Intelligent prediction method for anti-seismic performance of reinforced concrete shear wall with optimized structure design

Abstract

The invention discloses an intelligent prediction method for the anti-seismic performance of a reinforced concrete shear wall for structural optimization design. The intelligent prediction method comprises the following steps: carrying out data collection on mechanical properties such as hysteretic performance, rigidity degradation and a skeleton curve of the shear wall by adopting a Python crawler technology and a low-cycle reciprocating load test of the shear wall; building a prediction model among the concrete strength grade, the geometric dimension of the shear wall, the reinforcement strength, the volume reinforcement ratio of the shear wall and the anti-seismic property by using a convolutional neural network deep learning model; training a convolutional neural network deep learning model according to the collected data; optimizing the convolutional neural network deep learning model, such as an early stop method, an Adam optimizer and an L2 regularization method; and predicting the anti-seismic performance of the reinforced concrete shear wall by using the trained convolutional neural network deep learning model, and comparing a prediction result with a test for verification.

Description

Technical Field

[0001] The present invention belongs to the technical field of seismic performance prediction of reinforced concrete shear walls, and in particular relates to an intelligent prediction method for seismic performance of reinforced concrete shear walls used for structural optimization design. Background Art

[0002] The steel frame-infill reinforced concrete shear wall structural system features a high degree of assembly, good seismic performance, and flexible building layout. As a key lateral force-resisting component, reinforced concrete shear walls are considered to be structural components with good seismic performance. In exploring the seismic performance of reinforced concrete shear walls, the traditional method is to obtain the hysteresis curve of the shear wall through low-cycle cyclic reciprocating load tests, and then extract the macro-computational constitutive model, skeleton curve, stiffness degradation, energy dissipation capacity, and ductility coefficient. However, this method is not only inefficient, but also difficult to effectively control cost and reliability.

[0003] In the field of structural optimization design, the currently commonly used macro-computational constitutive model for reinforced concrete shear walls oversimplifies the shear wall stiffness changes according to the elastic and elastoplastic stages. This makes it impossible to track the nonlinear behavior of the shear wall in real time, making detailed analysis difficult. This results in designers lacking a full understanding of the structural system they are using, which not only leads to low economic benefits but also to irrational structural selection. Although researchers have conducted in-depth discussions on the elastic stage of this structural system, the lack of a refined macro-computational constitutive model for reinforced concrete shear walls has prevented detailed performance research and optimization design in the elastoplastic stage. The development and widespread application of artificial intelligence and deep learning technologies have provided a new approach and method for the design of macro-computational constitutive models for reinforced concrete shear walls.

[0004] In order to improve the efficiency, accuracy and economy of structural design, it is necessary to propose an intelligent prediction method for the seismic performance of reinforced concrete shear walls for structural optimization design. Summary of the Invention

[0005] In response to the problems existing in the existing technology, the present invention provides an intelligent prediction method for the seismic performance of reinforced concrete shear walls for structural optimization design. Through the convolutional neural network model, the intelligent prediction of the seismic performance of reinforced concrete shear walls is achieved, and the structural design is optimized.

[0006] The technical solution of the present invention is achieved as follows:

[0007] An intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design includes the following steps:

[0008] S1. Data preparation: Data collection is performed using crawler technology and experiments. The collected data includes obtaining low-cycle reciprocating load test data of reinforced concrete shear walls through crawler technology, and preparing reinforced concrete shear wall specimens in the laboratory and conducting low-cycle reciprocating load tests to obtain and collect test data; pre-processing and dividing the data into training, validation, and test sets;

[0009] S2. Construct a convolutional neural network model, wherein the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer;

[0010] S3. Use the data of the training set to train the constructed model until the predetermined convergence condition or number of iterations is reached;

[0011] S4. Optimize the convolutional neural network model and use cross-validation on the training set, validation set, and test set to evaluate the performance of the model;

[0012] S5. Use the trained convolutional neural network model to predict the seismic performance of reinforced concrete shear walls.

[0013] Furthermore, in said S1, the collected data include different concrete strength grades, geometric dimensions of shear walls, steel bar strength, volume reinforcement ratio and corresponding hysteresis curves, skeleton curves, stiffness degradation, energy dissipation capacity and ductility coefficient of the shear walls;

[0014] The concrete strength grade, geometric dimensions of the shear wall, steel bar strength, and volume reinforcement ratio are taken as characteristic data; the hysteresis curve, skeleton curve, stiffness degradation, energy dissipation capacity, and ductility coefficient of the shear wall are taken as seismic performance data.

[0015] Furthermore, the data is preprocessed, specifically including cleaning the collected data, removing invalid data, and then normalizing it. The Min-Max normalization method is used to convert the feature data to the range of [0,1]. The normalization formula is:

[0016]

[0017] Among them, x norm is the normalized data, x is the original data, and x max and x min are the maximum and minimum values of the feature data respectively;

[0018] The feature data is converted into a two-dimensional matrix and divided into training set, validation set and test set.

[0019] Furthermore, in S2, a convolutional neural network is used as the main architecture of the model, and a deep learning framework Tensorflow or PyTorch is used to construct a convolutional neural network model.

[0020] Furthermore, the convolutional neural network model includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer; the specific construction method includes:

[0021] S2.1. Input the feature data into the input layer of the convolutional neural network model based on the two-dimensional matrix converted from the feature data.

[0022] S2.2. Design at least two convolutional layers to extract features of the input data. Each convolutional layer contains multiple convolution kernels. Each convolution kernel slides over the input data and calculates the convolution result to extract local features.

[0023] S2.3, use the ReLU activation function to map the output of the convolution kernel to a nonlinear space;

[0024] S2.4. A pooling layer is established after each convolutional layer to reduce the dimension and extract the main features of the output of the convolutional layer. The pooling function adopts the maximum pooling.

[0025] S2.5. After the output of the pooling layer is flattened, it is connected to multiple fully connected layers. The fully connected layers perform nonlinear transformations through multiple layers of neurons to learn the relationship between the input data and the seismic performance.

[0026] S2.6. Design the corresponding output nodes based on the output data corresponding to the required seismic performance to obtain the output layer.

[0027] Furthermore, in S3, the prediction result of the model is calculated by forward propagation, and the gradient is calculated by back propagation to update the weight and bias parameters of the model.

[0028] Furthermore, the mean square error loss function is used to evaluate the gap between the seismic performance data output by the model prediction results and the actual collected data, and the Adam optimization algorithm is used to update the model weights and bias parameters for model training.

[0029] Furthermore, during model training, validation set data is regularly used to evaluate the performance of the model, and hyperparameters are adjusted to improve the generalization ability of the model.

[0030] Furthermore, in S4, the model is optimized based on the data on the validation set and the seismic performance data, the hyperparameters are adjusted using random search or Bayesian optimization methods, the model is prevented from overfitting by the early stopping method, the complexity of the model is limited by regularization, and the performance of the model is evaluated by cross-validation using the training set, validation set, and test set.

[0031] Furthermore, in S5, the seismic performance data is used as the output of the convolutional neural network deep learning model, and the trained model is used for prediction to obtain input parameters that meet the requirements; then the results are verified based on the collected data; and the trained convolutional neural network model is used to predict the seismic performance of the reinforced concrete shear wall.

[0032] Compared with the prior art, the present invention achieves the following beneficial effects:

[0033] This paper addresses the lack of a refined macroscopic constitutive model for reinforced concrete shear walls used in structural optimization design. By using crawlers and experimental data to collect data, a convolutional neural network model is constructed and trained, optimized, and tested. Through deep learning, this method achieves intelligent prediction of the seismic performance of reinforced concrete shear walls, optimizing structural design.

[0034] Intelligent prediction of the hysteretic performance of shear walls under varying conditions, such as concrete strength grades, shear wall geometry, steel strength, and volumetric reinforcement ratio, can improve design accuracy and efficiency, shortening the design cycle. This reduces the number and cost of tests, minimizes waste of raw materials, and improves economic efficiency and reduces costs. Improving the mechanical properties of reinforced concrete shear walls: By combining intelligent optimization algorithms with the predictive power of the model, we identify the optimal parameter combination and improve shear wall performance indicators, such as energy dissipation, bearing capacity, and economic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of an intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design provided by an embodiment of the present invention;

[0036] Figure 2 This is a diagram of a crawler technology framework used for data preparation in an intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design provided by an embodiment of the present invention;

[0037] Figure 3 This is a convolutional neural network model architecture diagram of an intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0039] Example

[0040] like Figures 1 to 3 , an intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design, comprising the following steps:

[0041] Data preparation: Data collection and preparation were performed using Python crawler technology, indoor experiments, existing experiments, and finite element models;

[0042] Model construction: Establish a convolutional neural network deep learning model and a prediction model between various input parameters and seismic performance;

[0043] Model training: Train the convolutional neural network deep learning model based on the collected data;

[0044] Model optimization: Use early stopping method, L2 regularization and other methods to optimize the model;

[0045] Performance prediction: The seismic performance of reinforced concrete shear walls is predicted based on the trained convolutional neural network deep learning model and verified by experiments.

[0046] The specific method is as follows:

[0047] S1. Data preparation: Data collection is performed using crawler technology and experiments. The collected data includes obtaining low-cycle reciprocating load test data of reinforced concrete shear walls through crawler technology, and preparing reinforced concrete shear wall specimens in the laboratory and conducting low-cycle reciprocating load tests to obtain and collect test data; pre-processing and dividing the data into training, validation, and test sets;

[0048] The collected data include different concrete strength grades, shear wall geometric dimensions, steel bar strength, volume reinforcement ratio and their corresponding hysteresis curves, skeleton curves, stiffness degradation, energy dissipation capacity and ductility coefficient of shear walls;

[0049] On the one hand, using Python crawler technology, we collected relevant literature published at home and abroad, and compiled historical samples of hysteresis curves for different concrete strength grades, different shear wall geometric dimensions, different steel bar strengths, and different shear wall volume reinforcement ratios. On the other hand, we produced different reinforced concrete shear wall test specimens in the laboratory to test the mechanical properties of reinforced concrete shear walls under low-cycle cyclic loading. We also constructed training sets, validation sets, and test sets.

[0050] The crawler technology framework is as follows Figure 2 As shown in the figure, by defining the input parameters such as concrete strength grade, shear wall geometry, steel bar strength and volume reinforcement ratio, Python crawler technology is used to obtain relevant test data on reinforced concrete shear wall low-cycle reciprocating load tests from data websites such as HowNet, Wanfang, Baidu Library, ScienceDirect, SAGE Journals, Wiley Online Library, and SpringerLink;

[0051] At the same time, reinforced concrete shear wall specimens were made in the laboratory, and low-cycle cyclic reciprocating load tests were carried out to collect data. The data covered the seismic performance of reinforced concrete shear walls under concrete strength grade, geometric dimensions of shear walls, steel strength and volume reinforcement ratio, including but not limited to energy absorption capacity, ductility coefficient, skeleton curve and hysteresis curve.

[0052] The concrete strength grade, geometric dimensions of the shear wall, steel bar strength, and volume reinforcement ratio are taken as characteristic data; the hysteresis curve, skeleton curve, stiffness degradation, energy dissipation capacity, and ductility coefficient of the shear wall are taken as seismic performance data.

[0053] The data is preprocessed, specifically including cleaning the collected data, removing invalid data such as outliers and duplicate values, and then normalizing it. The Min-Max normalization method is used to convert the feature data to the range of [0,1]. The normalization formula is:

[0054]

[0055] Among them, x norm is the normalized data, x is the original data, and x max and x min are the maximum and minimum values of the feature data respectively;

[0056] The feature data is converted into a two-dimensional matrix and divided into training set, validation set and test set.

[0057] Assuming there are N samples, the feature data is converted into a two-dimensional matrix, as shown in Table 1.

[0058] Table 1:

[0059] Sample No. Concrete strength grade Geometric dimensions Steel bar strength Volume reinforcement ratio Sample 1 <![CDATA[a1]]> <![CDATA[b1]]> <![CDATA[c1]]> <![CDATA[d1]]> Sample 2 <![CDATA[a2]]> <![CDATA[b2]]> <![CDATA[b2]]> <![CDATA[d2]]> …… …… …… …… …… Sample N <![CDATA[a n ]]> <![CDATA[b n ]]> <![CDATA[c n ]]> <![CDATA[d n ]]>

[0060] The data is divided into training, validation, and test sets, with the training set accounting for 70% to 80% of the total data, the validation set 10% to 15%, and the test set 10% to 15%. The training set is used for model training, the validation set is used for model tuning and preventing overfitting, and the test set is used to evaluate the final model performance.

[0061] S2. Construct a convolutional neural network model, wherein the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer;

[0062] A convolutional neural network (CNN) model was developed. The model used concrete strength grade, shear wall geometry, steel strength, and volume reinforcement ratio as input features, and the hysteresis curve, skeleton curve, stiffness degradation, energy dissipation capacity, and ductility coefficient of reinforced concrete shear walls as output targets. A prediction model was established between strength grade, geometry, steel strength, and volume reinforcement ratio and seismic performance.

[0063] Use convolutional neural network as the main architecture of the model and use deep learning framework Tensorflow or PyTorch to build a convolutional neural network model.

[0064] The convolutional neural network model includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer. The specific construction method includes:

[0065] S2.1. Input the feature data into the input layer of the convolutional neural network model based on the two-dimensional matrix converted from the feature data.

[0066] According to the two-dimensional matrix of characteristic data, the type and quantity of raw materials, load environment and other characteristic data are input into the input layer of the model;

[0067] S2.2. Design at least two convolutional layers to extract features from the input data. Each convolutional layer contains multiple convolution kernels. Each convolution kernel slides over the input data and calculates the convolution result to extract local features. The size and number of convolution kernels should be designed based on the dimension and complexity of the input features. Considering computational efficiency, a 3×3 convolution kernel is recommended.

[0068] S2.3. Use the ReLU activation function to map the output of the convolution kernel to a nonlinear space, thereby enhancing the expressive power of the model;

[0069] S2.4. A pooling layer is built after each convolutional layer to reduce the dimension of the output of the convolutional layer and extract the main features. The pooling function uses maximum pooling to reduce the dimension of the data.

[0070] S2.5. After the output of the pooling layer is flattened, it is connected to multiple fully connected layers. The fully connected layers perform nonlinear transformations through multiple layers of neurons to learn the relationship between the input data and the seismic performance. Each layer uses the ReLU activation function.

[0071] S2.6. Finally, according to the output data corresponding to the required seismic performance, the corresponding output nodes are designed to obtain the output layer.

[0072] S3. Use the data of the training set to train the constructed model until the predetermined convergence condition or number of iterations is reached;

[0073] During the training process, the model's prediction results, that is, the output values, are calculated through forward propagation, and the gradients are calculated through backpropagation to update the model's weights and bias parameters.

[0074] During model training, the mean square error loss function is used to evaluate the gap between the seismic performance data output by the model prediction results and the actual collected data, and the Adam optimization algorithm is used to update the model weights and bias parameters to perform model training.

[0075] During model training, validation set data is regularly used to evaluate the performance of the model, and hyperparameters (such as learning rate, convolution kernel size, etc.) are adjusted to improve the generalization ability of the model.

[0076] S4. Optimize the convolutional neural network model and use cross-validation on the training set, validation set, and test set to evaluate the performance of the model;

[0077] Early stopping is introduced to prevent model overfitting, and L2 regularization or Dropout technology is used to further improve the generalization ability of the model;

[0078] The model is optimized based on the data on the validation set and the seismic performance data. The hyperparameters are adjusted using random search or Bayesian optimization methods. The early stopping method is used to prevent the model from overfitting. The complexity of the model is limited by regularization. The performance of the model is evaluated by cross-validation using the training set, validation set, and test set.

[0079] Adjust hyperparameters, add regularization terms, and use early stopping techniques to prevent overfitting. Hyperparameters include the learning rate, number of iterations, and batch size. Set the initial learning rate to 0.0001, the number of iterations to at least 100, and the batch size to a power of 2, which is 128. Add regularization terms to limit the complexity of the model's parameters, and use early stopping primarily to save training time.

[0080] S5. Use the trained convolutional neural network model to predict the seismic performance of reinforced concrete shear walls.

[0081] The seismic performance data is used as the output of the convolutional neural network deep learning model, and the trained model is used for prediction to obtain input parameters that meet the requirements; the results are then verified based on the collected data.

[0082] The trained convolutional neural network model is used to predict the seismic performance of reinforced concrete shear walls. The concrete strength grade, geometric dimensions, steel strength and volume reinforcement ratio of the reinforced concrete shear wall to be designed are input into the trained convolutional neural network model. The model outputs predicted energy absorption capacity, ductility coefficient, skeleton curve, hysteresis curve and other data.

[0083] The present invention provides an intelligent prediction method for the seismic performance of reinforced concrete shear walls for structural optimization design, in which the Python crawler technology, deep learning framework Tensorflow or PyTorch, Adam optimization algorithm, early stopping method, L2 regularization or Dropout technology, etc. used are all known existing technologies and will not be repeated here.

[0084] The present invention provides an intelligent prediction method for the seismic performance of reinforced concrete shear walls for structural optimization design, which can realize the prediction of the seismic performance of reinforced concrete shear walls, improve the accuracy and reliability of structural optimization design, shorten the design cycle, facilitate structural selection in the preliminary design stage, and significantly improve economic benefits.

[0085] Based on the disclosure and teachings of the above description, those skilled in the art may also make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and modifications and variations of the present invention should also fall within the scope of protection of the claims of the present invention. In addition, although certain specific terms are used in this description, these terms are only for convenience of description and do not constitute any limitation to the present invention.

Claims

1. An intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design, characterized in that: The following steps are involved: S1. Data preparation: Data collection is performed using crawler technology and experiments. The collected data includes obtaining low-cycle reciprocating load test data of reinforced concrete shear walls through crawler technology, and preparing reinforced concrete shear wall specimens in the laboratory and conducting low-cycle reciprocating load tests to obtain and collect test data; pre-processing and dividing the data into training, validation, and test sets; S2. Construct a convolutional neural network model, wherein the convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer; S3. Use the data of the training set to train the constructed model until the predetermined convergence condition or number of iterations is reached; S4. Optimize the convolutional neural network model and use cross-validation on the training set, validation set, and test set to evaluate the performance of the model; S5. Use the trained convolutional neural network model to predict the seismic performance of reinforced concrete shear walls.

2. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 1, characterized in that: In said S1, the collected data include different concrete strength grades, geometric dimensions of shear walls, steel bar strength, volume reinforcement ratio and the corresponding hysteresis curves, skeleton curves, stiffness degradation, energy dissipation capacity and ductility coefficients of the shear walls; The concrete strength grade, geometric dimensions of the shear wall, steel bar strength, and volume reinforcement ratio are taken as characteristic data; the hysteresis curve, skeleton curve, stiffness degradation, energy dissipation capacity, and ductility coefficient of the shear wall are taken as seismic performance data.

3. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 2 is characterized in that: The data is preprocessed, specifically including cleaning the collected data, removing invalid data, and then normalizing it. The Min-Max normalization method is used to convert the feature data to the range of [0,1]. The normalization formula is: x norm =x-x min ; x max -x min Among them, x norm is the normalized data, x is the original data, and x max and x min are the maximum and minimum values of the feature data respectively; The feature data is converted into a two-dimensional matrix and divided into training set, validation set and test set.

4. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 3 is characterized in that: In S2, a convolutional neural network is used as the main architecture of the model, and a deep learning framework Tensorflow or PyTorch is used to build a convolutional neural network model.

5. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 4, characterized in that: The convolutional neural network model includes an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer. The specific construction method includes: S2.

1. Input the feature data into the input layer of the convolutional neural network model based on the two-dimensional matrix converted from the feature data. S2.

2. Design at least two convolutional layers to extract features of the input data. Each convolutional layer contains multiple convolution kernels. Each convolution kernel slides over the input data and calculates the convolution result to extract local features. S2.3, use the ReLU activation function to map the output of the convolution kernel to a nonlinear space; S2.

4. A pooling layer is established after each convolutional layer to reduce the dimension and extract the main features of the output of the convolutional layer. The pooling function adopts the maximum pooling. S2.

5. After the output of the pooling layer is flattened, it is connected to multiple fully connected layers. The fully connected layers perform nonlinear transformations through multiple layers of neurons to learn the relationship between the input data and the seismic performance. S2.

6. Design the corresponding output nodes based on the output data corresponding to the required seismic performance to obtain the output layer.

6. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 1, characterized in that: In S3, the prediction result of the model is calculated by forward propagation, and the gradient is calculated by back propagation to update the weight and bias parameters of the model.

7. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 6, characterized in that: The mean square error loss function is used to evaluate the gap between the seismic performance data output by the model prediction results and the actual collected data, and the Adam optimization algorithm is used to update the model weights and bias parameters for model training.

8. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 6, characterized in that: During model training, validation set data is regularly used to evaluate the performance of the model and to improve the generalization ability of the model by adjusting hyperparameters.

9. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 8, characterized in that: In S4, the model is optimized based on the data on the validation set and the seismic performance data, the hyperparameters are adjusted using random search or Bayesian optimization methods, the model is prevented from overfitting by the early stopping method, the complexity of the model is limited by regularization, and the performance of the model is evaluated by cross-validation using the training set, validation set, and test set.

10. The intelligent prediction method for seismic performance of reinforced concrete shear walls for structural optimization design according to claim 1, characterized in that: In S5, the seismic performance data is used as the output of the convolutional neural network deep learning model, and the trained model is used to perform predictions to obtain input parameters that meet the requirements; the results are then verified based on the collected data; and the trained convolutional neural network model is used to predict the seismic performance of the reinforced concrete shear wall.

Images