Convolutional neural network-based structure displacement response prediction method and device, and medium
Through the structural displacement response prediction method based on convolutional neural network, the problem of lack of universality of deep learning methods in the prior art is solved, efficient prediction of multiple structures and loading conditions is achieved, and the accuracy and efficiency of prediction are improved.
Patent Information
- Application Number
- CN202411420756.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-12
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing deep learning methods lack universality in structural displacement response prediction, require frequent modification of network structure, and input data requires manual screening, which limits its promotion in different application scenarios.
The structural displacement response prediction method based on convolutional neural network is adopted, and the structure displacement response prediction is achieved by obtaining the input data of the structure for discretization and preprocessing, and the convolutional neural network model is constructed, and the model weight is optimized through iterative training to achieve the prediction of structural displacement response.
Improves the universality and prediction accuracy of the model, reduces the workload of model modification, can be quickly applied to a variety of structure types and loading conditions, and improves prediction efficiency and speed.
Smart Images

Figure CN118940387A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of structural analysis, and in particular to a method, device and medium for predicting structural displacement response based on a convolutional neural network. Background Art
[0002] In the field of structural engineering, the prediction of structural displacement response is an important topic. Traditional methods are mainly based on numerical simulation and analysis, among which commonly used methods include finite element analysis, boundary element analysis, analytical solution, etc. These methods can provide accurate results, but the amount of calculation is large when dealing with large-scale structures and complex loading conditions, and the structure needs to be simplified and assumed, which has certain limitations. In recent years, methods based on machine learning and deep learning have been widely used in the field of structural displacement response prediction. Among them, convolutional neural network (CNN) is a deep learning model that has been widely explored and applied, and it has achieved remarkable results in the fields of image processing and pattern recognition. However, in the field of structural analysis and calculation, the application and integration of deep learning methods are still in the exploratory stage. Most of the existing methods use relatively primitive feedforward neural networks, and a small number of them use deep neural networks or convolutional neural networks for global structural displacement response prediction. However, these studies are mainly aimed at specific tasks, lack of general models, and need to modify and quickly adjust the network structure to adapt to new target tasks. In addition, the input data usually needs to be manually screened and lacks universality. The characteristics of manually processed data are inconsistent with the "end-to-end" intelligent method, which limits the promotion of deep learning methods in different application scenarios. Therefore, it is necessary to further study and develop new methods to improve the application of deep learning methods in the prediction of structural displacement response. Summary of the invention
[0003] The purpose of the present invention is to provide a method, device and medium for predicting structural displacement response based on a convolutional neural network in order to address the problem that deep learning methods in the prior art lack universality.
[0004] The objective of the present invention is achieved through the following technical solutions: In a first aspect, an embodiment of the present invention provides a method for predicting structural displacement response based on a convolutional neural network, comprising the following steps: (1) Obtaining input data for the structure, including defining the geometry and dimensions of the structure, determining material properties, and clarifying the boundary conditions of the structure; (2) Discretize the geometric shape of the structure to divide it into nodes and units, and determine the connection relationship between the nodes; preprocess the input data of the structure and the obtained node and unit information to convert them into the input data representation required by the structural displacement response prediction model, and obtain the preprocessed structural data matrix to be input; (3) Select a convolutional neural network and adjust and configure its specific structure to construct a structural displacement response prediction model; (4) Using the preprocessed structural data matrix to be input obtained in step (2) as input samples, obtaining the actual structural displacement response as target output data, and constructing a training sample set based on the input samples and the target output data; (5) Iteratively train the structural displacement response prediction model using the training sample set. According to the loss function of the structural displacement response prediction model, the weight and bias parameters of the structural displacement response prediction model are optimized and updated using the back propagation method until the preset training rounds are reached to obtain the trained structural displacement response prediction model. (6) Verifying the trained structural displacement response prediction model, and fine-tuning the trained structural displacement response prediction model based on the verification results to obtain the final structural displacement response prediction model for predicting the structural displacement response.
[0005] Furthermore, the step (1) specifically includes: The input data of the structure is obtained, specifically including: firstly, the geometric shape and size of the structure are defined, specifically including the length, width and height parameters of the structure, which are modeled and determined by computer-aided design software; then, the material property parameters are determined, specifically including the elastic modulus and density parameters of the material; finally, the boundary conditions of the structure are clarified, specifically including constraint conditions and loading conditions, wherein the constraint conditions are used to describe the fixed parts of the structure or the restricted degrees of freedom of movement, and the loading conditions are used to describe the external effects applied to the structure.
[0006] Furthermore, the discretization of the geometric shape of the structure to divide it into nodes and units and determining the connection relationship between the nodes specifically includes: First, a meshing method is used to divide the geometric shape of the structure, i.e., a simply supported plate, into discrete nodes and units; wherein the nodes are discrete points on the structure; the units are discrete regions connecting the nodes; the meshing method includes an orthogonal meshing method and a triangulation method; Then, in the discretization process, the connection relationship between nodes is determined through the node numbers and topological relationships.
[0007] Furthermore, the preprocessing of the input data of the structure and the obtained node and unit information to convert them into the input data representation required by the structural displacement response prediction model specifically includes: The discretized node and unit information is converted into the input data representation required by the structural displacement response prediction model and used as the structural data matrix to be input; The input data of the structure is standardized; wherein, when the boundary conditions in the input data of the structure are standardized, an approximate distance function is used for processing; The geometric shape and size, material property parameters and boundary conditions in the standardized input data are mapped to different channels of the structure data matrix to be input, so as to obtain the preprocessed structure data matrix to be input.
[0008] Furthermore, the step (3) specifically includes: First, a convolutional neural network is selected to process the structural data matrix and extract structural features; Then, according to the requirements and structural characteristics of the structural problem, the specific structure of the convolutional neural network is configured and adjusted, including: selecting combinations of convolutional layers, pooling layers, fully connected layers, and activation functions of different types, numbers, and connection orders.
[0009] Furthermore, the step (5) specifically includes: First, the weights and bias parameters of the structural displacement response prediction model are initialized; then, the input samples in the training sample set are used as the input data of the structural displacement response prediction model to obtain the predicted structural displacement response output by the structural displacement response prediction model; the loss function of the structural displacement response prediction model is calculated according to the predicted structural displacement response and the target output data in the training sample set, and the weights and bias parameters of the structural displacement response prediction model are optimized and updated by the back propagation method with the minimization of the loss function of the structural displacement response prediction model as the optimization goal, until the preset training rounds are reached, and the trained structural displacement response prediction model is obtained; wherein, the loss function of the structural displacement response prediction model uses the mean square error or the mean absolute error.
[0010] A second aspect of an embodiment of the present invention provides a structural displacement response prediction device based on a convolutional neural network, comprising one or more processors and a memory, wherein the memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above-mentioned structural displacement response prediction method based on a convolutional neural network.
[0011] A third aspect of an embodiment of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, is used to implement the above-mentioned method for predicting structural displacement response based on a convolutional neural network.
[0012] The beneficial effects of the present invention are as follows: the present invention has higher versatility and can be applied to a variety of structural types and loading conditions; the present invention can solve a class of homogeneous problem families and reduce the workload of model modification and adjustment; the present invention makes full use of the advantages of convolutional neural networks in processing images and spatial data, and by extracting features such as the geometric shape, material properties and loading conditions of the structure, it can better capture the relevant information of the structural displacement response, thereby improving the accuracy and precision of the prediction; in addition, the present invention predicts the structural displacement response based on the convolutional neural network and has the ability to be quickly promoted and applied; through the training and parallel calculation of large-scale data sets, the present invention can quickly predict the structural displacement response and improve the efficiency and speed of the prediction; the present invention can improve the prediction performance of the structural displacement response by fully training and optimizing the convolutional neural network model, which is of great significance to the fields of structural design, analysis and evaluation; in summary, the present invention has obvious advantages in versatility, feature extraction, rapid promotion and application, and prediction performance, and provides an efficient, accurate and reliable solution for research and practice in the field of structural analysis and calculation. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 It is a technical circuit flow chart of the structural displacement response prediction method based on convolutional neural network of the present invention; Figure 2 It is a schematic diagram of the structure discretization and data preprocessing results of the present invention; Figure 3 is a schematic diagram of the network structure of the structural displacement response prediction model of the present invention; Figure 4 It is a schematic diagram of the structural displacement response prediction result of the present invention; Figure 5 It is a structural schematic diagram of a convolutional neural network-based structural displacement response prediction device of the present invention. DETAILED DESCRIPTION
[0014] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. It should be understood that the above general description and the detailed description that follows are merely exemplary and explanatory and cannot limit the present application.
[0015] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.
[0016] It should be understood that, although the term first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of..." or "when..." or "in response to determination". Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process and method including a series of elements include not only those elements, but also include other elements not clearly listed, or also include elements inherent to such process and method. In the absence of more restrictions, the elements limited by the statement "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements.
[0017] The present invention is described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.
[0018] The structural displacement response prediction method based on convolutional neural network described in the present invention is illustrated by taking a simply supported plate as an example. Figure 1 As shown in the figure, in order to realize the displacement response prediction of the simply supported plate structure, the following steps are specifically included: (1) Obtaining the input data of the structure, including defining the geometric shape and size of the structure, determining the material properties, and clarifying the boundary conditions of the structure.
[0019] Specifically, the input data of the structure is obtained, including: first, the geometry and size of the structure are defined, that is, the geometry and size of the simply supported plate are defined, including parameters such as the length, width and height of the structure, which can be modeled and determined by computer-aided design (CAD) software or other related tools. Then, the material property parameters are determined, including other necessary material parameters such as the elastic modulus and density of the material, so as to be used in subsequent analysis and prediction. These material property parameters have an important influence on the behavior and response of the structure. Finally, the boundary conditions of the structure are clarified, including constraints and loading conditions. Constraints are used to describe the fixed parts of the structure or the restricted degrees of freedom of movement. Loading conditions are used to describe external effects such as force, pressure or displacement applied to the structure, that is, the simply supported plate. These boundary conditions are crucial for predicting the displacement response of the structure.
[0020] (2) Discretize the geometric shape of the structure to divide it into nodes and units, and determine the connection relationship between the nodes; preprocess the input data of the structure and the obtained node and unit information to convert them into the input data representation required by the structural displacement response prediction model, and obtain the preprocessed structural data matrix to be input.
[0021] Furthermore, the geometric shape of the structure is discretized to divide it into nodes and units, and the connection relationship between the nodes is determined, specifically including: first, the geometric shape of the structure, i.e., the simply supported plate, is divided into discrete nodes and units using a meshing method; wherein, a node is a discrete point on the structure, i.e., the simply supported plate, which can be selected on the surface or inside of the simply supported plate; a unit is a discrete area connecting nodes, usually represented by a simple geometric unit such as a triangle or a quadrilateral; the meshing method includes an orthogonal meshing method and a triangulation method, etc. Then, in the discretization process, the connection relationship between the nodes is determined by the node numbering and topological relationship, which can facilitate the establishment of the connection and interaction relationship between the nodes in subsequent analysis and prediction.
[0022] It should be noted that the node number is the identifier of each discrete node, which is a continuous integer starting from 1; the topological relationship refers to the connection between nodes, which is represented by the node number list contained in each unit. By checking the node number of each unit, you can determine which nodes have a direct connection relationship, thereby determining the node number of each unit and the connection between them.
[0023] For example, assume that there are four nodes, which are numbered 1, 2, 3, and 4 respectively; there are two triangular units, each of which is composed of three nodes, that is: unit 1 is composed of nodes 1, 2, and 3, and unit 2 is composed of nodes 2, 3, and 4; the corresponding topological relationship list is shown in Table 1.
[0024] Table 1: List of topological relationships between nodes of each unit
[0025] According to the topological relationship list shown in Table 1, the following connection relationships can be determined: for unit 1, nodes 1, 2, and 3 are connected to each other, that is, there is a connection relationship between node 1 and node 2, there is a connection relationship between node 1 and node 3, and there is a connection relationship between node 2 and node 3; for unit 2, nodes 2, 3, and 4 are connected to each other, that is, there is a connection relationship between node 2 and node 3 (repeated with unit 1), there is a connection relationship between node 2 and node 4, and there is a connection relationship between node 3 and node 4. In summary, the node pairs with connection relationships are: node 1-node 2, node 1-node 3, node 2-node 3, node 2-node 4, and node 3-node 4; among them, there is no direct connection relationship between node 1 and node 4 because they do not appear in the same unit. Therefore, the connection relationship between nodes can be determined by the node number and topological relationship.
[0026] Furthermore, the input data of the structure and the obtained node and unit information are preprocessed to convert them into the input data representation required by the structural displacement response prediction model, specifically including: converting the discretized node and unit information into the input data representation required by the structural displacement response prediction model, such as a matrix or a tensor, to facilitate subsequent calculations and processing, and using it as the structural data matrix to be input; standardizing the input data of the structure to ensure that the data is within the same scale range; mapping the geometric shape and size, material property parameters and boundary conditions in the standardized input data to different channels of the structural data matrix to be input, and obtaining the preprocessed structural data matrix to be input, which is conducive to the training and convergence of the structural displacement response prediction model, as shown in the schematic diagram. Figure 2 As shown in the figure, when the boundary conditions in the input data of the structure are standardized, the approximate distance functions (ADFs) are used to improve the recognition ability of the structural displacement response prediction model for the changes in boundary conditions.
[0027] Specifically, the process of using approximate distance functions (ADFs) to handle boundary conditions includes: ① defining R functions to represent various logical operations, the signs of these functions are determined by the signs of their parameters, and do not depend on the size of the parameters. ② constructing a signed distance function f(x) and a pruning function t(x) for a specific geometric shape (such as a line segment) to accurately represent the boundary of the geometric shape. The approximate distance function ϕ(x) is constructed by combining the distance function f(x) and the pruning function t(x). ③ using these ADFs as an additional parameter field λ(x) to impose boundary conditions in numerical calculations.
[0028] (3) Select an appropriate convolutional neural network and adjust and configure its specific structure to construct a structural displacement response prediction model.
[0029] Specifically, we first select an appropriate convolutional neural network to process the structural data matrix and extract structural features. Then, we configure and adjust the specific structure of the convolutional neural network according to the requirements of the structural problem and the structural features, including: selecting different types, numbers, and connection orders of convolutional layers, pooling layers, fully connected layers, and activation function combinations. Common activation functions include ReLU, Leaky ReLU, and sigmoid.
[0030] It should be understood that the structural displacement response prediction models corresponding to different structures have different specific structural components, including but not limited to convolutional layers, pooling layers, fully connected layers and activation functions. Of course, other network layers such as regularization layers can be added according to actual needs. The number, type, connection order, etc. of each network layer will be different. Figure 3 The structural displacement response prediction model shown is only an example.
[0031] Exemplarily, in this embodiment, an appropriate convolutional neural network is first selected so that the convolutional neural network can effectively process the structural data matrix and extract useful structural features, such as U-Net. Since U-Net is characterized by an encoder-decoder structure and contains jump connections, U-Net is selected as the structure of the convolutional neural network, which can extract multi-scale features and retain detail information. When using U-Net, the specific structure of the convolutional neural network can be configured and adjusted according to the requirements and structural characteristics of the specific structural problem. For example, depending on the size and complexity of the simply supported plate, the number of convolutional layers and pooling layers in the encoder and decoder can be increased or decreased; similarly, the size of the convolution kernel, the choice of activation function, and the setting of other hyperparameters can be adjusted as needed. After configuring and adjusting the specific structure of the convolutional neural network in this example, the structure of the structural displacement response prediction model is finally obtained as shown in the figure. Figure 3 shown.
[0032] (4) The preprocessed structural data matrix to be input obtained in step (2) is used as an input sample, the actual structural displacement response is obtained as the target output data, and a training sample set is constructed based on the input sample and the target output data.
[0033] Specifically, the preprocessed structural data matrix to be input obtained in step (2) is used as the input sample. The real structural displacement response is obtained as the target output data through experimental measurement, numerical simulation or existing data sets from other reliable data sources; wherein, when obtaining the real structural displacement response, it is necessary to ensure that the real structural displacement response corresponds to the structural data matrix to establish a corresponding relationship between input and output. A training sample set is generated based on the input sample and the target output data. The structural data information is used as the input data, and the real displacement response is used as the target output data. According to the simply supported plate problem, multiple samples are generated to cover different structural conditions and displacement response modes.
[0034] (5) The structural displacement response prediction model is iteratively trained using the training sample set. According to the loss function of the structural displacement response prediction model, the weight and bias parameters of the structural displacement response prediction model are optimized and updated using the back propagation method until the preset training rounds are reached. The trained structural displacement response prediction model is obtained so that the trained structural displacement response prediction model can learn the characteristics and behavior of the structure and perform accurate prediction and analysis.
[0035] Specifically, the weights and bias parameters of the structural displacement response prediction model are first initialized, and the initialization weights and bias parameters can use random initialization or pre-trained weights. Then, the input samples in the training sample set are used as the input data of the structural displacement response prediction model, and the predicted structural displacement response output by the structural displacement response prediction model is obtained by the forward propagation method. Specifically, the input samples are passed to the input layer of the structural displacement response prediction model, and after the calculation and activation functions of each network layer, the predicted structural displacement response output by the structural displacement response prediction model is finally obtained. The loss function of the structural displacement response prediction model is calculated according to the predicted structural displacement response and the target output data in the training sample set, and the loss function of the structural displacement response prediction model is minimized as the optimization goal, which is used to minimize the difference between the predicted output and the target output. The weights and bias parameters of the structural displacement response prediction model are optimized and updated by the back propagation method until the preset training rounds are reached, and the trained structural displacement response prediction model is obtained. Among them, the loss function of the structural displacement response prediction model is selected from the mean square error (MSE) or the mean absolute error (MAE) or other commonly used loss functions.
[0036] (6) Verifying the trained structural displacement response prediction model, and fine-tuning the trained structural displacement response prediction model based on the verification results to obtain the final structural displacement response prediction model for predicting the structural displacement response.
[0037] Specifically, the trained structural displacement response prediction model is verified and evaluated using a validation set or a test set, its performance and accuracy are evaluated, its prediction accuracy and generalization ability on unseen data are checked, and the structural displacement response prediction model is fine-tuned and improved as needed, which solves the universality problem of structural response prediction and provides an efficient and accurate prediction method based on convolutional neural network. Specifically, first, the parameters (weights and biases) of the previously trained structural displacement response prediction model are loaded. Using the input data in the validation set, the predicted output of the structural displacement response prediction model is obtained by the forward propagation method. The predicted output of the structural displacement response prediction model is compared with the target output data in the validation set, and the evaluation index is calculated to measure the performance of the structural displacement response prediction model. The results of the evaluation index are analyzed to evaluate the performance of the structural displacement response prediction model on the validation set, which helps to understand the prediction accuracy and generalization ability of the structural displacement response prediction model and check whether there are problems such as overfitting or underfitting. According to the results of the validation set, the structural displacement response prediction model can be adjusted and improved. After the evaluation and adjustment of the validation set, the test set can be used to make a final evaluation of the structural displacement response prediction model. The test set is data that the model has never seen before and is used to evaluate the performance of the structural displacement response prediction model in real scenarios. The test data set will be independent of the training data set and the validation data set. Figure 4 shown.
[0038] It should be noted that, in the above examples, a simply supported plate is used as a structure to illustrate its displacement prediction, but the present invention is not limited to this, and may also include displacement response prediction of civil engineering structures.
[0039] Corresponding to the above-mentioned embodiment of the structural displacement response prediction method based on convolutional neural network, the present invention also provides an embodiment of the structural displacement response prediction device based on convolutional neural network.
[0040] See also Figure 5 An embodiment of the present invention provides a structural displacement response prediction device based on a convolutional neural network, comprising one or more processors and a memory, wherein the memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the structural displacement response prediction method based on a convolutional neural network in the above embodiment.
[0041] The embodiment of the structural displacement response prediction device based on convolutional neural network of the present invention can be applied to any device with data processing capability, and the device with data processing capability can be a device or apparatus such as a computer. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of any device with data processing capability in which it is located reading the corresponding computer program instructions in the non-volatile memory into the internal memory for execution. From the hardware level, if Figure 5 As shown in the figure, it is a hardware structure diagram of any device with data processing capability where the structural displacement response prediction device based on convolutional neural network of the present invention is located, except Figure 5 In addition to the processor, memory, network interface, and non-volatile memory shown, any device with data processing capabilities in which the apparatus in the embodiments is located may also include other hardware, generally based on the actual functions of the device with data processing capabilities, which will not be described in detail.
[0042] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.
[0043] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative work.
[0044] An embodiment of the present invention further provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the method for predicting structural displacement response based on a convolutional neural network in the above embodiment is implemented.
[0045] The computer-readable storage medium may be an internal storage unit of any device with data processing capability described in any of the aforementioned embodiments, such as a hard disk or a memory. The computer-readable storage medium may also be any device with data processing capability, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capability and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capability, and may also be used to temporarily store data that has been output or is to be output.
[0046] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting structural displacement response based on convolutional neural network, characterized in that: The following steps are involved: (1) Obtaining input data for the structure, including defining the geometry and dimensions of the structure, determining material properties, and clarifying the boundary conditions of the structure; (2) Discretize the geometric shape of the structure to divide it into nodes and units and determine the connection relationship between nodes; Preprocessing the input data of the structure and the obtained node and unit information to convert them into the input data representation form required by the structural displacement response prediction model, and obtaining the preprocessed structural data matrix to be input; (3) Select a convolutional neural network and adjust and configure its specific structure to construct a structural displacement response prediction model; (4) Using the preprocessed structural data matrix to be input obtained in step (2) as input samples, obtaining the actual structural displacement response as target output data, and constructing a training sample set based on the input samples and the target output data; (5) Iteratively train the structural displacement response prediction model using the training sample set. According to the loss function of the structural displacement response prediction model, the weight and bias parameters of the structural displacement response prediction model are optimized and updated using the back propagation method until the preset training rounds are reached to obtain the trained structural displacement response prediction model. (6) Verifying the trained structural displacement response prediction model, and fine-tuning the trained structural displacement response prediction model based on the verification results to obtain the final structural displacement response prediction model for predicting the structural displacement response.
2. The structural displacement response prediction method based on convolutional neural network according to claim 1 is characterized in that: The step (1) specifically includes: The input data of the structure is obtained, specifically including: firstly, the geometric shape and size of the structure are defined, specifically including the length, width and height parameters of the structure, which are modeled and determined by computer-aided design software; then, the material property parameters are determined, specifically including the elastic modulus and density parameters of the material; finally, the boundary conditions of the structure are clarified, specifically including constraint conditions and loading conditions, wherein the constraint conditions are used to describe the fixed parts of the structure or the restricted degrees of freedom of movement, and the loading conditions are used to describe the external effects applied to the structure.
3. The structural displacement response prediction method based on convolutional neural network according to claim 1 is characterized in that: Discretizing the geometric shape of the structure to divide it into nodes and units and determining the connection relationship between the nodes specifically includes: First, a meshing method is used to divide the geometric shape of the structure, i.e., a simply supported plate, into discrete nodes and units; wherein the nodes are discrete points on the structure; the units are discrete regions connecting the nodes; the meshing method includes an orthogonal meshing method and a triangulation method; Then, in the discretization process, the connection relationship between nodes is determined through the node numbers and topological relationships.
4. The structural displacement response prediction method based on convolutional neural network according to claim 1 is characterized in that: The preprocessing of the input data of the structure and the obtained node and unit information to convert them into the input data representation required by the structural displacement response prediction model specifically includes: The discretized node and unit information is converted into the input data representation required by the structural displacement response prediction model and used as the structural data matrix to be input; The input data of the structure is standardized; wherein, when the boundary conditions in the input data of the structure are standardized, an approximate distance function is used for processing; The geometric shape and size, material property parameters and boundary conditions in the standardized input data are mapped to different channels of the structure data matrix to be input, so as to obtain the preprocessed structure data matrix to be input.
5. The structural displacement response prediction method based on convolutional neural network according to claim 1, characterized in that: The step (3) specifically includes: First, a convolutional neural network is selected to process the structural data matrix and extract structural features; Then, according to the requirements and structural characteristics of the structural problem, the specific structure of the convolutional neural network is configured and adjusted, including: selecting combinations of convolutional layers, pooling layers, fully connected layers, and activation functions of different types, numbers, and connection orders.
6. The structural displacement response prediction method based on convolutional neural network according to claim 1 is characterized in that: The step (5) specifically includes: First, the weights and bias parameters of the structural displacement response prediction model are initialized; then, the input samples in the training sample set are used as the input data of the structural displacement response prediction model to obtain the predicted structural displacement response output by the structural displacement response prediction model; the loss function of the structural displacement response prediction model is calculated according to the predicted structural displacement response and the target output data in the training sample set, and the weights and bias parameters of the structural displacement response prediction model are optimized and updated by the back propagation method with the minimization of the loss function of the structural displacement response prediction model as the optimization goal, until the preset training rounds are reached, and the trained structural displacement response prediction model is obtained; wherein, the loss function of the structural displacement response prediction model uses the mean square error or the mean absolute error.
7. A structural displacement response prediction device based on convolutional neural network, comprising one or more processors and a memory, characterized in that: The memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the structural displacement response prediction method based on convolutional neural network described in any one of claims 1-6.
8. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, it is used to implement the structural displacement response prediction method based on a convolutional neural network as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Convolution prediction method for impact displacement time history diagram of concrete-filled steel tube composite member
CN116629080A
Wind-bridge system buffeting response prediction method based on lightweight Transform
CN117910120A
Convolutional neural network optimization method and system based on sparse regularization theory
CN118114734A
System and methods employing prefabricated volumetric construction modules including transforming truss elements
US11802402B1
