AI-based methods, equipment, and storage media for deploying seismic-resistant components

Abstract

The present application relates to artificial intelligence technology, and discloses an artificial intelligence-based seismic component deployment method, device and storage medium, including: obtaining a building architecture diagram; using a graph neural network pre-trained by a pre-trained model to analyze seismic stress points on the building architecture diagram, and obtaining stress analysis results corresponding to each building area; the stress analysis results include stress point distribution information, and seismic attribute indicators of each stress point; based on the stress analysis results, using the post-network of the pre-trained model, deploy corresponding seismic components for each building area, and combine them into a seismic system; the post-network has seismic components adapted to various seismic attributes, and pre-trains the first association features between various seismic components and corresponding stress points, and the second association features between the combination of multiple seismic components and corresponding stress points. The present application also discloses a control device. The present application aims to improve the efficiency and reliability of seismic system design for buildings.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based earthquake-resistant component deployment method, control device, computer equipment, and computer-readable storage medium. Background Art

[0002] Due to the unpredictability of earthquakes, in order to reduce the damage to buildings caused by earthquakes, it is necessary to add earthquake protection mechanisms to buildings (especially medium and large buildings). At present, the design of earthquake-resistant systems for buildings is generally done by engineers based on relevant earthquake-resistant design specifications and industry experience, and they artificially plan and configure corresponding earthquake-resistant components (such as supporting structures, shock-absorbing devices, reinforcement materials, etc.) for each key point of the building to achieve the purpose of building earthquake resistance.

[0003] However, the traditional deployment method of building earthquake-resistant components is often limited by the personal experience and intuition of engineers, and these experiences may not fully take into account the specific conditions and complexity of the building, making it difficult to make more accurate and effective deployment decisions. In addition, the manual deployment method also requires a lot of design time and human resources, especially when designing earthquake-resistant deployment plans for medium-sized and large buildings or building complexes. Not only is the design workload huge, but it often requires the collaboration of multiple people, making it even more difficult to ensure the effectiveness and reliability of the deployment.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present application is to provide an artificial intelligence-based earthquake-resistant component deployment method, control device, computer equipment and computer-readable storage medium, aiming to improve the efficiency and reliability of earthquake-resistant systems designed for buildings.

[0006] To achieve the above objectives, the present application provides an artificial intelligence-based earthquake-resistant component deployment method, comprising the following steps:

[0007] Obtaining a building structure diagram, wherein the building structure diagram is marked with material types of relevant building structures;

[0008] Performing seismic stress point analysis on the building structure diagram using a graph neural network pre-trained by a pre-trained model to obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point;

[0009] Based on the force analysis results, the post-network of the pre-trained model is used to deploy corresponding earthquake-resistant components for each building area and combine them into an earthquake-resistant system; wherein the post-network is loaded with earthquake-resistant components that are adapted to various earthquake-resistant properties, and pre-trains the first association features between various earthquake-resistant components and corresponding force points, as well as the second association features between the combination of multiple earthquake-resistant components and corresponding force points.

[0010] Optionally, the step of performing seismic stress point analysis on the building structure diagram using a graph neural network pre-positioned with a pre-trained model to obtain stress analysis results corresponding to each building area includes:

[0011] Extracting graphic data related to each building area from the building architecture diagram, the graphic data including node features and edge features;

[0012] Inputting the graphic data related to each building area into the graph neural network in front of the pre-trained model to perform seismic stress point analysis to generate stress analysis results corresponding to each building area;

[0013] The graph neural network is pre-trained and learns a third correlation feature between the building graphic data and the corresponding force analysis results.

[0014] Optionally, the post-network includes an input layer, an intermediate layer and an output layer;

[0015] The input layer is used to extract feature vectors related to the force analysis results;

[0016] A multilayer perceptron structure with multiple hidden layers is designed in the middle layer for learning and applying the first association feature and the second association feature respectively;

[0017] The output layer is used to generate a deployment plan for earthquake-resistant components and to generate an earthquake-resistant system based on a combination of the deployed earthquake-resistant components.

[0018] Optionally, before the step of using the graph neural network with a pre-trained model to perform seismic stress point analysis on the building structure diagram to obtain stress analysis results corresponding to each building area, the step further includes:

[0019] Obtaining the seismic requirement level corresponding to the building structure diagram;

[0020] Adjust relevant network parameters of the graph neural network according to the seismic resistance requirement level;

[0021] Among them, the higher the earthquake resistance requirement level, the greater the depth and width of the adjusted graph neural network.

[0022] Optionally, the step of obtaining the seismic requirement level corresponding to the building architecture diagram includes:

[0023] The corresponding seismic demand level is determined based on the building strength information, design service life and local historical earthquake data corresponding to the building structure diagram.

[0024] Optionally, after the step of deploying corresponding earthquake-resistant components for each building area based on the force analysis result and combining them into an earthquake-resistant system by using the post-network of the pre-trained model, the method further includes:

[0025] The generated earthquake-resistant system is exported to the CAD (Computer Aided Design) software for display, and the number of consumables required for the earthquake-resistant system is counted in the CAD software.

[0026] Optionally, before the step of counting the number of consumables required for the earthquake-resistant system in the CAD software, the step further includes:

[0027] Objects corresponding to various types of earthquake-resistant components are created in the CAD software, and the corresponding consumables attributes are associated, wherein the consumables attributes at least include part type and part material.

[0028] To achieve the above object, the present application also provides a control device, comprising:

[0029] An acquisition module, used to acquire a building structure diagram, wherein the building structure diagram is marked with a material type of a related building structure;

[0030] A processing module, configured to perform seismic stress point analysis on the building structure diagram using a graph neural network pre-trained by a pre-trained model, and obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point;

[0031] The processing module is also used to deploy corresponding earthquake-resistant components for each building area based on the force analysis results and combine them into an earthquake-resistant system using the post-network of the pre-trained model; wherein the post-network is loaded with earthquake-resistant components that are adapted to various earthquake-resistant properties, and pre-trains the first association features between various earthquake-resistant components and corresponding force points, as well as the second association features between the combination of multiple earthquake-resistant components and the corresponding force points.

[0032] To achieve the above objectives, the present application also provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the above-mentioned artificial intelligence-based earthquake-resistant component deployment method are implemented.

[0033] To achieve the above objectives, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned artificial intelligence-based earthquake-resistant component deployment method are implemented.

[0034] The artificial intelligence-based seismic component deployment method, control device, computer equipment and computer-readable storage medium provided in this application are based on artificial intelligence technology to intelligently analyze the building structure and stress conditions, automatically deploy seismic components customized for each building area, and generate a corresponding seismic system. This can not only improve the design efficiency of the seismic system, save design time and human resources, but also reduce the impact of human factors on deployment decisions to a certain extent, and increase the reliability and stability of the deployed seismic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of the steps of a method for deploying earthquake-resistant components based on artificial intelligence in one embodiment of the present application;

[0036] Figure 2 This is a schematic diagram of the architecture of a pre-training model in one embodiment of the present application;

[0037] Figure 3 This is a schematic diagram of the steps of a method for deploying earthquake-resistant components based on artificial intelligence in another embodiment of the present application;

[0038] Figure 4 This is a schematic diagram of a control device in an embodiment of the present application;

[0039] Figure 5 A schematic diagram of the internal structure of a computer device according to an embodiment of the present application.

[0040] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0041] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0042] In addition, if the descriptions of "first", "second", etc. are involved in this application, they are only used for descriptive purposes (such as for distinguishing the same or similar features), and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0043] Reference Figure 1 In one embodiment, the method for deploying earthquake-resistant components based on artificial intelligence includes:

[0044] Step S10: Obtain a building structure diagram, wherein the building structure diagram is marked with material types of relevant building structures;

[0045] Step S20, using the graph neural network pre-trained by the pre-trained model to perform seismic stress point analysis on the building structure diagram, and obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point;

[0046] Step S30: Based on the force analysis results, the post-network of the pre-trained model is used to deploy corresponding earthquake-resistant components for each building area and combine them into an earthquake-resistant system; wherein the post-network is loaded with earthquake-resistant components that are adapted to various earthquake-resistant properties, and pre-trains the first association features between various earthquake-resistant components and corresponding force points, as well as the second association features between the combination of multiple earthquake-resistant components and the corresponding force points.

[0047] In this embodiment, the execution terminal of the embodiment may be a computer device (such as a computer, a cloud server, etc.), or may be other equipment or devices (such as a control device) that controls the computer device.

[0048] As described in step S10, the building structure diagram can be obtained through various means, including existing architectural design drawings, CAD files, point cloud data obtained by laser scanning technology, etc. It is sufficient to ensure that the obtained structure diagram has sufficient clarity and accuracy to support subsequent analysis and processing (for example, it is converted based on the CAD drawing corresponding to the building structure diagram).

[0049] For the material types of building structures marked in building structure drawings, manual marking or semi-automatic marking can be used. For manual marking, professional engineers can manually mark the material type of each component based on design planning and field investigation. For semi-automatic marking, computer vision technology and image processing algorithms can be used to automatically identify and mark the building components and material types in the structure drawings.

[0050] In addition to marking the material type, you can further describe the properties of each material, such as density, elastic modulus, tensile strength, compressive strength, etc. This property information will help in subsequent force analysis and selection of earthquake-resistant components.

[0051] As described in step S20, an artificial intelligence model (i.e., a pre-trained model) for deploying earthquake-resistant components is pre-trained. Figure 2 ,The pre-trained model consists of two parts, the front network and the back network. Among them, the front network part of the ,pretrained model is responsible for the seismic stress point analysis of the building ,structure diagram.

[0052] Optionally, a front network is constructed based on a graph neural network (GNN), that is, a suitable graph neural network is selected and pre-trained to adapt to the task of seismic stress point analysis.

[0053] Optionally, the design strategy for the graph neural network used to build the front network is as follows:

[0054] (1) Network architecture design: Select a suitable graph neural network architecture, such as Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), and GraphSAGE framework (this can be done based on the requirements of the task and the characteristics of the architectural structure data);

[0055] (2) Input feature design: When designing a network, the input features include the graph data of the building structure, where the graph data includes node features and edge features (node ​​features represent the structural components in each building area and may include information such as the material, location, and size of the structural components; edge features represent the connections between the structural components in each building area and may include information such as the connection type and length);

[0056] (3) Output feature design: The network output includes the force analysis results of the building structure (such as the force value, force direction, seismic property index, etc. of each node). These output features will be used to guide the subsequent deployment of seismic components.

[0057] (4) Design of number of layers and activation function: Select the appropriate number of layers according to the complexity of the task, and the activation function can use nonlinear functions such as ReLU and Leaky ReLU;

[0058] (5) Loss function design: Select an appropriate loss function to evaluate the prediction effect of the model. This can use mean square error or cross entropy loss to measure the gap between the predicted value and the true value;

[0059] (6) Regularization design: Consider adding regularization methods, such as L2 regularization or Dropout, to prevent overfitting and improve the generalization ability of the model;

[0060] (7) Training strategy design: Select a suitable optimizer and learning rate, and adjust these hyperparameters based on the training results, such as using optimizers such as Adam (Adaptive Moment Estimation optimizer) and SGD (Stochastic Gradient Descent).

[0061] Graph neural networks can effectively learn and represent complex graph structure data, including nodes (components) and edges (connection relationships) in building structure diagrams, which enables the front-end network to better understand the topological relationships and characteristics of building structures. By combining graph neural networks with seismic stress point analysis tasks, an end-to-end learning process can be achieved, that is, a complete learning and inference process from raw input data to final output results, which can simplify the design and deployment process of the entire system and improve the overall efficiency of the system. In addition, graph neural networks have strong adaptability and can adapt to building structure diagram data of different scales and complexities. Whether it is a small residential building or a large commercial building, the front-end network can perform effective seismic stress point analysis. In addition, graph neural networks can effectively extract and learn feature information in structural diagrams, including node features and graph structure features. These features are very important for seismic stress point analysis and can help accurately identify stress points and perform effective stress analysis.

[0062] Optionally, an example of the pre-training steps for the front network is as follows:

[0063] (1) Data collection: First, a large amount of building structure data needs to be collected, which can include CAD drawings of the building, structural parameters (including the material type of the building structure), geographic location information, etc. This data can cover buildings of different types, sizes, and geographic locations;

[0064] (2) Data preprocessing: preprocess the collected data, including cleaning the data, removing noise, standardizing the data format, etc., to ensure the quality and consistency of the data;

[0065] (3) Data annotation: annotate the collected building structure data, that is, mark the stress conditions and seismic property indicators of each building area. This can be done by human experts or automatic annotation tools based on existing knowledge bases.

[0066] (4) Model training: Use the labeled data set to train the front-end network. During the training process, the network parameters can be continuously adjusted through the back propagation algorithm and optimizer so that the network can accurately predict the stress conditions of various areas in the building structure.

[0067] (5) Model evaluation: After training, the model can be evaluated to ensure its accuracy and generalization ability. This can be done using techniques such as cross-validation to evaluate the model's performance on unseen data.

[0068] (6) Model optimization: Based on the evaluation results, the model is tuned and optimized, including adjusting the network structure, improving the training algorithm, increasing training data, etc., to improve the performance and stability of the model.

[0069] In this way, an accurate and reliable front network can be pre-trained for seismic stress point analysis of building structures.

[0070] Optionally, after obtaining the building structure diagram corresponding to the building to be analyzed, the graph neural network in the pre-trained model is used to process the building structure diagram, identify the stress points and accurately locate them, and obtain the distribution information of the stress points. These stress points can be structural connection points, support nodes, load concentration points, etc., which are of great significance for subsequent seismic analysis.

[0071] Furthermore, the force information of each stress point is extracted, including the force magnitude, direction, frequency, etc. At the same time, according to the structural position and material type of the stress point, the seismic property index of each stress point is calculated, such as mass, stiffness, seismic resistance level, etc. Then, according to the distribution of stress points in the building structure (including dense areas, high-load areas, vulnerable areas, etc.), the force analysis results corresponding to each building area are divided.

[0072] It should be noted that seismic property indices are quantitative indicators used to describe the seismic performance and characteristics of building structures under earthquakes. These indices are usually calculated and evaluated based on factors such as the geometric shape, material properties, structural stiffness, and mass distribution of the structure, which helps to understand the seismic capacity of the building structure and formulate appropriate seismic design plans accordingly.

[0073] In this way, the graph neural network in front of the pre-trained model can be effectively used to analyze the seismic stress points of the building structure diagram, and the corresponding stress analysis results of each building area can be obtained, including stress point distribution information and seismic attribute indicators. This provides important basic data and guidance for the subsequent deployment of seismic components and system design.

[0074] As described in step S30, the post-network part of the pre-trained model is responsible for the deployment of earthquake-resistant components and the generation of a corresponding earthquake-resistant system.

[0075] Among them, the post-network has inputs of seismic components that are adapted to various seismic properties, and for each type of seismic component, extracts features related to seismic properties as needed. These features may include information such as the geometry, material properties, stiffness, mass, load-bearing capacity of the seismic component, as well as features related to its function and role.

[0076] For example, according to the corresponding earthquake-resistant property indicators, select suitable earthquake-resistant components. These components may include but are not limited to:

[0077] Shock absorbers: such as hydraulic shock absorbers and friction shock absorbers, used to reduce the impact of seismic loads on structures;

[0078] Reinforcement components: such as reinforcing steel bars, steel plates, brackets, etc., used to enhance the bearing capacity of the structure;

[0079] Seismic isolation system: such as seismic isolation rubber, spring isolators, etc., used to isolate the transmission of earthquake energy to the structure.

[0080] Optionally, the post-network is pre-trained and learns the first association features between various types of seismic components and corresponding stress points (including seismic property indicators related to the stress points). This requires collecting and preparing a data set containing various types of seismic components, as well as performance data of these components under stress conditions. These data may include information such as the geometry, material properties, strength parameters, etc. of different types of seismic components, as well as the location, load-bearing conditions, seismic property indicators, etc. of the corresponding stress points. Multiple iterative training is performed based on the prepared data set to learn and extract the first association features between the seismic components and the corresponding stress points (one seismic component can be adapted to at least one stress point). The first association features may include properties such as deformation, displacement, and stress distribution of the seismic components under stress conditions, as well as seismic property indicators related to the stress points, such as acceleration and displacement response of the stress points.

[0081] Optionally, the post-network realizes multi-task processing based on a multi-layer perceptron structure, and while training and learning the first associated features, pre-training of the second associated features can be performed for various combinations of earthquake-resistant components.

[0082] The pre-training process of the second associated feature can consider the stress conditions under different combinations (including a combination of multiple seismic components that can adapt to one or more stress points; each combination includes multiple seismic components of the same or different types), and train the post-network to learn the synergy between multiple seismic components and the associated features between the corresponding stress points as the second associated features.

[0083] Among them, the training and learning of the second associated features can be carried out based on the extended data set. The extended data set contains various combinations of seismic components and their performance data under different stress conditions, which involves different configurations of various combinations of seismic components and their responses under various stress points (single or multiple) (these stress points all have their corresponding seismic attribute indicators). Based on the extended data set, the second associated features are learned and extracted, which includes the correlation between the combination of various seismic components and at least one stress point, such as the interaction between seismic components, synergy effect and response under complex stress environment.

[0084] It should be noted that the combination of earthquake-resistant components can be selected according to the specific situation, either directly or through connecting components, to meet the earthquake-resistant requirements of different building structures. Direct combination means that different earthquake-resistant components can be directly spliced ​​after being placed in the position that needs to be reinforced or supported, so as to enhance the overall earthquake resistance of the building. Combination through connecting components means that different earthquake-resistant components are connected through connectors (such as brackets) to form an overall earthquake-resistant system, which can better coordinate the functions between the various components and improve the overall earthquake-resistant effect.

[0085] Compared with the first association feature, the second association feature can provide a more comprehensive understanding of the overall performance of the earthquake-resistant system. The learning of the second association feature enables the pre-trained model to better understand the interaction between multiple earthquake-resistant components, thereby optimizing the design of the earthquake-resistant system. By considering the synergy between the various components, the most effective combination of earthquake-resistant components can be selected to improve the overall earthquake-resistant performance. And because the second association feature takes into account the relationship between multiple earthquake-resistant components, the position and configuration of each component can be determined more accurately to avoid excessive use of earthquake-resistant consumables, which helps save costs and reduce waste.

[0086] Optionally, after using the front-end graph neural network to obtain the force analysis results for the building structure diagram, the back-end network is used to determine the specific types of earthquake-resistant components required for each building area, such as shock absorbers, reinforcement components, isolation systems, etc., based on the force analysis results. The back-end network intelligently selects and places appropriate earthquake-resistant components based on the pre-learned first association features and force analysis results. This includes determining the location, quantity, and connection method of each component to obtain preliminary deployment results.

[0087] Optionally, the post-network can then intelligently combine the seismic components based on the preliminary deployment results according to the pre-learned second association features, and adjust the deployment positions, quantities and types of the seismic components to improve the overall seismic resistance effect, optimize the deployment results, and obtain the deployment results of the seismic components corresponding to each building area.

[0088] Finally, based on the deployment results after post-network optimization, the selected earthquake-resistant components are combined into a complete earthquake-resistant system. This process ensures that the connections between the various components are firm and that they can work together effectively to improve the overall earthquake resistance.

[0089] In one embodiment, based on artificial intelligence technology, intelligent analysis of building structures and stress conditions is achieved, earthquake-resistant components are automatically deployed for each building area in a customized manner, and a corresponding earthquake-resistant system is generated. This can not only improve the design efficiency of the earthquake-resistant system, save design time and human resources, but also reduce the impact of human factors on deployment decisions to a certain extent, and increase the reliability and stability of the deployed earthquake-resistant system.

[0090] In one embodiment, based on the above embodiment, the step of using the graph neural network with a pre-trained model to perform seismic stress point analysis on the building structure diagram to obtain stress analysis results corresponding to each building area includes:

[0091] Extracting graphic data related to each building area from the building architecture diagram, the graphic data including node features and edge features;

[0092] Inputting the graphic data related to each building area into the graph neural network in front of the pre-trained model to perform seismic stress point analysis to generate stress analysis results corresponding to each building area;

[0093] The graph neural network is pre-trained and learns a third correlation feature between the building graphic data and the corresponding force analysis results.

[0094] In this embodiment, graphic data related to each building area is extracted from the building structure diagram, and these data include node features and edge features. Among them, the node features represent the structural components in each building area, and may include information such as the material, location, and size of the structural components; the edge features represent the connection between the structural components in each building area, and may include information such as the connection type and length.

[0095] Among them, the building area can be divided according to floors, functional areas, buildings, etc. For example, dividing by floors can give a clearer understanding of the stress conditions of each floor, dividing by functional areas can better understand the stress characteristics of different functional areas, and dividing by buildings can better consider the integrity and connectivity of the building structure. Taking these factors into consideration, the seismic resistance of the building structure can be more comprehensively evaluated, providing more targeted solutions for subsequent seismic design and reinforcement.

[0096] The extracted graphic data is input into the graph neural network in front of the pre-trained model for seismic stress point analysis. In the pre-trained model, the graph neural network has been pre-trained to learn the third correlation feature between the building graphic data and the corresponding force analysis results. Therefore, based on the third correlation feature and combined with the input graphic data, the graph neural network will analyze and obtain the corresponding force analysis results.

[0097] The purpose of learning the third correlation feature is to better understand and model the complex relationship between data and improve the accuracy and generalization of the task. By learning the third correlation feature, the complex relationship between the morphological characteristics of the building structure and its stress conditions can be taken into account in the model, thereby improving the model's ability to model the building structure. By considering the relationship between the building morphology and the stress, the model can understand the characteristics of the building structure more accurately, thereby improving the accuracy of the prediction.

[0098] Specifically, the graph neural network will first identify the stress points in each building area and accurately locate them (i.e., obtain the distribution information of the stress points), and extract information such as the magnitude, direction, and frequency of the stress from the identified stress points. According to the structural position and material type of the stress points, the seismic property indicators of each stress point are calculated, such as mass, stiffness, and seismic resistance level. These indicators can reflect the response ability and seismic performance of the stress points to earthquakes (i.e., obtain the seismic property indicators of the stress points).

[0099] In one embodiment, multiple aspects of building structure information (including node features and edge features) are combined to achieve a more comprehensive and accurate analysis of seismic stress points of building structures. Using a graph neural network with a pre-trained model to analyze seismic stress points can fully utilize the advantages of deep learning in graph data processing, improve the ability to understand and analyze complex structures, and improve analysis efficiency and accuracy.

[0100] In one embodiment, based on the above embodiment, referring to Figure 2 , the post-network includes an input layer, an intermediate layer and an output layer;

[0101] The input layer is used to extract feature vectors related to the force analysis results;

[0102] A multilayer perceptron structure with multiple hidden layers is designed in the middle layer for learning and applying the first association feature and the second association feature respectively;

[0103] The output layer is used to generate a deployment plan for earthquake-resistant components and to generate an earthquake-resistant system based on a combination of the deployed earthquake-resistant components.

[0104] In this embodiment, the structure of the post-network includes an input layer, an intermediate layer, and an output layer, and each layer has its specific function and role:

[0105] The input layer is used to receive and extract feature vectors related to the stress analysis results. These feature vectors include the location information of the stress point, stress conditions, seismic property indicators, etc., to provide a basis for subsequent feature learning.

[0106] The middle layer adopts a multi-layer perceptron structure with multiple hidden layers. These hidden layers are used to learn and apply the first association feature and the second association feature. The hidden layer here can effectively learn the first association feature between the seismic components and the stress points in the force analysis results, as well as the second association feature between the combination of multiple seismic components and the corresponding stress points, so as to provide a more accurate deployment plan for the output layer.

[0107] Among them, each hidden layer converts the input signal into an output signal through an activation function, and is connected and adjusted through weights and biases to gradually extract and combine feature information.

[0108] The feature vector extracted from the input layer will be transformed linearly and nonlinearly through the hidden layer multiple times and propagated to the output layer layer by layer. In this process, the neurons in the hidden layer will learn the high-order feature representation in the data and pass it to the next layer. Each layer of neurons in the hidden layer will learn different abstract features in the data, including first-association features and second-association features. This is achieved through the network's back-propagation algorithm, which adjusts the network parameters by minimizing the error between the predicted value and the true value, so that the network can better fit the training data. The learned association features will be applied to the output layer to perform related tasks.

[0109] It should be noted that the multilayer perceptron has multiple nonlinear hidden layers, which can learn and model complex nonlinear relationships in input data. In the design of earthquake-resistant systems, there are complex nonlinear relationships between the attributes of the stress points, location information, and the associations between earthquake-resistant components, and the multilayer perceptron can effectively fit these relationships. In addition, the multilayer perceptron can adapt to different problems and data sets by adjusting the number of hidden layers and the connection mode of neurons. In the design of earthquake-resistant systems, considering that there may be different types of combinations of earthquake-resistant components and stress points, a multilayer perceptron structure is needed to adapt to this diversity.

[0110] The output layer is used to generate the deployment plan of earthquake-resistant components and the design plan of earthquake-resistant systems. In the output layer, the model will comprehensively consider the feature vectors corresponding to the force analysis results provided by the input layer, as well as the first and second associated features learned by the intermediate layer, to generate the final deployment plan of earthquake-resistant components. This plan includes determining the type, quantity, location and connection method of each component, and how to combine these components to form a complete earthquake-resistant system.

[0111] During the training of the post-network, appropriate loss functions can also be designed, including classification loss functions and combination loss functions. The classification loss function can use the cross entropy loss function to optimize the selection probability of earthquake-resistant components. The combination loss function can use a multi-task learning method to learn a better combination of earthquake-resistant components by jointly optimizing multiple tasks, while optimizing the accuracy and earthquake-resistant performance of the combination.

[0112] These loss functions can effectively guide the model to learn better feature representations and deployment schemes, improving the model's generalization and anti-interference capabilities.

[0113] In summary, the post-network will first intelligently select and place appropriate earthquake-resistant components (including determining the location, quantity, and connection method of each component) based on the pre-learned first association features and force analysis results to obtain preliminary deployment results. Then, based on the preliminary deployment results, the post-network will intelligently combine the earthquake-resistant components based on the pre-learned second association features and adjust the deployment location, quantity, and type of the earthquake-resistant components to improve the overall earthquake-resistant effect, optimize the deployment results, and obtain the deployment results of the earthquake-resistant components corresponding to each building area. Finally, according to the deployment results optimized by the post-network, the selected earthquake-resistant components are combined into a complete earthquake-resistant system.

[0114] In some optional embodiments, the optimized deployment results of the earthquake-resistant components can also be associated with the force analysis results and fed back to the input layer of the post-network, and the corresponding feature vectors are extracted by the input layer and input into the middle layer to strengthen and optimize the learning of the first associated features. In this way, when the subsequent pre-trained model deploys the corresponding earthquake-resistant components for the new building architecture diagram, the preliminary deployment results of the earthquake-resistant components can be obtained more efficiently and accurately.

[0115] In one embodiment, the post-network uses a multi-layer perceptron structure with multiple hidden layers, which has strong expressive power and can learn and model complex nonlinear relationships in input data. In the design of earthquake-resistant systems, there are often complex nonlinear relationships between the attributes of the stress points, location information, and the associations between earthquake-resistant components. The multi-layer perceptron can effectively fit these relationships and improve the expressive power of the model. Moreover, the feature learning ability of the hidden layer enables the model to understand the input data more accurately and provide a more accurate deployment plan for the output layer, thereby improving the predictive power and accuracy of the model.

[0116] Moreover, the post-network adopts an end-to-end training method, and adjusts the network parameters through the back-propagation algorithm, so that the network can better fit the training data. This end-to-end training method can directly optimize the performance of the model on the task, simplify the model training process, and better adapt to different earthquake-resistant system design tasks.

[0117] In one embodiment, based on the above embodiment, before the step of using the graph neural network with a pre-trained model to perform seismic stress point analysis on the building structure diagram to obtain the stress analysis results corresponding to each building area, the step further includes:

[0118] Obtaining the seismic requirement level corresponding to the building structure diagram;

[0119] Adjust relevant network parameters of the graph neural network according to the seismic resistance requirement level;

[0120] Among them, the higher the earthquake resistance requirement level, the greater the depth and width of the adjusted graph neural network.

[0121] In this embodiment, before analyzing the stress points of the building, the seismic demand level of the building corresponding to the building structure diagram must first be obtained. This usually depends on factors such as the building's geographical location, building type, building height, and purpose of use. The higher the seismic demand level, the greater the earthquake risk in the area where the building is located, or the building itself requires a higher seismic resistance. For example, the seismic demand level is divided into three levels: high, medium, and low.

[0122] Optionally, according to the obtained earthquake resistance requirement level, relevant network parameters of the graph neural network are adjusted. The direction of network parameter adjustment may include: network depth and network width.

[0123] The corresponding network depth and width can be set for high, medium and low earthquake resistance demand levels, and the higher the earthquake resistance demand level, the greater the network depth and width of the graph neural network.

[0124] For example, for high seismic requirements, the depth of the graph neural network is increased to better capture the complex features and relationships in the building structure. For medium and low requirements, the depth of the network can be appropriately reduced to avoid overfitting or excessive computational burden.

[0125] For example, for high earthquake resistance requirements, the number of neurons in each layer is increased to improve the representation and computational capabilities of the network, while for medium and low requirements, the width of the network can be appropriately reduced to reduce computational complexity.

[0126] Optionally, while adjusting the network depth and width, the activation function and regularization method may be appropriately adjusted according to the earthquake resistance requirement level to ensure the stability and generalization ability of the network.

[0127] For example, for high seismic requirements, you can choose activation functions that are more suitable for complex data and deep networks, such as Leaky, ELU, etc., to ensure that the network can better capture the nonlinear relationship in the data. For medium and low requirements, commonly used activation functions (such as ReLU) are sufficient, and there is no need for overly complex activation functions.

[0128] Optionally, while adjusting the network depth and width, the learning rate of the model can also be adjusted according to the earthquake resistance requirement level, and the higher the earthquake resistance requirement level, the smaller the adjusted learning rate is, so as to ensure the stability and accuracy of the training.

[0129] In this way, it can ensure that the graph neural network can make corresponding adjustments according to the seismic demand level of the building when analyzing the seismic stress points, providing higher accuracy and flexibility. This method can also help the model maintain sufficient adaptability and effectiveness when dealing with buildings of different sizes and complexities.

[0130] In one embodiment, based on the above embodiment, the step of obtaining the seismic requirement level corresponding to the building structure diagram includes:

[0131] The corresponding seismic demand level is determined based on the building strength information, design service life and local historical earthquake data corresponding to the building structure diagram.

[0132] In this embodiment, the building strength information includes the building material, structure type (such as steel structure, concrete structure), building structure system (such as frame structure, shear wall structure), etc. This information can help evaluate the basic earthquake resistance of the building when facing an earthquake.

[0133] The design life of a building indicates the expected service life of the building. Generally speaking, the longer the design life of a building, the higher the earthquake design standard will be, because the structure needs to remain safe for a longer period of time.

[0134] Local historical earthquake data includes the intensity, frequency and damage of earthquakes in the area where the building is located. These data are crucial for assessing the earthquake risk of the region. In areas where there has been frequent earthquake activity or major earthquakes in the past, the seismic resistance requirements of buildings should be higher.

[0135] Combining these three aspects of information, the seismic demand level can be determined by following the steps below:

[0136] (1) Data collection and analysis: First, collect the above information, including specific building data, design documents, and historical earthquake records related to the geographical environment.

[0137] (2) Assess the potential risk of the building: Conduct a risk assessment based on the strength and service life of the building and local earthquake history data. This can be done through detailed calculations performed by structural engineers or using an already developed assessment model.

[0138] (3) Determine the seismic resistance requirement level: Based on the results of risk assessment, buildings are classified into high, medium, and low seismic resistance requirement levels.

[0139] The collected building strength information, design service life and local historical earthquake data are input into the seismic demand level assessment model pre-built using big data technology. Such a model usually uses a large amount of data and algorithms to assess the seismic demand level, and can more accurately consider the combined impact of multiple factors. The model will comprehensively consider these data according to different weights and algorithms, and give the corresponding seismic demand level.

[0140] This can provide a scientific basis for the subsequent parameter adjustment of the graph neural network, making the seismic stress point analysis more accurate and effective.

[0141] In one embodiment, based on the above embodiment, referring to Figure 3 , after the step of deploying corresponding earthquake-resistant components for each building area based on the force analysis result and combining them into an earthquake-resistant system by using the post-network of the pre-trained model, the method further includes:

[0142] Step S40: export the generated earthquake-resistant system to the CAD software for display, and count the number of consumables required for the earthquake-resistant system in the CAD software.

[0143] In this embodiment, an export module is designed to convert the generated earthquake-resistant system into a format that can be recognized by the CAD software, such as DXF, DWG, etc. According to the import requirements of the CAD software, the structure, component layout, connection method and other information of the earthquake-resistant system are imported into the CAD software for subsequent visualization and further design.

[0144] Optionally, the imported seismic system data can be loaded into the CAD software to ensure that the position and connection of each component are displayed accurately. At the same time, the seismic system can be presented in three dimensions through the 3D modeling function of the CAD software, which is convenient for engineers and designers to view and adjust.

[0145] Optionally, based on the seismic system displayed in the CAD software, the number of various seismic components can be counted using the measurement tools or plug-ins provided by the CAD software. This includes information such as the number, length, and area of ​​materials required for various seismic components (such as support structures, reinforcement materials, steel pipe length, etc.), and a detailed list of consumables can be generated.

[0146] By exporting the earthquake-resistant system to CAD software to display and count the number of consumables required, visual management and material estimation of earthquake-resistant engineering design can be achieved, providing an important reference for subsequent construction and material procurement. Such a process can improve the efficiency and accuracy of engineering design, and also help control engineering costs and resource investment.

[0147] In one embodiment, based on the above embodiment, before the step of counting the number of consumables required for the earthquake-resistant system in the CAD software, the step further includes:

[0148] Objects corresponding to various types of earthquake-resistant components are created in the CAD software, and the corresponding consumables attributes are associated, wherein the consumables attributes at least include part type and part material.

[0149] In this embodiment, corresponding graphic objects are created in the CAD software for various preset earthquake-resistant components, such as support structures, reinforcement materials, etc. At the same time, according to actual conditions, the drawing tools provided by the CAD software can also be used to draw the cross-section of the support structure, the shape of the reinforcement material, etc.

[0150] Optionally, for each seismic component object created, associate the corresponding consumable properties. These properties should at least include information such as part type and bracket material. This can be done through the property editing function of the CAD software or a custom property management tool to add attribute information related to each object.

[0151] Optionally, the part type can refer to the part type of the connecting bracket, or the component necessary to form the corresponding seismic component; in the CAD software, the part type is defined for each seismic component object so that subsequent design and statistical work can be carried out accurately. The selection of these part types can be based on the specific design and engineering requirements of the seismic system to ensure that the load-bearing requirements and construction feasibility of the structure are met. The part material specifies the specific materials used for each type of part, such as the material of the bracket is steel pipe.

[0152] In one embodiment, by creating a seismic component object in the CAD software and associating the corresponding consumables attributes, the necessary basic data can be provided for the subsequent consumables quantity statistics, which not only helps to ensure the accuracy of the consumables statistics, but also provides more comprehensive and reliable information support for engineering design.

[0153] In addition, refer to Figure 4 In an embodiment of the present application, a control device Z10 is further provided, comprising:

[0154] An acquisition module Z11 is used to acquire a building structure diagram, wherein the building structure diagram is marked with a material type of a relevant building structure;

[0155] The processing module Z12 is used to perform seismic stress point analysis on the building structure diagram using the graph neural network pre-trained by the pre-trained model to obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point;

[0156] The processing module Z12 is also used to deploy corresponding seismic components for each building area based on the force analysis results and combine them into a seismic system by using the post-network of the pre-trained model; wherein the post-network is input with seismic components that are adapted to various types of seismic properties, and pre-trains the first association features between various types of seismic components and corresponding force points, as well as the second association features between the combination of multiple seismic components and corresponding force points.

[0157] Optionally, the control device Z10 may be a virtual control device (such as a virtual machine) or a physical device (such as a physical device other than a computer device that can execute the corresponding method).

[0158] In addition, a computer device is also provided in an embodiment of the present application. The internal architecture of the computer device can be as follows: Figure 5 As shown, it includes a processor, a memory, a communication interface and an input interface connected by a system bus. Among them, the processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database is used to store data called by the computer program. The communication interface is used to communicate data with an external terminal. The input interface is used to receive a signal input by an external device. When the computer program is executed by the processor, a method for deploying earthquake-resistant components based on artificial intelligence as described in the above embodiment is implemented.

[0159] Those skilled in the art will understand that Figure 5The structure shown in the figure is only a block diagram of a part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied. For example, in some optional embodiments, the computer device may also include an output interface (not shown in the figure), and the output interface is also connected to the system bus and is used to output corresponding signals to the peripheral device.

[0160] In addition, the present application also proposes a computer-readable storage medium, the computer-readable storage medium including a computer program, and the computer program, when executed by a processor, implements the steps of the method for deploying earthquake-resistant components based on artificial intelligence as described in the above embodiment. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0161] In summary, the artificial intelligence-based seismic component deployment method, control device, computer equipment, and computer-readable storage medium provided in the embodiments of the present application, based on artificial intelligence technology, can intelligently analyze the building structure and stress conditions, automatically deploy seismic components customized for each building area, and generate a corresponding seismic system. This can not only improve the design efficiency of the seismic system, save design time and human resources, but also reduce the impact of human factors on deployment decisions to a certain extent, and increase the reliability and stability of the deployed seismic system.

[0162] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0163] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0164] The above description is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for deploying earthquake-resistant components based on artificial intelligence, characterized in that: include: Obtaining a building structure diagram, wherein the building structure diagram is marked with material types of relevant building structures; Performing seismic stress point analysis on the building structure diagram using a graph neural network pre-trained by a pre-trained model to obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point; Based on the force analysis results, the post-network of the pre-trained model is used to deploy corresponding earthquake-resistant components for each building area and combine them into an earthquake-resistant system; wherein the post-network is loaded with earthquake-resistant components adapted to various earthquake-resistant properties, and the first association features between various earthquake-resistant components and corresponding force points, and the second association features between the combination of multiple earthquake-resistant components and corresponding force points are pre-trained; Among them, the post-network performs pre-training of the second associated feature for the combination of multiple seismic components; the pre-training process of the second associated feature takes into account the stress conditions under different combinations, and trains the post-network to learn the synergy between the multiple seismic components and the associated features between the corresponding stress points as the second associated features; the training and learning of the second associated features are based on an extended data set, which contains the combination of multiple seismic components and performance data under different stress conditions, involving different configurations of the combination of multiple seismic components, and the response of the seismic component combination under a single or multiple stress points; learning and extracting the second associated features based on the extended data set includes learning and extracting the correlation between the combination of multiple seismic components and at least one stress point, so as to learn and extract the interaction, synergy and response to the stress points between the seismic components; The steps of deploying corresponding earthquake-resistant components for each building area and combining them into an earthquake-resistant system based on the force analysis results and utilizing the post-network of the pre-trained model include: Using the post-network, the type of earthquake-resistant components required for each building area is determined according to the force analysis results; based on the first association feature and the force analysis results, appropriate earthquake-resistant components are selected and placed, including determining the location, quantity and connection method of each earthquake-resistant component, and obtaining a preliminary deployment result; Based on the second association feature, on the basis of the preliminary deployment results, the synergy between the various seismic resistant components is considered, the seismic resistant components are combined, and the deployment positions, quantities and types of the seismic resistant components are adjusted to optimize the deployment results, and the deployment results of the seismic resistant components corresponding to each building area are obtained; according to the optimized deployment results, the selected seismic resistant components are combined into a complete seismic resistant system.

2. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 1, characterized in that: The step of using the graph neural network with a pre-trained model to analyze the seismic stress points of the building structure diagram to obtain the stress analysis results corresponding to each building area includes: Extracting graphic data related to each building area from the building architecture diagram, the graphic data including node features and edge features; Input the graphic data related to each building area into the graph neural network in front of the pre-trained model to perform seismic stress point analysis to generate stress analysis results corresponding to each building area; The graph neural network is pre-trained and learns a third correlation feature between the building graphic data and the corresponding force analysis results.

3. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 1, characterized in that: The post-network includes an input layer, an intermediate layer and an output layer; The input layer is used to extract feature vectors related to the force analysis results; A multilayer perceptron structure with multiple hidden layers is designed in the middle layer for learning and applying the first association feature and the second association feature respectively; The output layer is used to generate a deployment plan for earthquake-resistant components and to generate an earthquake-resistant system based on a combination of the deployed earthquake-resistant components.

4. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 1, characterized in that: Before the step of using the graph neural network with a pre-trained model to analyze the seismic stress points of the building structure diagram to obtain the stress analysis results corresponding to each building area, the method further includes: Obtaining the seismic requirement level corresponding to the building structure diagram; Adjust relevant network parameters of the graph neural network according to the seismic resistance requirement level; Among them, the higher the earthquake resistance requirement level, the greater the depth and width of the adjusted graph neural network.

5. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 4, characterized in that: The step of obtaining the seismic requirement level corresponding to the building structure diagram comprises: The corresponding seismic demand level is determined based on the building strength information, design service life and local historical earthquake data corresponding to the building structure diagram.

6. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 1, characterized in that: After the step of deploying corresponding earthquake-resistant components for each building area based on the force analysis result and combining them into an earthquake-resistant system by using the post-network of the pre-trained model, the method further includes: The generated earthquake-resistant system is exported to the CAD software for display, and the number of consumables required for the earthquake-resistant system is counted in the CAD software.

7. The method for deploying earthquake-resistant components based on artificial intelligence according to claim 6, characterized in that: Before the step of counting the number of consumables required for the earthquake-resistant system in the CAD software, the method further includes: Objects corresponding to various types of earthquake-resistant components are created in the CAD software, and the corresponding consumables attributes are associated, wherein the consumables attributes at least include part type and part material.

8. A control device, characterized in that: include: An acquisition module, used to acquire a building structure diagram, wherein the building structure diagram is marked with a material type of a related building structure; A processing module, configured to perform seismic stress point analysis on the building structure diagram using a graph neural network pre-trained by a pre-trained model, and obtain stress analysis results corresponding to each building area; wherein the stress analysis results include stress point distribution information and seismic property indicators of each stress point; The processing module is also used to deploy corresponding earthquake-resistant components for each building area based on the force analysis results and combine them into an earthquake-resistant system using the post-network of the pre-trained model; wherein the post-network is loaded with earthquake-resistant components that are adapted to various earthquake-resistant properties, and pre-trains the first association features between various earthquake-resistant components and corresponding force points, as well as the second association features between the combination of multiple earthquake-resistant components and the corresponding force points; the post-network performs pre-training of the second association features for the combination of multiple earthquake-resistant components; the pre-training process of the second association features considers the force conditions under different combinations, and trains the post-network to learn the synergy between multiple earthquake-resistant components and the association features between the corresponding force points as the second association features; the training and learning of the second association features are based on an extended data set, which contains the combination of multiple earthquake-resistant components and performance data under different force conditions, involving different configurations of multiple combinations of earthquake-resistant components, and The method comprises the following steps: learning and extracting the second associated features based on the extended data set, including learning and extracting the association between the combination of multiple seismic components and at least one stress point, so as to learn and extract the interaction, synergy and response of the seismic components to the stress point; using the post-network, according to the stress analysis results, determining the type of seismic components required for each building area; selecting and placing appropriate seismic components based on the first associated features and the stress analysis results, including determining the position, quantity and connection method of each seismic component, and obtaining a preliminary deployment result; based on the second associated features, on the basis of the preliminary deployment result, considering the synergy between the seismic components, combining the seismic components, and adjusting the deployment position, quantity and type of the seismic components to optimize the deployment result, and obtaining the deployment result of the seismic components corresponding to each building area; according to the optimized deployment result, combining the selected seismic components into a complete seismic system.

9. A computer device, characterized in that: The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the artificial intelligence-based earthquake-resistant component deployment method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the artificial intelligence-based earthquake-resistant component deployment method according to any one of claims 1 to 7 are implemented.

Images