Structure topology optimization method, device and equipment based on TOPO-U-Net and storage medium

Abstract

The invention discloses a TOPO-U-Net-based structure topological optimization method, device and equipment and a storage medium, and the method comprises the steps: setting initial conditions, and obtaining an initial topological optimization result as sample data according to each initial condition; constructing a data set according to the sample data, and preprocessing the data set; setting an initial parameter of the TOPO-U-Net network and an initial parameter of a training mode; the preprocessed data set is input into the TOPO-U-Net network model after the initial parameters are set for training, and a structure topology optimization model based on TOPO-U-Net is constructed; and finally, performing mechanical analysis on the initial condition to be predicted to obtain an input condition, and obtaining a structure topological optimization result through the structure topological optimization model according to the input condition. According to the method, the structure topological optimization result can be generated in an extremely short time, the gray area of the structure topological optimization result is reduced, the boundary is clearer, and the structure optimization quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of structural optimization, and in particular to a structural topology optimization method, device, equipment and storage medium based on TOPO-U-Net. Background Art

[0002] In recent years, with the continuous development of the civil aviation field, weight reduction of civil aircraft has gradually become an important way to reduce costs and increase efficiency. The structural weight of civil aircraft not only affects its flexibility, but also directly affects fuel consumption and passenger capacity, thereby affecting economic benefits. The large structures on the aircraft constitute the main structural weight. Therefore, lightweight design of large structures is the key to weight reduction, and the most important method used for structural lightweighting is topological optimization.

[0003] Traditional topology optimization methods are mainly based on finite elements. After a long period of development, common methods such as variable density method (SIMP), bidirectional evolutionary structural optimization method (BESO), level set method (LS) and moving deformable component method (MMC) have been formed. Among the current topology optimization methods, the solid isotropic material with penalization model (SIMP) is widely used.

[0004] The inventors have found through research that the topology optimization method in the prior art still has at least the following defects:

[0005] A large number of grayscale units are often generated at the boundaries of the generated structure. These grayscale units constitute grayscale areas, which reduce the accuracy of the structure boundaries and thus affect the quality of structural optimization. In addition, the existing topology optimization methods are based on finite element iterations, which makes it take a long time to generate optimization results.

[0006] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention

[0007] The purpose of the present invention is to generate a structural topology optimization result in a very short time, while also solving the gray area problem and improving the quality of structural optimization.

[0008] The present invention provides a structural topology optimization method based on TOPO-U-Net, comprising:

[0009] Setting initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0010] Obtaining initial topology optimization results as sample data according to each of the initial conditions; constructing a data set according to the sample data, and preprocessing the data set;

[0011] Setting the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; inputting the preprocessed data set into the TOPO-U-Net network model after setting the initial parameters for training, and constructing a structural topology optimization model based on TOPO-U-Net;

[0012] A mechanical analysis is performed on the initial conditions to be predicted to obtain input conditions, and a structural topology optimization result is obtained through the structural topology optimization model according to the input conditions.

[0013] Preferably, in an embodiment of the present invention, the method further comprises: performing visual output and calculating evaluation index results based on the structural topology optimization results.

[0014] Preferably, in an embodiment of the present invention, constructing a data set according to the sample data includes:

[0015] Multiple groups of samples, each group of samples includes input data and labels; among them,

[0016] The input data include X-direction force, Y-direction force, X-direction displacement, Y-direction displacement and volume fraction;

[0017] The labels are the optimization results obtained by the variable density method.

[0018] Preferably, in an embodiment of the present invention, the preprocessing of the data set includes:

[0019] Extracting multiple parameters from an input data set and synthesizing a feature map according to the parameters;

[0020] Convert each of the parameters and the feature map into a tensor form;

[0021] The optimization results are extracted and converted into tensor form.

[0022] Preferably, in an embodiment of the present invention, the construction of a structural topology optimization model based on TOPO-U-Net includes:

[0023] Inputting the preprocessed data set into the input conversion module in the TOPO-U-Net network to normalize the feature map size and perform preliminary feature extraction;

[0024] Extract features and fuse features using a multi-layer perception module that uses downsampling, upsampling, and fusion HLFE modules;

[0025] The topology optimization results are predicted according to the extracted features; and the topology optimization results are corrected by using the DICELoss loss function to obtain a structural topology optimization model.

[0026] Preferably, in the embodiment of the present invention, the evaluation index results include:

[0027] Predict the time, number of intermediate density units, and intermediate density drift rate.

[0028] In another aspect of the present invention, a structural topology optimization device based on TOPO-U-Net is also provided, comprising:

[0029] An initial condition setting unit, used to set initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0030] A sample data acquisition unit, used to obtain an initial topology optimization result as sample data according to each of the initial conditions; construct a data set according to the sample data, and preprocess the data set;

[0031] A structural topology optimization model building unit is used to set the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; the preprocessed data set is input into the TOPO-U-Net network model after the initial parameters are set for training, and a structural topology optimization model based on TOPO-U-Net is built;

[0032] The structural topology optimization result generating unit is used to perform mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtain the structural topology optimization results through the structural topology optimization model according to the input conditions.

[0033] Preferably, in an embodiment of the present invention, it further comprises: an evaluation index result calculation unit, which is used to perform visual output and calculate the evaluation index result according to the structural topology optimization result.

[0034] On the other hand, a structural topology optimization device based on TOPO-U-Net is provided; the structural topology optimization device based on TOPO-U-Net includes a computer program stored on a medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the methods described in the above aspects and achieves the same technical effects.

[0035] On another aspect of the embodiment of the present invention, a storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, each step of the structural topology optimization method based on TOPO-U-Net as described in any one of the above items is implemented.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] In the present invention, initial conditions are first set, and the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor; initial topology optimization results are obtained as sample data according to the initial conditions; a data set is constructed according to the sample data, and the data set is preprocessed; initial parameters of a TOPO-U-Net network and initial parameters of a training method are set; the preprocessed data set is input into a TOPO-U-Net network model after setting various initial parameters for training, and a structural topology optimization model based on TOPO-U-Net is constructed; finally, mechanical analysis is performed on the initial conditions to be predicted to obtain input conditions, and a structural topology optimization result is obtained through a structural topology optimization model according to the input conditions; in the structural topology optimization result generated by the present invention, the grayscale area is reduced and the boundary is clearer, thereby improving the quality of structural optimization.

[0038] Furthermore, the present invention can generate accurate topology optimization results in a very short time, solving the problem of long calculation time of the traditional SIMP method.

[0039] In addition, the present invention can be used for structural topology optimization services, web applications and mobile APPs, thereby achieving expected benefits and commercial value.

[0040] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, and to make the above and other purposes, technical features and advantages of the present invention easier to understand, one or more preferred embodiments are listed below and described in detail with the accompanying drawings as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments 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.

[0042] Figure 1 It is a step diagram of the structural topology optimization method based on TOPO-U-Net described in the present invention;

[0043] Figure 2 It is a schematic diagram of the structure principle of the TOPO-U-Net network model described in the present invention;

[0044] Figure 3 It is a schematic diagram of the principle of the high and low order feature extraction blocks described in the present invention;

[0045] Figure 4 It is a schematic diagram of the loss change of the complete training process described in the present invention;

[0046] Figure 5 It is a schematic diagram of the structural topology optimization model based on TOPO-U-Net described in the present invention;

[0047] Figure 6 It is a schematic diagram of the final complete output of the drawing module of the present invention;

[0048] Figure 7 It is a schematic diagram of a structure calculated by using the SIMP method of dividing the same number of units described in the present invention;

[0049] Figure 8 It is a schematic diagram of the structure of the TOPO-U-Net-based structure topology optimization device described in the present invention;

[0050] Fig. 9 It is a schematic diagram of the structure of the structural topology optimization equipment based on TOPO-U-Net described in the present invention. DETAILED DESCRIPTION

[0051] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings, but it should be understood that the protection scope of the present invention is not limited by the specific implementation modes.

[0052] Unless explicitly stated otherwise, throughout the specification and claims, the term “comprise” or variations such as “include” or “comprising”, etc., will be understood to include the stated elements or components but not to exclude other elements or components.

[0053] In this document, the terms "first", "second", etc. are used to distinguish two different elements or parts, and are not used to limit a specific position or relative relationship. In other words, in some embodiments, the terms "first", "second", etc. can also be interchangeable.

[0054] Embodiment 1

[0055] Traditional topology optimization methods have many shortcomings, including high computational cost, numerical instability and grayscale areas. With the improvement of computer computing power and the development of artificial intelligence, algorithms represented by deep learning are becoming more mature, and many problems existing in traditional topology optimization methods have new solutions. A major advantage of machine learning is that they can discover the potential relationship between data and construct a highly nonlinear function mapping relationship between input variables and targets, which can then assist in predicting the probability of events occurring in new situations. Once the model is trained, the output result can be obtained in a very short time given the input conditions.

[0056] In order to solve the gray area problem and improve the quality of structure optimization, such as Figure 1 As shown, in an embodiment of the present invention, a structural topology optimization method based on TOPO-U-Net is provided, comprising the steps of:

[0057] S11, setting initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0058] The inventive ideas of the present invention include: TOPO-U-Net is a convolutional neural network for topology optimization, which is obtained by introducing attention mechanism, normalization method, depth-separable convolution and other technologies into U-Net. By replacing the loss function and using a large number of topology optimization results as data sets, an end-to-end model for topology optimization is trained. The model can obtain accurate topology optimization results in a very short time and will not occupy more system resources. At the same time, the problem of grayscale areas is also solved.

[0059] In the embodiment of the present invention, the initial conditions set include: initial structure, boundary conditions, external load size and position F (which can be represented by F), volume fraction, filter radius in the grid-independent filter (which can be represented by r) and penalty factor. These initial conditions are both parameterizations of the topology optimization problem and constitute the parameters of the sample, providing a basis for generating a large number of samples later.

[0060] S12, obtaining initial topology optimization results as sample data according to each of the initial conditions; constructing a data set according to the sample data, and preprocessing the data set;

[0061] In the embodiment of the present invention, a large number of topology optimization results are obtained as sample data and a data set is constructed by using a traditional method; the traditional method can use a variable density method (SIMP), set the parameters of a large number of different samples according to the initial conditions in step S11, and use Matlab code to complete the topology optimization and save the results. It should be noted here that there are many ways to implement the variable density method. In addition to using the optimization code, the finite element software can also select the variable density method as an optimization method to complete this step.

[0062] The initial conditions are divided into fixed conditions and variable conditions in step S11 according to user needs. The fixed conditions have fixed values ​​in each sample data, while the variable conditions have different values ​​in each sample data, which constitute the difference between the sample data.

[0063] The data set is obtained by inputting fixed conditions and changing conditions, using traditional methods to obtain multiple groups of samples, each group of samples consists of input and labels, thus forming a data set, and all samples in the data set are not repeated. In addition, in general, the more samples there are, the better the performance of the model, but the time used to train the model will also be longer. Users can choose the size of the data set according to the situation to balance the relationship between model performance and training time.

[0064] In an embodiment of the present invention, a data set is constructed based on the sample data, including: multiple groups of samples, each group of samples including input data and labels; wherein,

[0065] The input data include X-direction force Fx, Y-direction force Fy, X-direction displacement Ux, Y-direction displacement Uy and volume fraction Vol; the label is the optimization result obtained by the variable density method.

[0066] Furthermore, in order to improve the accuracy of the structural topology optimization model based on TOPO-U-Net, in an embodiment of the present invention, the data set is preprocessed, including:

[0067] Extract multiple parameters from the input data set and synthesize a feature map based on each parameter;

[0068] Convert all parameters and feature maps into tensor form;

[0069] Extract the optimization results and convert them into tensor form.

[0070] In an embodiment of the present invention, the data set is preprocessed by converting the saved file into a tensor form that can be recognized by the deep learning network. In a specific example, five parameters in the input can be extracted and synthesized into a feature map, each parameter is a tensor of (x+1)*(y+1)*1, and the synthesized feature map is a tensor with a size of (x+1)*(y+1)*5; the optimization result in the label is extracted, and the size of the optimization result is a tensor of x*y*1.

[0071] S13, setting the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; inputting the preprocessed data set into the TOPO-U-Net network model after setting the initial parameters for training, and constructing a structural topology optimization model based on TOPO-U-Net;

[0072] In an embodiment of the present invention, a structural topology optimization model based on TOPO-U-Net is constructed, including:

[0073] The preprocessed data set is input into the input conversion module in the TOPO-U-Net network for normalization of feature map size and preliminary feature extraction;

[0074] Extract features and fuse features using a multi-layer perception module that uses downsampling, upsampling, and fusion HLFE modules;

[0075] The topology optimization results are predicted according to the extracted features; and the topology optimization results are corrected by using the DICELoss loss function to obtain a structural topology optimization model.

[0076] Preferably, the training parameters in the structural topology optimization model include optimizer, learning rate, batch size, and training rounds, and these training parameters can be flexibly adjusted according to different situations.

[0077] like Figure 2 As shown in the figure, the structure of TOPO-U-Net network mainly consists of two parts: encoder and decoder. The encoder is responsible for extracting features from the input data, and the decoder generates segmentation maps based on these features. At the same time, the high-low-order feature extraction block (HLFE) is used to replace the jump connection in the original U-Net to retain spatial information and improve segmentation accuracy. The entire model contains 1 input conversion block, 8 depth-separable convolution blocks, 3 maximum pooling layers, 3 HLFE modules, 3 masks, and 3 upsampling layers.

[0078] The input conversion module uses 3*3 ordinary convolution and adaptive pooling to convert the size of the input tensor from (x+1)*(y+1)*5 to x*y*c. This is because when the number of units is divided into x*y, there are (x+1)*(y+1) nodes in total, and the nodes carry information such as displacement, force position, and force angle. Therefore, it is necessary to convert the size of the tensor in advance before passing it to the network to ensure that the input and label feature map sizes are consistent.

[0079] The depth-wise separable convolution blocks, maximum pooling layers, masks, and upsampling layers are all operations used to extract features and generate segmentation maps.

[0080] like Figure 3As shown in the figure, the high-low-level feature extraction block (HLFE) converts the high-dimensional feature map into a size and number of channels that match the low-dimensional feature map through a pre-defined convolution kernel size of 1x1 and bilinear upsampling, and further extracts spatial features from the two feature maps by setting the dilated convolution layer with a dilation rate of [1,2,5,7]. Each dilated convolution layer extracts features for a specific spatial range, which can cover information from local details to a wider spatial context. In particular, the HLFE module also considers mask information in the feature fusion process, splicing the processed high-order feature map with the corresponding low-order feature map and mask, and adjusting the number of channels through a convolution layer to complete the fusion of high-order and low-order features. Finally, all fused feature maps are unified through splicing and a normal convolution. In addition, the HLFE module uses the LayerNorm normalization method for the channel dimension after each convolution layer to ensure the stability and consistency of feature information in the multi-scale fusion process.

[0081] The LayerNorm normalization method can perform normalization independently for each channel, thereby more effectively controlling the internal covariate shift during training. In specific implementation, layer normalization is performed in the following steps: First, the mean μ and variance σ of each input feature channel are calculated 2 , and then normalize using the following formula:

[0082]

[0083] In the formula, x i is the input feature; μ is the mean; σ 2 is the variance; ε is a small constant to prevent division by zero errors. This step ensures that each feature channel can be effectively standardized even when the features of each channel are quite different, thereby improving the adaptability of the structural topology optimization model to different feature scales.

[0084] The DICE loss function is based on the Dice coefficient, which is an indicator for measuring the similarity between two sample sets. For image segmentation tasks, the Dice coefficient can be used to measure the overlap between the predicted segmentation results and the true labels. The calculation formula of the Dice coefficient is:

[0085]

[0086] In the formula, X and Y are the flattened vectors of the predicted results and true labels of the structural topology optimization model, respectively. |X∩Y| is the sum of the elements of X and Y multiplied bit by bit, representing the size of the intersection of the predicted results and the true labels. |X| and |Y| are the sum of the elements of X and Y, respectively, representing their respective sizes. In order to avoid the situation where the denominator is zero, a non-negative smoothing term is introduced. Based on the Dice coefficient, the Dice loss function is defined as:

[0087] Dice Loss=1-Dice Coefficient

[0088] Specifically, the calculation of Dice loss involves flattening the predicted probability map and the true label map output by the structural topology optimization model into a four-dimensional tensor form [N, C, H, W] (where N is the batch size, C is the number of channels, and H and W are the height and width of the image, respectively) into a two-dimensional [N, C*H*W] format in order to perform element-by-element multiplication to determine the intersection. Subsequently, by adding a smoothing term to prevent division by zero errors, the Dice coefficient is calculated by the ratio of twice the intersection to the sum of the elements of the prediction and the target. Finally, the Dice loss is defined as the difference between 1 and the Dice coefficient. The smaller the value, the higher the similarity between the predicted result and the true label, that is, the better the segmentation performance of the structural topology optimization model.

[0089] The DICE loss function provides an optimization target. The training process of the structural topology optimization model is to continuously adjust the parameters to minimize the loss function value. Through the back-propagation algorithm, the gradient of the loss function is used to guide the adjustment direction and amplitude of the parameters in each iteration. During the training and verification stages, the value of the loss function can be used as an indicator of the performance of the structural topology optimization model to help determine whether the structural topology optimization model is overfitting or underfitting.

[0090] The structural topology optimization model is obtained by looping all the operations in step S13. All samples in the data set are input into the TOPO-U-Net network in turn. The loop is completed according to the training rounds. After the loss function converges, the structural topology optimization model can be obtained. The loss function changes during the training process as shown in Figure 4 As shown in the figure, the horizontal axis represents the training rounds, and the vertical axis represents the loss value. The verification loss converges rapidly with the increase in the number of iterations. After 25 epochs, it has stabilized within 0.07, and then slowly decreases. After 120 epochs, it gradually stabilizes and finally converges to 0.06. Generally speaking, when the loss function converges, the model training is considered complete. Setting the training rounds to 200 takes redundancy into account to ensure convergence.

[0091] S14. Performing mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtaining structural topology optimization results through the structural topology optimization model according to the input conditions.

[0092] In the embodiment of the present invention, the structural topology optimization result is obtained by predicting the probability distribution of each pixel point through the structural topology optimization model to obtain the final density distribution. Figure 5 As shown in the figure, the optimization results under certain initial conditions are shown. The topology optimization result of the structure has a very clear structure. A large number of intermediate density units disappear and are replaced by solid units with a density of 1 or hollow units with a density of 0. Only a very small number of intermediate density units exist at the boundaries and corners of the structure.

[0093] Mechanical analysis is to convert a topology optimization problem into corresponding input conditions. Different from generating data set samples, here we only need to get the input conditions, including five parameters: X-direction force Fx, Y-direction force Fy, X-direction displacement Ux, Y-direction displacement Uy and volume fraction Vol.

[0094] The time required for generating the structural topology optimization results of the present invention is 0.01 s on average, while the generation time of the variable density method is 1.18 s, thus greatly reducing the generation time of the structural topology optimization results.

[0095] In order to better display the effect of the structural topology optimization result, in the embodiment of the present invention, the following may also be included:

[0096] S15. Visualize and output the structural topology optimization results and calculate evaluation index results.

[0097] In the embodiment of the present invention, the evaluation index results include: prediction time (Time), number of intermediate density units (Num) and intermediate density deviation rate (IDDR).

[0098] The number of intermediate density units is the total number of all units with density values ​​between 0.1 and 0.9 (including 0.1 and 0.9) in the structural topology optimization results.

[0099] The formula for the intermediate density shift rate is:

[0100]

[0101] In the formula, T S T is the number of cells with intermediate density between 0.1 and 0.5 in the label value whose intermediate density decreases in the corresponding position in the output; b F is the number of cells with intermediate density between 0.5 and 0.9 in the label value whose intermediate density increases in the corresponding position in the output; SF is the number of cells with intermediate density between 0.1 and 0.5 in the label value whose intermediate density increases in the corresponding position in the output; b It is the number of cells with intermediate density between 0.5 and 0.9 in the label value whose intermediate density decreases in the corresponding position in the output.

[0102] All evaluation criteria are calculated and saved by the evaluation module, and the saved evaluation index results are: prediction time Time=0.01s, number of intermediate density units Num=13.966, intermediate density deviation rate IDDR=85.42%.

[0103] In the embodiment of the present invention, the topology optimization result is visualized and outputted through a drawing module, and the drawing module outputs the final structure, such as Figure 6 As shown, it includes the density information on the unit. By canceling the display of the grid and the annotation of density details, we can better understand the shape of the structure. Figure 7 The structure calculated by the SIMP method with the same number of units is shown, from which it can be seen that the grayscale area in the structural topology optimization result generated by the present invention is reduced and the boundary is clearer.

[0104] In summary, in an embodiment of the present invention, initial conditions are first set, and the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor; initial topology optimization results are obtained as sample data according to the initial conditions; a data set is constructed according to the sample data, and the data set is preprocessed; initial parameters of the TOPO-U-Net network and initial parameters of the training method are set; the preprocessed data set is input into the TOPO-U-Net network model after setting the initial parameters for training, and a structural topology optimization model based on TOPO-U-Net is constructed; finally, a mechanical analysis is performed on the initial conditions to be predicted to obtain input conditions, and a structural topology optimization result is obtained according to the input conditions through the structural topology optimization model; in the structural topology optimization result generated by the present invention, the grayscale area is reduced and the boundary is clearer, thereby improving the quality of structural optimization.

[0105] Furthermore, the present invention can generate accurate topology optimization results in a very short time, solving the problem of long calculation time of the traditional SIMP method.

[0106] In addition, the present invention can be used for structural topology optimization services, web applications, and mobile APPs, achieving expected benefits and commercial value.

[0107] Embodiment 2

[0108] Corresponding to the method embodiment, on the other side of the embodiment of the present invention, a structure topology optimization device based on TOPO-U-Net is also provided. Figure 8A schematic diagram of a structure topology optimization device based on TOPO-U-Net provided in an embodiment of the present invention is shown. The structure topology optimization device based on TOPO-U-Net is Figure 1 The device corresponding to the structure topology optimization method based on TOPO-U-Net described in the corresponding embodiment, that is, implemented by means of a virtual device Figure 1 In the TOPO-U-Net-based structural topology optimization method in the corresponding embodiment, each virtual module constituting the TOPO-U-Net-based structural topology optimization device can be executed by an electronic device, such as a network device, a terminal device or a server. Specifically, the TOPO-U-Net-based structural topology optimization device in the embodiment of the present invention includes:

[0109] An initial condition setting unit 01 is used to set initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0110] In the embodiment of the present invention, the initial conditions set include: initial structure, boundary conditions, external load size and position (which can be represented by F), volume fraction, filter radius in the grid-independent filter (which can be represented by r) and penalty factor. These initial conditions are both parameterizations of the topology optimization problem and constitute the parameters of the sample, providing a basis for generating a large number of samples later.

[0111] The sample data acquisition unit 02 is used to obtain the initial topology optimization results as sample data according to the initial conditions; construct a data set according to the sample data, and pre-process the data set;

[0112] In the embodiment of the present invention, a data set is constructed, including: multiple groups of samples, each group of samples includes input data and labels; wherein,

[0113] The input data include X-direction force Fx, Y-direction force Fy, X-direction displacement Ux, Y-direction displacement Uy and volume fraction Vol; the label is the optimization result obtained by the variable density method.

[0114] Furthermore, in order to improve the accuracy of the structural topology optimization model based on TOPO-U-Net, in an embodiment of the present invention, the data set is preprocessed, including:

[0115] Extract multiple parameters from the input data set and synthesize a feature map based on each parameter;

[0116] Each parameter and feature map is converted into tensor form;

[0117] Extract the optimization results and convert them into tensor form.

[0118] In an embodiment of the present invention, the data set is preprocessed by converting the saved file into a tensor form that can be recognized by the deep learning network. In a specific example, five parameters in the input can be extracted and synthesized into a feature map, each parameter is a tensor of (x+1)*(y+1)*1, and the synthesized feature map is a tensor with a size of (x+1)*(y+1)*5; the optimization result in the label is extracted, and the size of the optimization result is a tensor of x*y*1.

[0119] The structural topology optimization model construction unit 03 is used to set the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; the preprocessed data set is input into the TOPO-U-Net network model after the initial parameters are set for training, and a structural topology optimization model based on TOPO-U-Net is constructed;

[0120] In an embodiment of the present invention, a structural topology optimization model based on TOPO-U-Net is constructed, including:

[0121] The preprocessed data set is input into the input conversion module in the TOPO-U-Net network for normalization of feature map size and preliminary feature extraction;

[0122] Extract features and fuse features using a multi-layer perception module that uses downsampling, upsampling, and fusion HLFE modules;

[0123] The topology optimization results are predicted based on the extracted features, and the DICE Loss function is used to correct the model to obtain the structural topology optimization model.

[0124] Preferably, the training parameters in the structural topology optimization model include optimizer, learning rate, batch size, and training rounds, and these training parameters can be flexibly adjusted according to different situations.

[0125] The structural topology optimization result generating unit 04 is used to perform mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtain the structural topology optimization results through the structural topology optimization model according to the input conditions.

[0126] In the embodiment of the present invention, the structural topology optimization result is obtained by predicting the probability distribution of each pixel point through the structural topology optimization model to obtain the final density distribution. Figure 5 As shown in the figure, the optimization result under a certain initial condition is shown as an example. The topology optimization result of the structure has a very clear structure. A large number of intermediate density units disappear and are replaced by solid units with a density of 1 or hollow units with a density of 0. Only a very small number of intermediate density units exist at the boundaries and corners of the structure.

[0127] Preferably, in the embodiment of the present invention, the following may also be included:

[0128] An evaluation index result calculation unit (not shown in the figure) is used to perform visual output and calculate evaluation index results based on the structural topology optimization results.

[0129] In the embodiment of the present invention, the evaluation index results include: prediction time (Time), number of intermediate density units (Num) and intermediate density deviation rate (IDDR).

[0130] The number of intermediate density units is the total number of all units with density values ​​between 0.1 and 0.9 (including 0.1 and 0.9) in the structural topology optimization results.

[0131] The formula for the intermediate density shift rate is:

[0132]

[0133] In the formula, T S T is the number of cells with intermediate density between 0.1 and 0.5 in the label value whose intermediate density decreases in the corresponding position in the output; b F is the number of cells with intermediate density between 0.5 and 0.9 in the label value whose intermediate density increases in the corresponding position in the output; S F is the number of cells with intermediate density between 0.1 and 0.5 in the label value whose intermediate density increases in the corresponding position in the output; b It is the number of cells with intermediate density between 0.5 and 0.9 in the label value whose intermediate density decreases in the corresponding position in the output.

[0134] All evaluation criteria are calculated and saved by the evaluation module, and the saved evaluation index results are: prediction time Time=0.01s, number of intermediate density units Num=13.966, intermediate density deviation rate IDDR=85.42%.

[0135] It should be noted that the specific implementation method and technical effect of the structure topology optimization device based on TOPO-U-Net in the embodiment of the present invention can be referred to Figure 1 The corresponding structural topology optimization method based on TOPO-U-Net will not be described here.

[0136] Embodiment 3

[0137] Corresponding to the method embodiment, in the embodiment of the present invention, a structural topology optimization device based on TOPO-U-Net is also provided, such as a terminal, a server, etc. Among them, the server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto.

[0138] An example of a hardware structure block diagram of a structure topology optimization device based on TOPO-U-Net provided in an embodiment of the present invention is shown in FIG. Fig. 9 As shown, it may include:

[0139] Processor 1, communication interface 2, memory 3 and communication bus 4;

[0140] The processor 1, the communication interface 2, and the memory 3 communicate with each other via the communication bus 4;

[0141] Optionally, the communication interface 2 may be an interface of a communication module, such as an interface of a GSM module;

[0142] The processor 1 may be a central processing unit CPU, or an application-specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0143] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0144] The processor 1 is specifically used to execute the computer program stored in the memory 3 to perform the following steps:

[0145] S11, setting initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0146] S12, obtaining initial topology optimization results as sample data according to each of the initial conditions; constructing a data set according to the sample data, and preprocessing the data set;

[0147] S13, setting the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; inputting the preprocessed data set into the TOPO-U-Net network model after setting the initial parameters for training, and constructing a structural topology optimization model based on TOPO-U-Net;

[0148] S14. Performing mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtaining structural topology optimization results through the structural topology optimization model according to the input conditions.

[0149] Preferably, in the embodiment of the present invention, the following may also be included:

[0150] S15. Visualize and output the structural topology optimization results and calculate evaluation index results.

[0151] The above product can execute the method provided in the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the structural topology optimization method based on TOPO-U-Net provided in the embodiment of the present invention.

[0152] Embodiment 4

[0153] In an embodiment of the present invention, a storage medium is further provided, which may store a program suitable for execution by a processor, wherein the program is used to:

[0154] S11, setting initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor;

[0155] S12, obtaining initial topology optimization results as sample data according to each of the initial conditions; constructing a data set according to the sample data, and preprocessing the data set;

[0156] S13, setting the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; inputting the preprocessed data set into the TOPO-U-Net network model after setting the initial parameters for training, and constructing a structural topology optimization model based on TOPO-U-Net;

[0157] S14. Performing mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtaining structural topology optimization results through the structural topology optimization model according to the input conditions.

[0158] Preferably, in the embodiment of the present invention, the following may also be included:

[0159] S15. Visualize and output the structural topology optimization results and calculate evaluation index results.

[0160] Optionally, the detailed functions and extended functions of the program may refer to the above description.

[0161] The above product can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the methods provided by other embodiments of the present invention.

[0162] 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.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0164] 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.

[0165] 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.

[0166] It should be understood that in the embodiments of the present application, the various embodiments and features can be combined with each other to solve the aforementioned technical problems.

[0167] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0168] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A structural topology optimization method based on TOPO-U-Net, characterized in that: include: Setting initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor; Obtaining initial topology optimization results as sample data according to each of the initial conditions; constructing a data set according to the sample data, and preprocessing the data set; Setting the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; inputting the preprocessed data set into the TOPO-U-Net network model after setting the initial parameters for training, and constructing a structural topology optimization model based on TOPO-U-Net; A mechanical analysis is performed on the initial conditions to be predicted to obtain input conditions, and a structural topology optimization result is obtained through the structural topology optimization model according to the input conditions.

2. The structural topology optimization method based on TOPO-U-Net according to claim 1, characterized in that: Also includes: Visual output is performed based on the structural topology optimization results and evaluation index results are calculated.

3. The structural topology optimization method based on TOPO-U-Net according to claim 2 is characterized in that: Constructing a data set according to the sample data includes: Multiple groups of samples, each group of samples includes input data and labels; among them, The input data include X-direction force, Y-direction force, X-direction displacement, Y-direction displacement and volume fraction; The labels are the optimization results obtained by the variable density method.

4. The structural topology optimization method based on TOPO-U-Net according to claim 3 is characterized in that: The preprocessing of the data set includes: Extracting multiple parameters from an input data set and synthesizing a feature map according to the parameters; Convert each of the parameters and the feature map into a tensor form; The optimization results are extracted and converted into tensor form.

5. The structural topology optimization method based on TOPO-U-Net according to claim 4 is characterized in that: The construction of a structural topology optimization model based on TOPO-U-Net includes: Inputting the preprocessed data set into the input conversion module in the TOPO-U-Net network to normalize the feature map size and perform preliminary feature extraction; Extract features and fuse features using a multi-layer perception module that uses downsampling, upsampling, and fusion HLFE modules; The topology optimization results are predicted according to the extracted features; and the topology optimization results are corrected by using the DICE Loss function to obtain a structural topology optimization model.

6. The structural topology optimization method based on TOPO-U-Net according to claim 2, characterized in that: The evaluation index results include: Predict the time, number of intermediate density units, and intermediate density drift rate.

7. A structural topology optimization device based on TOPO-U-Net, characterized in that: include: An initial condition setting unit, used to set initial conditions, wherein the initial conditions include: initial structure, load conditions, volume fraction, filter radius and penalty factor; A sample data acquisition unit, used to obtain an initial topology optimization result as sample data according to each of the initial conditions; construct a data set according to the sample data, and preprocess the data set; A structural topology optimization model building unit is used to set the initial parameters of the TOPO-U-Net network and the initial parameters of the training method; the preprocessed data set is input into the TOPO-U-Net network model after the initial parameters are set for training, and a structural topology optimization model based on TOPO-U-Net is built; The structural topology optimization result generating unit is used to perform mechanical analysis on the initial conditions to be predicted to obtain input conditions, and obtain the structural topology optimization results through the structural topology optimization model according to the input conditions.

8. The TOPO-U-Net-based structural topology optimization device according to claim 7, characterized in that: Also includes: The evaluation index result calculation unit is used to perform visual output and calculate the evaluation index result according to the structural topology optimization result.

9. A structural topology optimization device based on TOPO-U-Net, characterized in that: include: Memory for storing computer programs; A processor is used to call and execute the computer program to implement the steps of the structural topology optimization method based on TOPO-U-Net as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The method comprises a software program, wherein the software program is suitable for executing, by a processor, the steps of the structural topology optimization method based on TOPO-U-Net as claimed in any one of claims 1 to 7.

Images