Building floor seismic response calculation method based on spline interpolation and machine learning

Through spline interpolation and machine learning methods, the dimensions of building earthquake damage results were processed, combined with XGBoost and Optuna algorithms, the problem of limitation of the output dimension of the agent model was solved, and the rapid and efficient calculation of detailed earthquake damage results of building complexes was achieved, and the existing In the technology, the high computational cost of physical models and the fixed dimensions of proxy models output.

CN118965884APending Publication Date: 2024-11-15HARBIN INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411021212.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Due to the limitation of output dimensions, existing machine learning agent models cannot obtain detailed seismic results of buildings (seismic responses of all layers), while physical models are difficult to use for earthquakes in large-scale building complexes due to the cumbersome modeling process and high computational cost. Simulation.

Method used

Using spline interpolation and machine learning methods, the detailed seismic results of buildings of different layers are processed into the same dimensions, and the proxy model is constructed, and the mapping relationship from input to output is established through XGBoost and Optuna algorithms, and the spline interpolation is used to interpolate and Reverse sampling realizes the restoration of the calculation results.

Benefits of technology

It realizes rapid calculation of detailed earthquake damage results for buildings of any number of buildings, with high accuracy, high efficiency and low cost, the calculation efficiency is close to that of the physical model, and the training of the agent model is low and easy to use.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965884A_ABST
    Figure CN118965884A_ABST
Patent Text Reader

Abstract

The invention provides a method for calculating seismic response of each floor of a building based on spline interpolation and machine learning. According to the method, detailed earthquake damage results (earthquake responses of all floors) of buildings of various floor numbers are processed through spline interpolation and equidistant sampling methods, and the detailed earthquake damage results are unified into the same dimension, so that the detailed earthquake damage results of the buildings of any floor number can be calculated by an improved agent model. The method can be applied to any building with a layer structure. The method is high in precision, high in efficiency and low in cost, the calculation precision is close to that of a physical model, the calculation efficiency is thousands of times that of the physical model, the requirement for computer configuration is low, and the training and optimization process of the agent model can be completed within several hours by using a common personal computer.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of rapid calculation of earthquake responses of multi-story buildings, and in particular to a method for calculating earthquake responses of each floor of a building based on spline interpolation and machine learning. Background Art

[0002] Building complexes are the core of urban infrastructure and the main place for human production activities. They are also faced with a high risk of earthquakes. Previous earthquake damage survey results show that the economic losses and casualties caused by earthquakes are mainly caused by damage to buildings. Accurate building complex earthquake damage simulation technology can be used for quantitative assessment of economic losses, preliminary estimation of casualties and resilience evaluation of building complexes.

[0003] At present, physical models based on finite elements or proxy models based on machine learning are usually used to calculate the seismic response of buildings. However, the current physical models have the following shortcomings: (1) The modeling process is cumbersome. It takes a lot of manpower to build a sophisticated physical model, and it has high requirements on the modeler's modeling technology level and mechanical knowledge reserve. Otherwise, it is difficult to guarantee the final calculation accuracy; (2) The calculation cost is high. Because the calculation process involves solving differential equations, the calculation efficiency of the physical model is low. It takes dozens of hours to perform a single dynamic analysis of a large single building on a conventional personal computer. Therefore, the physical model is difficult to use for earthquake damage simulation of large-scale building complexes.

[0004] The proxy model is a black box model that directly builds a mapping from ground motion (input) to building seismic response (output) based on machine learning or deep learning algorithms, skipping the process of solving differential equations. Therefore, the computational efficiency of the proxy model far exceeds that of the physical model, and its computational accuracy is close to that of the physical model in specific scenarios. However, the output dimension of the proxy model is fixed and cannot change with the number of floors of the building. Therefore, when the research object is a building complex, the current proxy model is only used to calculate the maximum value of the seismic response of all floors of the building or the seismic response of all floors of a building with a specific number of floors, and cannot be used to calculate the seismic response of all floors of a building with any number of floors.

[0005] In summary, the use of physical models can obtain detailed earthquake damage results for buildings (seismic responses of all layers), but the feasibility of earthquake damage simulation for building complexes based on physical models is low, because it requires a huge amount of manpower and computing resources to complete the physical modeling and dynamic analysis steps. Although the proxy model has high computational efficiency, the output dimension cannot be changed at will after the modeling is completed, so it cannot be used to calculate the detailed earthquake damage results of the building complex. If the data dimensions of the detailed earthquake damage results of buildings with different numbers of floors can be unified, so that samples from buildings with any number of floors can be trained in the same proxy model, the detailed earthquake damage results of buildings with any number of floors can be quickly calculated based on the proxy model, thereby combining the advantages of existing proxy models and physical models, that is, obtaining detailed earthquake damage results of the building complex while maintaining ultra-high computational efficiency. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that the existing machine learning proxy model cannot obtain detailed earthquake damage results (seismic responses of all layers) of the building due to the limitation of the output dimension, and proposes a method for calculating the earthquake response of each layer of the building based on spline interpolation and machine learning. The method processes the detailed earthquake damage results of buildings with various numbers of floors by spline interpolation and equidistant sampling, unifies them into the same dimension, so that the improved proxy model can calculate the detailed earthquake damage results of buildings with any number of floors. The scope of application of the present invention is any building with a layer structure.

[0007] The present invention is implemented by the following technical scheme. The present invention proposes a method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning. The method comprises the following steps:

[0008] Step 1: Select some buildings with different numbers of floors and a certain number of earthquake records, establish a physical model of the building, use the earthquake records as external excitation to perform dynamic analysis on the physical model, and obtain the seismic response of all floors of each building under each earthquake;

[0009] Step 2: Calculate the seismic parameters based on the seismic records, and use the seismic parameters and building parameters as the proxy model input; process the seismic responses of all layers of each building into the same dimension as the proxy model output through spline interpolation and equidistant sampling methods; thereby construct the data set required for training the proxy model;

[0010] Step 3: Divide the data set in step 2 into a training set, a validation set, and a test set in a specific ratio, which are used for proxy model training, parameter adjustment, and accuracy evaluation, respectively. Then, a proxy model is established based on the XGBoost machine learning algorithm and the Optuna optimization algorithm to construct a mapping relationship from input to output.

[0011] Step 4: Perform spline interpolation on the output of the proxy model and then perform reverse sampling to restore it to its original dimension, thereby obtaining the seismic response of each floor of the building calculated based on the proxy model;

[0012] Step 5: Compare the calculation results of the proxy model with those of the physical model to evaluate the accuracy and calculation efficiency of the proxy model.

[0013] Furthermore, the physical model is an MDOF shear model.

[0014] Furthermore, the MDOF shear model is established based on the finite element analysis software OpenSees, and the “TwoNodeLink” unit is used to establish the shear spring. The mechanical parameters of the shear spring include the skeleton curve and the stress pinching coefficient.

[0015] Furthermore, in step 2, N s The seismic response of the building is processed to a predetermined dimension, and the specific implementation method is as follows: ① The floor number x is 1 and the seismic response y of each layer 1 As independent and dependent variables, fit the continuous function f based on spline interpolation 1 (x), where ② Generate an equidistant sampling sequence x based on equation 1 2 , in f 1 (x) is sampled to obtain a one-dimensional vector y of length N 2 ; With earthquake motion parameters and building parameters as input and N earthquake response values ​​after equidistant sampling as output, the data set can be constructed;

[0016]

[0017] Furthermore, the dataset is divided into training set, validation set and test set in a ratio of 8:1:1.

[0018] Furthermore, the step 3 is specifically as follows:

[0019] (1) Select the parameters to be tuned and set their adjustment range;

[0020] (2) Set the objective function, optimization method, and optimization times, where the objective function is the decision coefficient R of the proxy model's prediction results on the validation set. 2 , the optimization method is maximization, and the number of optimizations is 1000, that is, the optimization goal is to make the proxy model achieve the maximum R on the validation set. 2 ; R 2 Calculated by formula (2), where y i , They are the values ​​calculated by the physical model, the values ​​calculated by the proxy model, and the average value of the values ​​calculated by the physical model;

[0021]

[0022] (3) Use the training set to train the proxy model, adjust the parameters with the goal of maximizing the accuracy of the proxy model on the validation set, take the parameter values ​​that make the proxy model achieve the highest accuracy on the validation set as the optimal parameters, and establish the final proxy model based on the optimal parameters.

[0023] Furthermore, in step 4, the building parameters and ground motion parameters of each sample in the test set are input into the proxy model to obtain the output of the proxy model. is a one-dimensional vector with a length of N, and then through spline interpolation and reverse sampling Restore to the original dimension, length N s , and obtain the calculation results of the seismic response of each floor of the building The specific implementation is as follows: (1) Take the sampling sequence x in step 2 2 As an independent variable, As the dependent variable, the spline interpolation fitting function f 2 (x); (2) Floor number x 1 As a sampling sequence, in the function f 2 (x) upsamples to get

[0024] Furthermore, in step 5, the accuracy index evaluated is the determination coefficient R 2 and mean relative error MRE, MRE is calculated by formula (3), where n is the number of samples;

[0025]

[0026] The present invention also proposes an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning are implemented.

[0027] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning.

[0028] Beneficial effects of the present invention:

[0029] 1. The prediction results are detailed and have a wider scope of application. When the research object is a building complex, the existing proxy model can only calculate the maximum value of the response of all layers of the building or the seismic response of all layers of a building with a specific number of floors, while the present invention can calculate the seismic response of all layers of a building with any number of floors.

[0030] 2. It has high precision, high efficiency and low cost. The calculation accuracy is close to that of the physical model, while the calculation efficiency is thousands of times that of the physical model. It has low requirements for computer configuration and can complete the training and optimization process of the proxy model within a few hours using an ordinary personal computer.

[0031] 3. The proxy model has low training difficulty and is easy to use. It adopts Bayesian automatic parameter adjustment technology and does not require manual adjustment. Compared with the exhaustive parameter adjustment method (grid search), it has higher optimization efficiency and can optimize the calculation accuracy of the proxy model to an acceptable range in a short time. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0033] Figure 1 This is a flow chart of the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning according to the present invention.

[0034] Figure 2 Schematic diagram of the multi-degree-of-freedom (MDOF) shear model, where (a) is the model architecture and (b) is the mechanical model.

[0035] Figure 3 This is a flowchart for unifying the dimension of building seismic response data based on spline interpolation and equidistant sampling.

[0036] Figure 4 This is a flowchart for obtaining the calculation results of the seismic response of each floor of a building based on spline interpolation and reverse sampling.

[0037] Figure 5 This is a schematic diagram of the calculation results (inter-story displacement angle) of the proxy model on the test set.

[0038] Figure 6 Schematic diagram of the calculation results (peak layer acceleration) of the proxy model on the test set. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the 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 creative work are within the scope of protection of the present invention.

[0040] Combination Figure 1-Figure 6The present invention proposes a method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning, and the method comprises the following steps:

[0041] Step 1: Select some buildings with different numbers of floors and a certain number of earthquake records, establish a physical model of the building (multi-degree-of-freedom shear model), use the earthquake records as external excitation to perform dynamic analysis on the physical model, and obtain the seismic response of all floors of each building under each earthquake;

[0042] 200 reinforced concrete frame buildings were selected, including 50 buildings with 7 to 10 floors each, and then a physical model of each building was established. The physical model used in the present invention is the MDOF shear model, which is suitable for simulating earthquake damage to medium and low-rise buildings with regular plane shapes. Figure 2 As shown in (a), the MDOF model is established based on the finite element analysis software OpenSees (Open System for Earthquake Engineering Simulation) and uses the "TwoNode Link" unit. Figure 2 (a) The mechanical parameters of the shear spring include the skeleton curve and the stress pinching coefficient, as shown in Figure 2 (b) is shown. 100 earthquake motions were downloaded from the Pacific Earthquake Engineering Research Center (PEER) website. Each earthquake motion was scaled 5 times. Each time the scale was made to make the peak acceleration (PGA) of the earthquake motion a random value between 0.05g and 0.8g. After scaling, 500 earthquake motion records were obtained. Each earthquake motion record was matched with 50 randomly selected MDOF models and dynamic analysis was performed to obtain the inter-story displacement angle (IDR) and peak story acceleration (PFA) of all floors of the building.

[0043] Step 2: Calculate the seismic parameters based on the seismic records, and use the seismic parameters and building parameters as the proxy model input; process the seismic responses of all layers of each building into the same dimension as the proxy model output through spline interpolation and equidistant sampling methods; thereby construct the data set required for training the proxy model;

[0044] In step 2, the seismic parameters are calculated based on the seismic records. The required parameters and their calculation methods are shown in Table 1. The building period, number of floors and mechanical parameters of the bottom shear spring are used as building parameters. The selected mechanical parameters of the shear spring include the coordinates of the yield point, peak point and limit point on the skeleton curve, as well as the stress pinching coefficient. Then, the seismic responses of all buildings are processed into a unified dimension through spline interpolation and equidistant sampling. For example, N s The seismic response of the building is processed to a predetermined dimension, and the specific implementation method is as follows: ① The floor number x is 1 and the seismic response y of each layer 1As independent and dependent variables, fit the continuous function f based on spline interpolation 1 (x), where ② Generate an equidistant sampling sequence x based on equation 1 2 , in f 1 (x) is sampled to obtain a one-dimensional vector y of length N 2 ; With the seismic parameters and building parameters as input and N seismic response values ​​after equidistant sampling as output, the data set can be constructed; see the specific process Figure 3 .

[0045]

[0046] Table 1 Detailed explanation of ground motion parameters

[0047]

[0048]

[0049] Step 3: Divide the data set in step 2 into a training set, a validation set, and a test set in a specific ratio, which are used for proxy model training, parameter adjustment, and accuracy evaluation, respectively. Then, a proxy model is established based on the XGBoost machine learning algorithm and the Optuna optimization algorithm to construct a mapping relationship from input to output.

[0050] The dataset is divided into training set, validation set and test set in a ratio of 8:1:1. A proxy model is established based on the XGBoost regression algorithm, and the parameters of the proxy model are adjusted using the Optuna toolkit to obtain higher calculation accuracy. The specific steps are as follows:

[0051] (1) Select the parameters to be tuned and set their adjustment ranges, as shown in Table 2;

[0052] Table 2 Parameters to be tuned and their value ranges

[0053]

[0054] (2) Set the objective function, optimization method, and optimization times, where the objective function is the decision coefficient R of the proxy model's prediction results on the validation set. 2 , the optimization method is maximization, and the number of optimizations is 1000, that is, the optimization goal is to make the proxy model achieve the maximum R on the validation set. 2 ; R 2 Calculated by formula (2), where y i , They are the values ​​calculated by the physical model, the values ​​calculated by the proxy model, and the average value of the values ​​calculated by the physical model;

[0055]

[0056] (3) Use the training set to train the proxy model, adjust the parameters with the goal of maximizing the accuracy of the proxy model on the validation set, take the parameter values ​​that make the proxy model achieve the highest accuracy on the validation set as the optimal parameters, and establish the final proxy model based on the optimal parameters.

[0057] Step 4: Perform spline interpolation on the output of the proxy model and then perform reverse sampling to restore it to its original dimension, thereby obtaining the seismic response of each floor of the building calculated based on the proxy model;

[0058] In step 4, the building parameters and ground motion parameters of each sample in the test set are input into the proxy model to obtain the output of the proxy model. is a one-dimensional vector with a length of N, and then through spline interpolation and reverse sampling Restore to the original dimension, length N s , and obtain the calculation results of the seismic response of each floor of the building The specific implementation is as follows: (1) Take the sampling sequence x in step 2 2 As an independent variable, As the dependent variable, the spline interpolation fitting function f 2 (x); (2) Floor number x 1 As a sampling sequence, in the function f 2 (x) upsamples to get For detailed process, see Figure 4 .

[0059] Step 5: Compare the calculation results of the proxy model with those of the physical model to evaluate the accuracy and calculation efficiency of the proxy model.

[0060] In step 5, the accuracy indicator evaluated is the determination coefficient R 2 and mean relative error MRE, MRE is calculated by formula (3), where n is the number of samples; the calculation results of the proxy model on the test set (inter-story displacement angle) and the calculation results of the proxy model on the test set (peak story acceleration) are shown in Figure 5 and Figure 6 .

[0061]

[0062] The present invention also proposes an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning are implemented.

[0063] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning.

[0064] The memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DRRAM). It should be noted that the memory of the method described in the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0065] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a high-density digital video disc (DVD)), or a semiconductor medium (eg, a solid state disc (SSD)).

[0066] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software. The steps of the method disclosed in conjunction with the embodiment of the present application can be directly embodied as a hardware processor for execution, or a combination of hardware and software modules in a processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in conjunction with its hardware. To avoid repetition, it is not described in detail here.

[0067] It should be noted that the processor in the embodiment of the present application can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiment can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor can be combined and performed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0068] The above is a detailed introduction to the method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning proposed in the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for calculating the seismic response of each floor of a building based on spline interpolation and machine learning, characterized in that: The method comprises the following steps: Step 1: Select some buildings with different numbers of floors and a certain number of earthquake records, establish a physical model of the building, use the earthquake records as external excitation to perform dynamic analysis on the physical model, and obtain the seismic response of all floors of each building under each earthquake; Step 2: Calculate the seismic parameters based on the seismic records, and use the seismic parameters and building parameters as the proxy model input; process the seismic responses of all layers of each building into the same dimension as the proxy model output through spline interpolation and equidistant sampling methods; thereby construct the data set required for training the proxy model; Step 3: Divide the data set in step 2 into a training set, a validation set, and a test set in a specific ratio, which are used for proxy model training, parameter adjustment, and accuracy evaluation, respectively. Then, a proxy model is established based on the XGBoost machine learning algorithm and the Optuna optimization algorithm to construct a mapping relationship from input to output. Step 4: Perform spline interpolation on the output of the proxy model and then perform reverse sampling to restore it to its original dimension, thereby obtaining the seismic response of each floor of the building calculated based on the proxy model; Step 5: Compare the calculation results of the proxy model with those of the physical model to evaluate the accuracy and calculation efficiency of the proxy model.

2. The method according to claim 1, characterized in that: The physical model is an MDOF shear model.

3. The method according to claim 2, characterized in that The MDOF shear model is established based on the finite element analysis software OpenSees, and the "TwoNode Link" unit is used to establish the shear spring. The mechanical parameters of the shear spring include the skeleton curve and the stress pinching coefficient.

4. The method according to claim 1, characterized in that: In step 2, N s The seismic response of a building with multiple floors is processed to a predetermined dimension. The specific implementation method is as follows: ① The floor number x1 and the seismic response y1 of each floor are used as independent variables and dependent variables respectively, and a continuous function f1(x) is fitted based on spline interpolation, where ② Generate an equidistant sampling sequence x2 based on equation 1, and sample at f1(x) to obtain a one-dimensional vector y2 of length N; take the seismic parameters and building parameters as input, and the N seismic response values ​​after equidistant sampling as output to construct the data set; 5. The method according to claim 1, characterized in that The dataset is divided into training set, validation set, and test set in a ratio of 8:1:

1.

6. The method according to claim 5, characterized in that The step 3 is specifically as follows: (1) Select the parameters to be tuned and set their adjustment range; (2) Set the objective function, optimization method, and optimization times, where the objective function is the decision coefficient R of the proxy model's prediction results on the validation set. 2 , the optimization method is maximization, and the number of optimizations is 1000, that is, the optimization goal is to make the proxy model achieve the maximum R on the validation set. 2 ; R 2 Calculated by formula (2), where y i , They are the values ​​calculated by the physical model, the values ​​calculated by the proxy model, and the average value of the values ​​calculated by the physical model; (3) Use the training set to train the proxy model, adjust the parameters with the goal of maximizing the accuracy of the proxy model on the validation set, take the parameter values ​​that make the proxy model achieve the highest accuracy on the validation set as the optimal parameters, and establish the final proxy model based on the optimal parameters.

7. The method according to claim 6, characterized in that In step 4, the building parameters and ground motion parameters of each sample in the test set are input into the proxy model to obtain the output of the proxy model. is a one-dimensional vector with a length of N, and then through spline interpolation and reverse sampling Restore to the original dimension, length N s , and obtain the calculation results of the seismic response of each floor of the building The specific implementation is as follows: (1) Take the sampling sequence x2 in step 2 as the independent variable, As the dependent variable, the function f2(x) is fitted based on spline interpolation; (2) the floor number x1 is used as the sampling sequence, and sampling is performed on the function f2(x), and we get 8. The method according to claim 7, characterized in that In step 5, the accuracy indicator evaluated is the determination coefficient R 2 and mean relative error MRE, MRE is calculated by formula (3), where n is the number of samples; 9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium for storing computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.