Earthquake response prediction method and device for seismic isolation structure

Abstract

The invention provides a seismic response prediction method and device for a seismic isolation structure, and the method comprises the steps: obtaining an actually-measured seismic oscillation acceleration time history record, and obtaining the flexion-to-weight ratio of the seismic isolation structure; the seismic oscillation acceleration time history record and the flexion-to-weight ratio are fused to form a fusion feature; inputting the fusion features into a trained seismic response prediction neural network model to obtain seismic response of each layer of the seismic isolation structure; wherein the seismic response prediction neural network model is obtained through training based on a fusion feature training set composed of historical seismic motion acceleration records and the ratio of flexion to weight. The calculation amount of seismic response prediction of the seismic isolation structure is reduced, the prediction time is saved, and the prediction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster prevention and reduction engineering, and in particular to a method and device for predicting earthquake response of a seismic isolation structure. Background Art

[0002] The prediction of seismic response of seismic isolation structures is of great significance for seismic isolation design and performance evaluation. Existing seismic response prediction technology usually uses time history analysis method to predict seismic response. However, the use of time history analysis method to calculate the seismic response of seismic isolation structures has the problem of large amount of calculation and long time consumption. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a method and device for predicting the seismic response of a seismic isolation structure, which can avoid the problem of large amount of calculation required for calculating the seismic response of the seismic isolation structure using a time-history analysis method, reduce the amount of calculation required for predicting the seismic response of the seismic isolation structure, save prediction time, and improve prediction efficiency.

[0004] In order to achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting seismic response of a seismic isolation structure, comprising:

[0006] Obtain the measured earthquake acceleration time history record and obtain the buckling-to-weight ratio of the seismic isolation structure;

[0007] fusing the seismic acceleration time history record with the buckling-to-weight ratio to form a fusion feature;

[0008] The fused features are input into the trained seismic response prediction neural network model to obtain the seismic response of each layer of the seismic isolation structure; wherein the seismic response prediction neural network model is trained based on the historical seismic acceleration records and the fused feature training set composed of the yield-to-weight ratio.

[0009] Furthermore, the embodiment of the present invention provides a first possible implementation of the first aspect, wherein the method for predicting seismic response of a seismic isolation structure further includes:

[0010] Obtain multiple historical seismic acceleration time history records;

[0011] Acquire an elastic-plastic analysis model of the seismic isolation structure, and input each of the historical seismic acceleration time history records into the elastic-plastic analysis model to determine the seismic response of each layer of the seismic isolation structure corresponding to each of the historical seismic acceleration time history records;

[0012] Acquire the buckling-to-weight ratio of the seismic isolation structure, and fuse each of the seismic acceleration time history records with the buckling-to-weight ratio to form a plurality of fusion features;

[0013] The fusion features and their corresponding seismic responses of each layer of the seismic isolation structure are used as a sample set, and a training set is selected from the sample set and input into the seismic response prediction neural network model for training to obtain a trained seismic response prediction neural network model.

[0014] Furthermore, the embodiment of the present invention provides a second possible implementation of the first aspect, wherein the step of fusing the seismic acceleration time history record with the buckling-to-weight ratio to form a fusion feature comprises:

[0015] Compile the buckling-to-weight ratio parameter of the seismic isolation structure into keyword attributes to obtain corresponding key attribute words;

[0016] Filling the matrix formed by the key attribute words so that the length of the matrix of the key attribute words is the same as the length of the matrix of the seismic acceleration time history record;

[0017] The filled matrix of the key attribute words is fused with the matrix of the seismic acceleration time history record to form the fusion feature.

[0018] Furthermore, an embodiment of the present invention provides a third possible implementation of the first aspect, wherein the step of performing matrix fusion on the filled matrix of the key attribute words and the matrix of the seismic acceleration time history record to form the fusion feature comprises:

[0019] The filled matrix of the key attribute words is horizontally spliced ​​with the matrix of the seismic acceleration time history record to form a feature fusion matrix; wherein the data of each row of the feature fusion matrix includes the seismic acceleration time history record and the buckling-to-weight ratio.

[0020] Furthermore, an embodiment of the present invention provides a fourth possible implementation of the first aspect, wherein the site category corresponding to the historical seismic acceleration time history record is the same as the site category corresponding to the seismic isolation structure.

[0021] Furthermore, an embodiment of the present invention provides a fifth possible implementation of the first aspect, wherein the historical seismic acceleration time history record and the measured seismic acceleration time history record both include seismic data of a set interval and a set duration.

[0022] Furthermore, an embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein the earthquake response prediction neural network model is a long short-term memory neural network.

[0023] Furthermore, the embodiment of the present invention provides a seventh possible implementation of the first aspect, wherein the method for predicting seismic response of a seismic isolation structure further includes:

[0024] The sample set is divided into a training set, a test set and a validation set, the trained earthquake response prediction neural network model is tested based on the test set, and the accuracy of the earthquake response prediction neural network model is determined based on the test result.

[0025] In a second aspect, an embodiment of the present invention further provides a seismic response prediction device for a seismic isolation structure, comprising:

[0026] An acquisition module is used to obtain the measured earthquake acceleration time history record and the buckling-to-weight ratio of the seismic isolation structure;

[0027] A fusion module, used for fusing the earthquake acceleration time history record with the buckling-to-weight ratio to form a fusion feature;

[0028] A prediction module is used to input the fusion features into a trained earthquake response prediction neural network model to obtain the earthquake response of each layer of the seismic isolation structure; wherein the earthquake response prediction neural network model is trained based on a fusion feature training set composed of historical seismic acceleration records and the yield-to-weight ratio.

[0029] In a third aspect, an embodiment of the present invention provides an electronic device, including: a processor and a storage device;

[0030] The storage device stores a computer program, and when the computer program is executed by the processor, the method according to any one of the first aspects is executed.

[0031] The embodiment of the present invention provides a method and device for predicting the seismic response of a seismic isolation structure, the method comprising: obtaining a measured seismic acceleration time history record, obtaining the buckling-weight ratio of the seismic isolation structure; fusing the seismic acceleration time history record with the buckling-weight ratio to form a fusion feature; inputting the fusion feature into a trained seismic response prediction neural network model to obtain the seismic response of each layer of the seismic isolation structure; wherein the seismic response prediction neural network model is trained based on a fusion feature training set composed of historical seismic acceleration records and buckling-weight ratios. The present invention can quickly predict the seismic response of each layer of the seismic isolation structure by pre-training the seismic response prediction neural network model and inputting the fusion feature formed by fusing the seismic acceleration time history record with the buckling-weight ratio into the model, providing important data support and reference for the design and performance evaluation of the seismic isolation structure, avoiding the problem of large amount of calculation when using the time history analysis method to calculate the seismic response of the seismic isolation structure, reducing the amount of calculation for predicting the seismic response of the seismic isolation structure, saving prediction time, and improving prediction efficiency.

[0032] Other features and advantages of the embodiments of the present invention will be described in the following description, or some features and advantages can be inferred or determined without doubt from the description, or can be learned by implementing the above-mentioned techniques of the embodiments of the present invention.

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A flow chart of a method for predicting seismic response of a seismic isolation structure provided by an embodiment of the present invention is shown;

[0036] Figure 2 A schematic diagram of the structure of a neural network model for earthquake response prediction provided by an embodiment of the present invention is shown;

[0037] Figure 3 A three-dimensional schematic diagram of a five-layer frame structure provided by an embodiment of the present invention is shown;

[0038] Figure 4a A schematic diagram of a seismic isolation arrangement scheme with a yield-to-weight ratio of 2% provided in an embodiment of the present invention is shown;

[0039] Figure 4b A schematic diagram of a seismic isolation arrangement scheme with a yield-to-weight ratio of 2.5% provided in an embodiment of the present invention is shown;

[0040] Figure 4c A schematic diagram of a seismic isolation arrangement scheme with a yield-to-weight ratio of 3% provided in an embodiment of the present invention is shown;

[0041] Figure 4d A schematic diagram of a seismic isolation arrangement scheme with a yield-to-weight ratio of 3.5% provided in an embodiment of the present invention is shown;

[0042] Figure 4e A schematic diagram of a seismic isolation arrangement scheme with a yield-to-weight ratio of 4% provided in an embodiment of the present invention is shown;

[0043] Figure 5 A loss function curve diagram in a training process provided by an embodiment of the present invention is shown;

[0044] Figure 6a to Figure 6fA comparison chart of predicted peak displacement and true value results under different bending-to-weight ratios and the entire test set provided by an embodiment of the present invention is shown;

[0045] Figure 7a to Figure 7e A comparison diagram of predicted time history and actual time history of displacement of the seismic isolation layer under different samples provided by an embodiment of the present invention is shown;

[0046] Figure 8 A schematic structural diagram of a seismic response prediction device for a seismic isolation structure provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be described below in conjunction with the drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0048] At present, the prediction of seismic response of seismic isolation structures is of great significance for seismic isolation design and performance evaluation. The time-history analysis method is faced with the challenges of large amount of calculation and long time consumption to calculate the response of seismic isolation structures. Deep learning methods provide an important means to solve this challenge. However, the research on the prediction of seismic isolation structure response based on deep learning is relatively limited.

[0049] In order to improve the above problems, an embodiment of the present invention provides a method and device for predicting seismic response of a seismic isolation structure. The embodiment of the present invention is described in detail below.

[0050] This embodiment provides a method for predicting the seismic response of a seismic isolation structure. The method can be applied to electronic devices such as computers. Figure 1 The flowchart of the method for predicting the seismic response of a seismic isolation structure is shown in FIG. 1 , and the method mainly comprises the following steps:

[0051] Step S102, obtaining the measured earthquake acceleration time history record and obtaining the yield-to-weight ratio of the seismic isolation structure;

[0052] Based on the measured earthquake acceleration time history records at the location of the isolation structure, the key parameter of the isolation structure to be predicted, the yield-to-weight ratio, is obtained.

[0053] Step S104, fusing the earthquake acceleration time history record with the buckling-to-weight ratio to form a fusion feature;

[0054] The above fusion features include two feature information: the seismic acceleration time history record and the yield weight ratio of the seismic isolation structure. The fusion features may be a feature fusion matrix.

[0055] The buckling-to-weight ratio parameters of the seismic isolation structure are compiled with corresponding key attribute words; the corresponding key attribute words are filled into a matrix consistent with the length of the seismic acceleration time history record, and each element in the matrix is ​​a corresponding key attribute word; the buckling-to-weight ratio key attribute word matrix of the seismic isolation frame building and the seismic acceleration time history record matrix are stacked and spliced ​​to form a feature fusion matrix, and each row of the feature fusion matrix contains two features: the seismic acceleration record and the buckling-to-weight ratio of the seismic isolation structure.

[0056] Step S106, inputting the fusion features into the trained earthquake response prediction neural network model to obtain the earthquake response of each layer of the seismic isolation structure;

[0057] The above-mentioned earthquake response prediction neural network model is trained based on a fusion feature training set consisting of historical seismic acceleration records and yield-to-weight ratio.

[0058] In a specific implementation, the earthquake response prediction neural network model may be a long short-term memory neural network (Long Short-Term Memory, LSTM).

[0059] See Figure 2 The schematic diagram of the earthquake response prediction neural network model structure is shown in FIG. The long short-term memory neural network includes 3 LSTM layers (each layer contains 100 neural units), 4 activation function layers and 2 fully connected layers (the number of units in the fully connected layers is 100 and 6 respectively), wherein the activation function layer can use the Relu activation function, the optimizer can use the Adam optimizer (stochastic gradient descent optimizer), and the learning rate can be set to 0.001. The input layer of the network is the earthquake acceleration time history and the buckling weight ratio of the seismic isolation structure, and the output layer is the displacement time history of each layer of the seismic isolation structure corresponding to the earthquake motion.

[0060] The seismic response prediction method for the above-mentioned seismic isolation structure provided in this embodiment can quickly predict the seismic response of each layer of the seismic isolation structure by pre-training the seismic response prediction neural network model and inputting the fusion feature formed by fusing the seismic acceleration time history record and the buckling-to-weight ratio into the model, providing important data support and reference for the design and performance evaluation of the seismic isolation structure, avoiding the problem of large amount of calculation when calculating the seismic response of the seismic isolation structure by using the time history analysis method, reducing the amount of calculation for predicting the seismic response of the seismic isolation structure, saving prediction time, and improving prediction efficiency.

[0061] In one embodiment, the method provided in this embodiment further includes:

[0062] Obtain multiple historical seismic acceleration time history records;

[0063] Obtaining an elastic-plastic analysis model of the seismic isolation structure, inputting each historical earthquake acceleration time history record into the elastic-plastic analysis model to determine the seismic response of each layer of the seismic isolation structure corresponding to each historical earthquake acceleration time history record;

[0064] Obtain the buckling-to-weight ratio of the seismic isolation structure, fuse the vibration acceleration time history records of various places with the buckling-to-weight ratio respectively, and form multiple fusion features;

[0065] Each fusion feature and its corresponding seismic response of each layer of the seismic isolation structure are taken as a sample set, and a training set is selected from the sample set and input into the seismic response prediction neural network model for training to obtain the trained seismic response prediction neural network model.

[0066] According to the building seismic isolation design standard, multiple seismic isolation layout schemes with different yield-to-weight ratios are designed, and the elastic-plastic analysis model of the seismic isolation structure with different yield-to-weight ratios is established based on Perform-3D. The beams and columns are simulated with fiber models, the material constitutive model of concrete and steel bars is defined in the form of five-fold line and three-fold line respectively, and the seismic isolation bearing is simulated with Seismic Isolator unit.

[0067] Input each historical earthquake acceleration time history record into the elastoplastic analysis model to obtain the seismic response of each layer of the seismic isolation structure corresponding to each historical earthquake acceleration time history record;

[0068] The acceleration time history records and the buckling-to-weight ratios of various locations are fused to form multiple fusion features, and each fusion feature and its corresponding seismic response of each layer of the seismic isolation structure are used as the sample set of the model. A training set is selected from the sample set and input into the seismic response prediction neural network model, that is, the fusion features marked with the seismic response of each layer of the seismic isolation structure are input into the model for training.

[0069] The established LSTM architecture is analyzed for network hyperparameters, the parameters used are analyzed, and the optimal parameters are determined. The network depth, number of LSTM layer units, batch size, learning rate, and optimizer type are changed to observe their effects on the structural accuracy of displacement time history prediction of each layer of the seismic isolation structure, thereby determining the optimal parameters.

[0070] For example, see Figure 3 A three-dimensional schematic diagram of a five-story frame structure is shown. Based on the three-dimensional model, isolation layout schemes with different yield-to-weight ratios are designed, and an elastic-plastic analysis model of isolation structures with different yield-to-weight ratios is established based on Perform3D.

[0071] Obtain the key parameter of the seismic isolation structure response, the yield-to-weight ratio, and fuse it with the ground motion acceleration time history record, as shown in Figure 4a to Figure 4eThe figure shows a schematic diagram of five seismic isolation layout schemes with different yield-to-weight ratios. N4 represents a natural rubber bearing with a diameter of 400 mm, R4 represents a lead rubber bearing with a diameter of 400 mm, N5 represents a natural rubber bearing with a diameter of 500 mm, R5 represents a lead rubber bearing with a diameter of 500 mm, and R6 represents a lead rubber bearing with a diameter of 600 mm. Different feature fusion matrices are obtained by combining the five different yield-to-weight ratios with the acquired seismic acceleration time history records.

[0072] The seismic acceleration time history is input into the elastoplastic analysis model of the seismic isolation structure, and the seismic response of each layer of the seismic isolation structure is calculated. Based on the above seismic motion database and seismic isolation analysis model, a total of 19,550 sets of data are calculated for subsequent model training.

[0073] The feature matrix integrating seismic motion time history and yield-to-weight ratio is used as the input of the deep neural network, and the seismic response of each layer of the structure is used as the output. An LSTM network is constructed and the seismic response prediction model of the seismic isolation structure is obtained through training.

[0074] In one embodiment, this embodiment provides a specific implementation method for fusing the seismic acceleration time history record with the buckling weight ratio to form a fusion feature:

[0075] Compile the buckling-to-weight ratio parameters of the seismic isolation structure into keyword attributes to obtain the corresponding key attribute words; based on the prior knowledge in the professional field, compile the buckling-to-weight ratio parameters of the seismic isolation building into key attribute words, for example, if the buckling-to-weight ratio of the seismic isolation building is 2%, the corresponding key attribute word is 0.02;

[0076] Fill the matrix formed by the key attribute words so that the matrix length of the key attribute words is the same as the matrix length of the seismic acceleration time history record; fill the corresponding key attribute words to fill the matrix to the same length as the seismic acceleration time history record, and each element in the matrix is ​​the corresponding key attribute word;

[0077] The matrix of the filled key attribute words is fused with the matrix of the seismic acceleration time history record to form a fusion feature.

[0078] In a specific implementation, the matrix of filled key attribute words and the matrix of seismic acceleration time history records are horizontally spliced ​​to form a feature fusion matrix; wherein the data of each row of the feature fusion matrix includes seismic acceleration time history records and yield weight ratios.

[0079] The key attribute word matrix of the yield weight ratio of the seismic isolation building and the seismic acceleration time history record matrix are stacked and spliced ​​along the y-axis direction to form a feature fusion matrix. Each row of the matrix contains two features: seismic acceleration record and seismic isolation building yield weight ratio. The input data size of the LSTM network is [batch size, 1500, 2], and the output size is [batch size, 1500, N+1], where 1500 is the number of seismic data points and N is the number of building floors.

[0080] In one implementation, the site category corresponding to the historical seismic acceleration time history records provided in this embodiment is the same as the site category corresponding to the seismic isolation structure.

[0081] The principles of the above-mentioned historical seismic acceleration time history records and measured seismic acceleration time history records include: selecting strong vibrations as input and selecting records with an epicenter distance of less than 50 km.

[0082] When selecting historical seismic acceleration time history records, the site type of the target building should be considered. s30 Between the venue category corresponding to V s30 The seismic motion record of the interval is the average shear wave velocity within a depth of 30m below the surface of the target building with the same site category as the seismic isolation structure.

[0083] In one implementation, the historical seismic acceleration time history records and the measured seismic acceleration time history records provided in this embodiment both include seismic data of a set interval and a set duration.

[0084] When selecting historical seismic acceleration time-history records and measured seismic acceleration time-history records, it is necessary to preprocess the selected seismic motion data, unify the durations of the different seismic motions selected, and intercept the seismic motion data in segments with half the duration before and after the seismic motion reaches the PGA moment. For the insufficient time before and after, add data points by filling "0" to obtain seismic motion data of a set duration, which can be 30s; unify the sampling frequency of strong motions, and adjust the sampling frequency of strong motion data to 50Hz (step size 0.02s); modulate the selected seismic motion, and select the amplitude modulation coefficients as 2.0 and 4.0.

[0085] In one implementation, the method provided in this embodiment also includes: dividing the sample set into a training set, a test set, and a validation set, testing the trained earthquake response prediction neural network model based on the test set, and determining the accuracy of the earthquake response prediction neural network model based on the test results.

[0086] The sample set is divided into training set, test set and validation set. The training set and validation set are used for LSTM training, and the test set is used as a data set to test the prediction ability of the trained LSTM model. The prediction accuracy of the model is measured by the mean square error, mean absolute error, root mean square error and Pearson correlation coefficient. Based on the better model, the entire data set is trained, and the number of training iterations is reasonably designed. When the LSTM model can achieve the convergence of the validation error after multiple iterations and the mean square error of the training set and validation set is reduced, it means that the training effect has been achieved.

[0087] The trained seismic response prediction neural network model of the isolation structure is used to predict the displacement time history of the test set and compare the results with the true value to obtain the mean square error, mean absolute error, root mean square error and Pearson correlation coefficient of the test set, and determine the accuracy of the seismic response prediction neural network model in predicting the displacement time history of the isolation structure.

[0088] For example, the sample set is divided into training set: test set: validation set = 0.8:0.1:0.1 ratio. The training set and validation set are used for LSTM training, and the test set is used as a data set to test the prediction ability of the trained LSTM model. The prediction accuracy of the model is then measured by the mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE) and Pearson correlation coefficient (r). Based on the optimal LSTM model parameter values, the entire data set (19550 samples) is trained, and the number of training iterations can be set to 1000 times, see Figure 5 The loss function curve during the training process is shown in the figure. After 1000 iterations, the LSTM model can achieve convergence of the verification error, and the mean square error of the training set and the verification set has been reduced, indicating that the training effect has been achieved, thereby establishing a seismic response prediction model for seismic isolation structures.

[0089] The trained seismic response prediction model of the seismic isolation structure is used to predict the displacement time history of the seismic isolation layer in the test set, and the 1955 samples of the test set are evaluated according to the evaluation index, see Figure 6a to Figure 6f The comparison chart of predicted peak displacement and true value results under various bending-to-weight ratios and the entire test set is shown. Figure 6a The predicted peak displacement of the seismically isolated building with a yield-to-weight ratio of 2% is compared with the actual value: R = 0.92, R 2 =0.84; Figure 6b The predicted peak displacement of the seismically isolated building with a yield-to-weight ratio of 2.5% is compared with the actual value: R = 0.97, R 2 =0.95; Figure 6c The comparison between the predicted peak displacement and the actual value for a seismically isolated building with a yield-to-weight ratio of 3% is shown: R = 0.99, R2 =0.99; Figure 6d The predicted peak displacement of the seismically isolated building with a yield-to-weight ratio of 3.5% is compared with the actual value: R = 0.99, R 2 =0.99; Figure 6e The predicted peak displacement of the seismically isolated building with a yield-to-weight ratio of 4% is compared with the actual value: R = 0.96, R 2 =0.92; through summary, the overall determination coefficient R = 0.97, R 2 =0.95, Figure 6f The comparison between the predicted peak displacement of the entire seismic isolation building and the actual value result is shown, which shows that the method provided in this embodiment can well predict the displacement peak of the structural seismic isolation layer.

[0090] The samples of the test set are arranged from high to low according to the evaluation index R, among which the number of samples with evaluation indexes above 90% accounts for 82.6% of the total samples. The samples with index R arranged at 90% (sample 1), 75% (sample 2), 50% (sample 3), 25% (sample 4) and 10% (sample 5) are selected to draw a comparison between the predicted time history of the isolation layer displacement and the actual time history, as shown in Figure 2. Figure 7a to Figure 7e The comparison chart of the predicted time history and the actual time history of the isolation layer displacement of samples 1 to 5 is shown in the figure. Figure 7a to Figure 7e The solid line in the middle is the actual time course, and the dotted line is the predicted time course. Figure 7a The coefficient of determination R of sample 1 is 2 It can reach 0.97, thus confirming that the method provided in this embodiment can better predict the displacement time history of the seismic isolation structure.

[0091] The seismic response prediction method of the seismic isolation structure provided in this embodiment establishes a seismic response prediction model for the seismic isolation frame, and determines the determination coefficient R through verification and testing. 2 , thus determining that the method can better predict the displacement time history of the isolation structure, providing an important means for the prediction of the seismic response and performance evaluation of the isolation structure.

[0092] Corresponding to the seismic response prediction method of the seismic isolation structure provided in the above embodiment, the embodiment of the present invention provides a seismic response prediction device of the seismic isolation structure, see Figure 8 The schematic diagram of the structure of a seismic response prediction device for a seismic isolation structure is shown, and the device includes the following modules:

[0093] An acquisition module 81 is used to acquire the measured earthquake acceleration time history record and the yield-to-weight ratio of the seismic isolation structure;

[0094] A fusion module 82, for fusing the seismic acceleration time history record with the buckling-to-weight ratio to form a fusion feature;

[0095] The prediction module 83 is used to input the fusion features into the trained earthquake response prediction neural network model to obtain the earthquake response of each layer of the seismic isolation structure; wherein the earthquake response prediction neural network model is trained based on the fusion feature training set composed of historical seismic acceleration records and yield-to-weight ratio.

[0096] The seismic response prediction device for the above-mentioned seismic isolation structure provided in the present embodiment can quickly predict the seismic response of each layer of the seismic isolation structure by pre-training the seismic response prediction neural network model and inputting the fusion feature formed by fusing the seismic acceleration time history record and the buckling-to-weight ratio into the model, thereby providing important data support and reference for the design and performance evaluation of the seismic isolation structure, avoiding the problem of large amount of calculation when calculating the seismic response of the seismic isolation structure by using the time history analysis method, reducing the amount of calculation for predicting the seismic response of the seismic isolation structure, saving prediction time, and improving prediction efficiency.

[0097] In one embodiment, the above device further comprises:

[0098] A training module is used to obtain multiple historical seismic acceleration time history records; obtain an elastic-plastic analysis model of a seismic isolation structure, input each historical seismic acceleration time history record into the elastic-plastic analysis model to determine the seismic response of each layer of the seismic isolation structure corresponding to each historical seismic acceleration time history record; obtain the buckling-to-weight ratio of the seismic isolation structure, fuse the local acceleration time history records with the buckling-to-weight ratio respectively, and form multiple fusion features; use each fusion feature and its corresponding seismic response of each layer of the seismic isolation structure as a sample set, select a training set from the sample set and input it into a seismic response prediction neural network model for training, and obtain a trained seismic response prediction neural network model.

[0099] In one embodiment, the above-mentioned fusion module is used to perform keyword attribute compilation on the buckling-to-weight ratio parameters of the seismic isolation structure to obtain corresponding key attribute words; fill the matrix formed by the key attribute words so that the matrix length of the key attribute words is the same as the matrix length of the seismic acceleration time history record; and perform matrix fusion on the filled matrix of the key attribute words and the matrix of the seismic acceleration time history record to form a fusion feature.

[0100] In one embodiment, the fusion module is used to horizontally splice the matrix of filled key attribute words with the matrix of seismic acceleration time history records to form a feature fusion matrix; wherein the data of each row of the feature fusion matrix includes seismic acceleration time history records and buckling-to-weight ratios.

[0101] In one embodiment, the site category corresponding to the historical seismic acceleration time history record is the same as the site category corresponding to the seismic isolation structure.

[0102] In one embodiment, the historical seismic acceleration time history records and the measured seismic acceleration time history records both include seismic data of a set interval and a set duration.

[0103] In one embodiment, the earthquake response prediction neural network model is a long short-term memory neural network.

[0104] In one embodiment, the above device further comprises:

[0105] The testing module is used to divide the sample set into a training set, a test set and a validation set, test the trained earthquake response prediction neural network model based on the test set, and determine the accuracy of the earthquake response prediction neural network model based on the test results.

[0106] The implementation principle and technical effects of the device provided in this embodiment are the same as those of the aforementioned embodiments. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0107] An embodiment of the present invention provides an electronic device, which includes a processor and a memory. The memory stores a computer program that can be run on the processor. When the processor executes the computer program, the steps of the method provided in the above embodiment are implemented.

[0108] An embodiment of the present invention provides a computer-readable medium, wherein the computer-readable medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method described in the above embodiment.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned embodiment, and will not be repeated here.

[0110] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0111] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0112] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0113] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting seismic response of a seismic isolation structure, characterized in that: include: Obtain the measured earthquake acceleration time history record and obtain the buckling-to-weight ratio of the seismic isolation structure; fusing the seismic acceleration time history record with the buckling-to-weight ratio to form a fusion feature; The fused features are input into the trained seismic response prediction neural network model to obtain the seismic response of each layer of the seismic isolation structure; wherein the seismic response prediction neural network model is trained based on the historical seismic acceleration records and the fused feature training set composed of the yield-to-weight ratio.

2. The method according to claim 1, characterized in that Also includes: Obtain multiple historical seismic acceleration time history records; Acquire an elastic-plastic analysis model of the seismic isolation structure, and input each of the historical seismic acceleration time history records into the elastic-plastic analysis model to determine the seismic response of each layer of the seismic isolation structure corresponding to each of the historical seismic acceleration time history records; Acquire the buckling-to-weight ratio of the seismic isolation structure, and fuse each of the seismic acceleration time history records with the buckling-to-weight ratio to form a plurality of fusion features; The fusion features and their corresponding seismic responses of each layer of the seismic isolation structure are used as a sample set, and a training set is selected from the sample set and input into the seismic response prediction neural network model for training to obtain a trained seismic response prediction neural network model.

3. The method according to claim 1 or 2, characterized in that: The step of fusing the seismic acceleration time history record with the buckling-to-weight ratio to form a fusion feature comprises: Compile the buckling-to-weight ratio parameter of the seismic isolation structure into keyword attributes to obtain corresponding key attribute words; Filling the matrix formed by the key attribute words so that the length of the matrix of the key attribute words is the same as the length of the matrix of the seismic acceleration time history record; The filled matrix of the key attribute words is fused with the matrix of the seismic acceleration time history record to form the fusion feature.

4. The method according to claim 3, characterized in that: The step of performing matrix fusion of the filled matrix of the key attribute words and the matrix of the seismic acceleration time history record to form the fusion feature comprises: The filled matrix of the key attribute words is horizontally spliced ​​with the matrix of the seismic acceleration time history record to form a feature fusion matrix; wherein the data of each row of the feature fusion matrix includes the seismic acceleration time history record and the buckling-to-weight ratio.

5. The method according to claim 2, characterized in that: The site category corresponding to the historical seismic acceleration time history record is the same as the site category corresponding to the seismic isolation structure.

6. The method according to claim 2, characterized in that The historical seismic acceleration time history record and the measured seismic acceleration time history record both include seismic data of a set interval and a set duration.

7. The method according to claim 1, characterized in that The earthquake response prediction neural network model is a long short-term memory neural network.

8. The method according to claim 2, characterized in that: Also includes: The sample set is divided into a training set, a test set and a validation set, the trained earthquake response prediction neural network model is tested based on the test set, and the accuracy of the earthquake response prediction neural network model is determined based on the test result.

9. A seismic response prediction device for a seismic isolation structure, characterized in that: include: An acquisition module is used to obtain the measured earthquake acceleration time history record and the buckling-to-weight ratio of the seismic isolation structure; A fusion module, used for fusing the earthquake acceleration time history record with the buckling-to-weight ratio to form a fusion feature; A prediction module is used to input the fusion features into a trained earthquake response prediction neural network model to obtain the earthquake response of each layer of the seismic isolation structure; wherein the earthquake response prediction neural network model is trained based on a fusion feature training set composed of historical seismic acceleration records and the yield-to-weight ratio.

10. An electronic device, characterized in that: include: processor and storage device; The storage device stores a computer program, which, when executed by the processor, performs the method according to any one of claims 1 to 8.

Images