neurons through the Dropout mechanism, and the splicing layer is used to splice the features extracted from different layers together and send the splicing results to the output layer.

[0079] Exemplarily, the deep learning model is based on convolutional neural network training, such as Figure 2 As shown in the figure, the convolutional neural network includes an input layer (imageinput), a convolution layer (conv), an activation function (relu), a pooling layer (maxpoling), a transposed convolution layer (transconv), a dropout layer, a concatenation layer (concatenate), and an output layer (regressionout).

[0080] The input of the neural network is a 33×65×3 tensor, which includes the horizontal displacement matrix, vertical displacement matrix and initial volume of the image.

[0081] After the input layer, multiple convolutional layers are added, and each convolutional layer is followed by an activation function (ReLU) and a maximum pooling layer. The ReLU activation function is used to perform nonlinear operations on the input tensor after convolution processing to avoid losing the gradient of calculation due to the deepening of the model. The maximum pooling layer is used to further extract important features and further reduce the size of the tensor and the amount of calculation.

[0082] In the second half of the network, a transposed convolution layer is used to upsample and increase the size of the feature map, restoring the spatial information of the tensor data after the size reduction in the previous step. In the second half, the Dropout mechanism is used to prevent overfitting and increase the generalization ability of the model by randomly discarding some neurons.

[0083] The final output layer splices the features extracted from different layers together and inputs more feature information into the Sigmoid function and regression function to output the final prediction result. The output of the model gives the optimized structure layout under given conditions in the form of relative density, that is, a 32×64 matrix with element values ​​ranging from 0 to 1, that is, the relative density value.

[0084] In other words, the convolutional neural network model outputs the density distribution of the designed product, combines it with the determined material density, matches the corresponding materials, and finally obtains the optimized solution for the designed product.

[0085] Furthermore, the application of variable density method in topology optimization can flexibly allocate materials in a given design space to achieve the optimization of structural performance. By adjusting the density function, the distribution of materials in the design domain can be controlled, thereby affecting the mechanical properties of the structure.

[0086] The variable density method describes the material distribution by introducing a density variable. The variable density method expresses the correspondence between the unit relative density and the material elastic modulus in the form of a density function of a continuous variable. The density referred to in the variable density method is a pseudo-density, which reflects the correspondence between the material density and the material properties. The value in each design unit is between 0 and 1. The flexible optimization topology optimization problem can be written as:

[0087]

[0088] F = K·U;

[0089]

[0090] 0xXmin≤x≤1;

[0091] Among them, U is the global displacement vector, F is the force vector, K is the global stiffness matrix, and u e and k 0 are the unit displacement vector and stiffness matrix respectively, x is the design variable vector, xmi n is the minimum relative density vector, N is the number of elements used to discretize the design domain, P is the penalty factor, V(x) and V0 are the material volume and design domain volume respectively, and f is the set volume fraction.

[0092] Furthermore, the multi-channel tensor is obtained by the following steps:

[0093] Based on the variable density method and the initial design variable vector of the material in the equipment model, the displacement vector of the model in the horizontal direction and the displacement vector in the vertical direction are determined;

[0094] Respectively converting the displacement vector in the horizontal direction and the displacement vector in the vertical direction into a displacement matrix in the horizontal direction and a displacement matrix in the vertical direction of the same size;

[0095] Extract the initial volume fraction of the design domain and expand it into an initial volume fraction matrix of the same size as the displacement matrix in the horizontal direction and the displacement matrix in the vertical direction;

[0096] The displacement matrix in the horizontal direction, the displacement matrix in the vertical direction and the initial volume fraction matrix are combined into the multi-channel tensor.

[0097] Furthermore, the convolutional neural network model is trained by the following steps:

[0098] Build an initial convolutional neural network model;

[0099] Acquire training data, wherein the training data includes multiple groups of optimized equipment models, and density distribution results and optimization parameters corresponding to each equipment module, wherein the optimization parameters include at least model geometry parameters, load parameters, and displacement boundary conditions;

[0100] Processing the optimization parameters to obtain a multi-channel tensor;

[0101] Taking the multi-channel tensor as input data and the density distribution result of the device model as the target output, the initial convolutional neural network model is trained using a preset loss function until the convergence condition is met, thereby obtaining a trained convolutional neural network model.

[0102] Exemplarily, 6000 samples with different boundary conditions were generated by topology optimization software to establish a data set for training deep learning models. The input of the neural network is a 33×65×3 tensor, which includes the horizontal displacement matrix, the vertical displacement matrix and the initial volume fraction. Therefore, for the input data, U1 and U2 are converted into matrices of size [number of row nodes, number of column nodes], and the initial volume fraction of the design domain is extracted and expanded into a matrix of the same size as U1 and U2, and then the U1, U2, and matrices are merged into three-channel image data of size [number of row nodes, number of column nodes, 3, number of samples].

[0103] The output of the deep learning model gives the optimal structure layout under given conditions in the form of relative density, that is, a 32×64 matrix with element values ​​ranging from 0 to 1. In terms of output data, in order to improve training efficiency and accuracy, the output vectors with values ​​between 0 and 1 are normalized and then converted into image data of [number of unit rows, number of unit columns, 1, number of samples]. In order to effectively train the neural network and accurately evaluate its performance, all samples are divided into training set, validation set and test set in a ratio of 8:1:1.

[0104] During model training, it is necessary to continuously adjust parameters such as learning rate, weight initialization, regularization, batch size, number of iterations, and data enhancement techniques to enhance the performance and generalization ability of the model, improve the convergence speed, avoid overfitting and other problems, and gradually minimize the loss function to improve the accuracy of the model. Compared with the traditional SIMP-based method, the prediction results based on the deep learning model have more confirmed density values ​​and more intuitive topological structure boundaries.

[0105] In one embodiment of this specification, the neural network uses Matlab's own unetLayers code to generate a U-Net network.

[0106] Furthermore, the method further comprises:

[0107] Normalizing the target density distribution of the device model;

[0108] The normalized device model is visualized according to the pre-set identifiers of different density values, wherein the identifiers include colors.

[0109] It can be understood that in order to improve training efficiency and accuracy, the output vectors with values ​​between 0 and 1 are normalized, and different density values ​​are represented by different colors, making the boundary conditions more obvious.

[0110] like Figure 3 The figure is a schematic diagram of the visualization effect in an example of an embodiment of this specification.

[0111] Based on the above-mentioned method, the embodiment of this specification also provides an industrial equipment topology optimization system based on deep learning, such as Figure 4 As shown, the system comprises:

[0112] The equipment model building module 410 is used to build the equipment model to be optimized based on the finite element modeling algorithm, and the model equipment at least includes the model geometry parameters and load parameters of the model in the design domain;

[0113] Condition setting module 420, used to set displacement boundary conditions for model optimization;

[0114] A convolutional neural network model construction module 430 is used to construct a convolutional neural network model, wherein the convolutional neural network model is used to process the input multi-channel tensor to obtain a density distribution optimization result of the device model, and the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input the multi-channel tensor, and the multiple convolution modules are connected in sequence, and the current convolution module is used to extract features according to the output result of the previous convolution module to obtain feature data of the multi-channel tensor in sequence, and the multiple transposed convolution modules are connected in sequence and correspond to the multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module;

[0115] An input module 440 is used to determine a multi-channel tensor input to a convolutional neural network based on a variable density method according to model geometry parameters and load parameters of the device model and the displacement boundary conditions, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction;

[0116] A target density distribution calculation module 450 is used to input the multi-channel tensor into the trained convolutional neural network model to obtain the target density distribution of the device model;

[0117] The determination module 460 is used to obtain a topology optimization result of the device model according to the target density distribution of the device model and the density parameters in the material set.

[0118] The effect obtained by the above-mentioned device is consistent with the effect obtained by the above-mentioned method, and will not be described in detail in the embodiments of this specification.

[0119] In the embodiments of this specification, an industrial equipment topology optimization system is also provided, such as Figure 5 As shown, the system comprises:

[0120] A modeling module 510, for constructing a geometric model of the industrial equipment to be optimized according to a finite element modeling algorithm, including at least the following controls for user input: geometric parameters, load parameters, and displacement boundary conditions;

[0121] An optimization module 520 is used to perform topology optimization processing according to the industrial equipment topology optimization method based on deep learning described above, to obtain a target density distribution of the industrial equipment to be optimized;

[0122] The visualization module 530 is used to visualize the industrial equipment according to the target density distribution and a preset identification matching rule.

[0123] For example, corresponding processing software can be configured inside the topology optimization system to realize the functions of the above modules. The software is divided into three modules: modeling, topology optimization algorithm and software visualization, such as Figure 6 shown.

[0124] The software can quickly and accurately create geometric models of industrial equipment through finite element modeling. Figure 7 As shown, the geometric parameters of the equipment product, the size, direction and location of the load can be specified, and the boundary parameters of the equipment can be passed to the topology optimization algorithm model through the "Apply" button. Displacement boundary conditions (MMB and cantilever beam) can be set through radio buttons and can be intuitively displayed in the image according to the corresponding displacement boundary conditions. With these inputs, the software can generate an accurate finite element model and prepare for the subsequent optimization process.

[0125] After entering the corresponding geometric lengths x and y in the geometry area, the geometric parameters are passed to the topology optimization algorithm model. In the load area, enter the load position coordinates Position_x and Position_y, as well as the load components Force_x and Force_y in the x and y directions, and then pass the load parameters to the topology optimization algorithm model.

[0126] According to the results of finite element analysis, optimization iterations are performed. The parameters required for topology optimization iterations include volume fraction, model iterations, penalty factor, and filter radius, as follows: Figure 8 The software will pass these parameters to the topology optimization algorithm model to calculate the topology optimization.

[0127] After each iteration, the optimization results are evaluated to check whether they meet the design requirements. The optimization process ends when the predetermined termination conditions are met, such as reaching the maximum number of iterations or the results converge. After the topology optimization calculation is completed, the results of the optimization iteration can be viewed, such as Figure 3 shown.

[0128] This embodiment provides a computer device, whose internal structure diagram can be as follows: Fig. 9 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection.

[0129] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0130] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0131] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0132] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0133] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0134] It should also be understood that in the embodiments of the present invention, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.

[0135] 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 in the present invention can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. 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 the present invention.

[0136] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0137] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the 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. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0138] 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 embodiments of the present invention.

[0139] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for industrial equipment topology optimization based on deep learning, characterized in that: The method comprises: Based on the finite element modeling algorithm, a device model to be optimized is constructed, wherein the model device at least includes model geometry parameters and load parameters of the model in the design domain; Set displacement boundary conditions for model optimization; Constructing a convolutional neural network model, the convolutional neural network model is used to process the input multi-channel tensor to obtain the density distribution optimization result of the device model, the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input the multi-channel tensor, the multiple convolution modules are connected in sequence, the current convolution module is used to extract features according to the output result of the previous convolution module to obtain the feature data of the multi-channel tensor in sequence, the multiple transposed convolution modules are connected in sequence and correspond to the multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module; Based on the variable density method, according to the model geometry parameters and load parameters of the device model and the displacement boundary conditions, a multi-channel tensor input to the convolutional neural network is determined, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction; Inputting the multi-channel tensor into a trained convolutional neural network model to obtain a target density distribution of the device model; According to the target density distribution of the device model and in combination with the density parameters in the material set, a topology optimization result of the device model is obtained.

2. The method according to claim 1, characterized in that The convolution module includes a convolution layer, an activation function, and a maximum pooling layer; the input layer is connected to multiple convolution layers, each convolution layer is sequentially connected to an activation function and a maximum pooling layer, the convolution layer is used to extract the feature map of the input multi-channel tensor, and the activation function is used to perform nonlinear operations on the multi-channel tensor after convolution processing; The transposed convolution module includes a transposed convolution layer, a Dropout layer and a splicing layer. The transposed convolution layer is used to upsample to increase the size of the feature map, the Dropout layer is used to randomly discard neurons through the Dropout mechanism, and the splicing layer is used to splice the features extracted from different layers together and send the splicing results to the output layer.

3. The method according to claim 1, characterized in that The variable density method is expressed by the following formula: F = K·U; 0<x min ≤x≤1; Among them, U is the global displacement vector, F is the force vector, K is the global stiffness matrix, and u e and k0 are the unit displacement vector and stiffness matrix respectively, x is the design variable vector, x min is the minimum relative density vector, N is the number of elements used to discretize the design domain, P is the penalty factor, and V (x) and V0 are the material volume and design domain volume respectively, and f is the set volume fraction.

4. The method according to claim 3, characterized in that The multi-channel tensor is obtained by the following steps: Based on the variable density method and the initial design variable vector of the material in the equipment model, the displacement vector of the model in the horizontal direction and the displacement vector in the vertical direction are determined; Respectively converting the displacement vector in the horizontal direction and the displacement vector in the vertical direction into a displacement matrix in the horizontal direction and a displacement matrix in the vertical direction of the same size; Extract the initial volume fraction of the design domain and expand it into an initial volume fraction matrix of the same size as the displacement matrix in the horizontal direction and the displacement matrix in the vertical direction; The displacement matrix in the horizontal direction, the displacement matrix in the vertical direction and the initial volume fraction matrix are combined into the multi-channel tensor.

5. The method according to claim 1, characterized in that The method further comprises: Normalizing the target density distribution of the device model; The normalized device model is visualized according to the pre-set identifiers of different density values, wherein the identifiers include colors.

6. The method according to claim 1, characterized in that The convolutional neural network model is trained by the following steps: constructing an initial convolutional neural network model; Acquire training data, wherein the training data includes multiple groups of optimized equipment models, and density distribution results and optimization parameters corresponding to each equipment module, wherein the optimization parameters include at least model geometry parameters, load parameters, and displacement boundary conditions; Processing the optimization parameters to obtain a multi-channel tensor; Taking the multi-channel tensor as input data and the density distribution result of the device model as the target output, the initial convolutional neural network model is trained using a preset loss function until the convergence condition is met, thereby obtaining a trained convolutional neural network model.

7. The method according to claim 1, characterized in that The convolutional neural network model is a U-net network model.

8. A deep learning-based industrial equipment topology optimization system, characterized in that: The system comprises: A device model building module, used to build a device model to be optimized based on a finite element modeling algorithm, wherein the model device at least includes model geometry parameters and load parameters of the model in the design domain; Condition setting module, used to set displacement boundary conditions for model optimization; A convolutional neural network model construction module is used to construct a convolutional neural network model, wherein the convolutional neural network model is used to process an input multi-channel tensor to obtain a density distribution optimization result of a device model, and the convolutional neural network model includes an input layer, multiple convolution modules, multiple transposed convolution modules and an output layer; the input layer is used to input a multi-channel tensor, and multiple convolution modules are connected in sequence, and the current convolution module is used to extract features according to the output result of the previous convolution module to obtain feature data of the multi-channel tensor in sequence, and multiple transposed convolution modules are connected in sequence and correspond to multiple convolution modules one by one, wherein the current transposed convolution module splices the output result of the previous transposed convolution module and the output result of the corresponding convolution module, and restores the spatial information of the multi-channel tensor according to the spliced ​​data, and the output layer is used to obtain the density distribution optimization result of the device model according to the processing result of the transposed convolution module; An input module, for determining a multi-channel tensor input to a convolutional neural network based on a variable density method according to model geometry parameters and load parameters of the device model and the displacement boundary conditions, wherein the multi-channel tensor includes at least a horizontal displacement matrix, a vertical displacement matrix and an initial volume fraction; A target density distribution calculation module, used for inputting the multi-channel tensor into the trained convolutional neural network model to obtain the target density distribution of the device model; The determination module is used to obtain a topology optimization result of the device model according to the target density distribution of the device model and the density parameters in the material set.

9. An industrial equipment topology optimization system, characterized in that: The system comprises: A modeling module, for constructing a geometric model of the industrial equipment to be optimized according to a finite element modeling algorithm, including at least the following controls for user input: geometric parameters, load parameters, and displacement boundary conditions; An optimization module, used for performing topology optimization processing according to the industrial equipment topology optimization method based on deep learning according to any one of claims 1 to 7, to obtain a target density distribution of the industrial equipment to be optimized; A visualization module is used to visualize the industrial equipment according to the target density distribution in combination with a preset identification matching rule.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the deep learning-based industrial equipment topology optimization method according to any one of claims 1 to 7 is implemented.

Images