CN-121980846-A - Construction method, device and equipment of wading underground structure seismic response prediction model

CN121980846ACN 121980846 ACN121980846 ACN 121980846ACN-121980846-A

Abstract

The application is suitable for the technical field of structural seismic engineering, and provides a construction method, a device and equipment of a wading underground structure seismic response prediction model. Then, based on a plurality of seismic sample feature vectors extracted from the plurality of seismic sample data, cluster analysis is performed on the plurality of seismic sample data to obtain a training data set. And finally, training the initial earthquake response prediction model by taking target earthquake motion sample data and water depth parameters in the training sample data as input data and taking earthquake response data of a plurality of monitoring points in the training sample data as labels of the input data to obtain the wading underground structure earthquake response prediction model. The prediction precision and the prediction efficiency of the multi-monitoring-point earthquake response of the wading underground structure are improved through the method.

Inventors

GUO YUTAO
CHEN YIMU
HU ZHENZHONG
HOU CHAO
Ai Hamaide
YU YANTAO

Assignees

清华大学深圳国际研究生院

Dates

Publication Date: 20260505
Application Date: 20251222

Claims (10)

1. The construction method of the wading underground structure earthquake response prediction model is characterized by comprising the following steps of: Generating a seismic response database based on simulation results of a structure-soil-fluid coupling finite element model, wherein the structure-soil-fluid coupling finite element model is used for simulating seismic responses of a plurality of monitoring points in a wading underground structure under working conditions of different earthquake motion intensities and water depths; Performing cluster analysis on the plurality of seismic sample data based on a plurality of seismic sample feature vectors extracted from the plurality of seismic sample data to obtain a training data set, wherein the training data set comprises a plurality of training sample data, each training sample data comprises target seismic sample data, the water depth parameter and seismic response data of a plurality of monitoring points corresponding to the target seismic sample data, and the target seismic sample data is data which is determined from the plurality of seismic sample data and serves as a training sample; and training an initial earthquake response prediction model by taking the target earthquake motion sample data and the water depth parameters in the training sample data as input data and taking the earthquake response data of a plurality of monitoring points in the training sample data as labels of the input data to obtain a wading underground structure earthquake response prediction model, wherein the initial earthquake response prediction model is a prediction model constructed based on an attention mechanism, a long-short-term memory neural network and a characteristic modulation module.
2. The method for constructing a prediction model of the seismic response of the wading subsurface structure according to claim 1, wherein the generating the seismic response database based on the simulation result of the structure-soil-fluid coupling finite element model comprises: Constructing the structure-soil-fluid coupling finite element model based on a two-dimensional plane strain model; Normalizing and time-aligning a plurality of actually measured earthquake motion data in a preset actually measured earthquake motion database to obtain a plurality of earthquake motion sample data; Inputting a plurality of seismic sample data into the structure-soil-fluid coupling finite element model, and performing simulation under the condition of different water depth parameters to obtain a corresponding seismic response simulation result; and extracting a plurality of seismic response data of a plurality of monitoring points in the wading underground structure from the seismic response simulation result to obtain the seismic response database.
3. The method for constructing a prediction model of the seismic response of the wading subsurface structure according to claim 2, wherein the constructing the structure-soil-fluid coupling finite element model based on the two-dimensional plane strain model comprises: Based on the two-dimensional plane strain model, constructing and obtaining an initial finite element model according to the cross section of the wading underground structure, a soil body area and a fluid area corresponding to the cross section of the wading underground structure; in the initial finite element model, a West God additional mass model is adopted to represent the dynamic water pressure of a water body, and an encryption grid division is adopted to represent the cross section and the surrounding area of the wading underground structure; And setting springs and damping at the boundary of the initial finite element model, loading earthquake motion at the bottom and two sides of the initial finite element model in an equivalent node force mode, and constructing to obtain the structure-soil-fluid coupling finite element model.
4. The method for constructing a prediction model of seismic response of a wading subsurface structure according to claim 1, wherein the performing cluster analysis on the plurality of seismic sample data based on a plurality of seismic sample feature vectors extracted from the plurality of seismic sample data to obtain a training data set comprises: Extracting a plurality of seismic amplitude characteristics, a plurality of seismic frequency spectrum characteristics and a plurality of seismic energy characteristics from each seismic sample data to obtain the seismic sample characteristic vector of each seismic sample data; based on the seismic sample feature vector of each seismic sample data, carrying out cluster analysis on a plurality of seismic sample data through a hierarchical clustering algorithm to obtain a clustering result corresponding to the seismic sample data; and carrying out hierarchical sampling division on the clustering result, and determining a plurality of target seismic sample data from a plurality of seismic sample data to obtain the training data set.
5. The method for constructing a prediction model of seismic response of a wading subsurface structure according to claim 4, wherein said extracting a plurality of seismic amplitude features, a plurality of seismic spectrum features, and a plurality of seismic energy features from each of said seismic sample data to obtain said seismic sample feature vector for each of said seismic sample data comprises: extracting a plurality of seismic amplitude characteristics in each seismic sample data based on acceleration time courses in each seismic sample data; Extracting a plurality of seismic vibration spectrum characteristics of each seismic vibration sample data from an acceleration response spectrum; obtaining a plurality of seismic energy characteristics of each seismic sample data by calculating an energy ratio in each seismic sample data; And obtaining the seismic sample feature vector of each seismic sample data based on a plurality of the seismic amplitude features, a plurality of the seismic frequency spectrum features and a plurality of the seismic energy features.
6. The method for constructing a prediction model of seismic response of a wading subsurface structure according to claim 4, wherein the clustering analysis of the plurality of seismic sample data by a hierarchical clustering algorithm based on the seismic sample feature vector of each of the seismic sample data to obtain a plurality of clustering results corresponding to the seismic sample data comprises: calculating the Euclidean distance between any two pieces of the seismic sample data based on the seismic sample feature vector of each piece of the seismic sample data; Performing first clustering analysis on all the seismic sample data according to the Euclidean distance and the minimized inter-class square error increment to obtain an initial clustering result; calculating a feature importance index of each feature vector in the seismic sample feature vectors according to the inter-category variance and the intra-category variance in the initial clustering result; removing the feature vector lower than a preset contribution value in the seismic sample feature vector according to the feature importance index to obtain a processed seismic sample feature vector; And performing second clustering analysis on all the seismic sample data according to the processed seismic sample feature vector until the clustering result meets a preset clustering termination condition to obtain the clustering result.
7. The method for constructing a prediction model of a seismic response of a wading subsurface structure according to claim 1, wherein the training the initial seismic response prediction model with the target seismic sample data and the water depth parameter in the training sample data as input data and the seismic response data of a plurality of monitoring points in the training sample data as labels of the input data, comprises: Extracting feature representations of a plurality of subspaces from the seismic vibration sample data through the attention mechanism, and splicing to obtain global feature vectors; converting the global feature vector according to a modulation coefficient to obtain a fusion feature vector, wherein the modulation coefficient is generated by the feature modulation module according to the water depth parameter; generating predicted seismic response data of a plurality of monitoring points of the wading underground structure according to the fusion feature vector through the long-short-term memory neural network; Calculating loss values between the seismic response data and the predicted seismic response data for a plurality of the monitoring points; And adjusting model parameters of the initial earthquake response prediction model based on the loss value, returning to execute the step of extracting feature representations of a plurality of subspaces from the earthquake motion sample data through the attention mechanism, and splicing to obtain global feature vectors and subsequent steps until a preset iteration round is completed to obtain the wading underground structure earthquake response prediction model.
8. The construction device of the wading underground structure earthquake response prediction model is characterized by comprising the following components: The system comprises a first building module, a first seismic response database, a second building module, a third building module, a fourth building module and a fourth building module, wherein the first building module is used for generating a seismic response database based on simulation results of a structure-soil-fluid coupling finite element model, the structure-soil-fluid coupling finite element model is used for simulating seismic response of a plurality of monitoring points in a wading underground structure under working conditions of different earthquake intensity and water depths; The system comprises a first construction module, a second construction module and a storage module, wherein the first construction module is used for carrying out cluster analysis on a plurality of seismic sample feature vectors extracted from a plurality of seismic sample data to obtain a training data set, the training data set comprises a plurality of training sample data, each training sample data comprises a target seismic sample data, a water depth parameter and seismic response data of a plurality of monitoring points corresponding to the target seismic sample data, and the target seismic sample data is data which is determined from the plurality of seismic sample data and serves as a training sample; The third construction module is used for training an initial earthquake response prediction model by taking the target earthquake motion sample data and the water depth parameters in the training sample data as input data and taking the earthquake response data of a plurality of monitoring points in the training sample data as labels of the input data to obtain a wading underground structure earthquake response prediction model, wherein the initial earthquake response prediction model is a prediction model constructed based on an attention mechanism, a long-short-time memory neural network and a characteristic modulation module.
9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the method according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a processor, implements the method according to any of claims 1 to 7.

Description

Construction method, device and equipment of wading underground structure seismic response prediction model Technical Field The application belongs to the technical field of structural seismic engineering, and particularly relates to a construction method, a construction device and construction equipment of a wading underground structure seismic response prediction model. Background The wading underground structure is an underground engineering structure which is partially or completely positioned below, inside or in direct contact with a water body. Wading underground structures (such as submarine tunnels, subway at the bottom of a river, coastal pipe galleries, river-crossing channels and the like) are used as important traffic and urban infrastructure and are usually in complex ocean or underground water environments and are influenced by the coupling effects of earthquake motion, water body hydrodynamic pressure and foundation nonlinear characteristics for a long time. Once a strong shock occurs, damage to the subsurface structure may result in serious secondary disasters and economic losses. At present, a deep learning model can improve prediction efficiency and reduce manual intervention by learning a nonlinear mapping relation between earthquake motion input and structural response, and is preliminarily applied to earthquake resistance analysis of an overground structure. However, in the application of the underground wading structure, the existing deep learning model does not fully consider the water coupling effect and the prediction model is complex, so that the earthquake response prediction structure has deviation and low prediction efficiency. Disclosure of Invention The embodiment of the application provides a construction method, a device and equipment of a wading underground structure earthquake response prediction model, which can solve the technical problems of low earthquake response prediction precision and low efficiency caused by insufficient consideration of water coupling and complex prediction model in the existing wading underground structure earthquake response prediction method. In a first aspect, an embodiment of the present application provides a method for constructing a seismic response prediction model of a wading subsurface structure, where the method includes: Generating a seismic response database based on simulation results of a structure-soil-fluid coupling finite element model, wherein the structure-soil-fluid coupling finite element model is used for simulating seismic responses of a plurality of monitoring points in a wading underground structure under working conditions of different earthquake motion intensities and water depths; Performing cluster analysis on the plurality of seismic sample data based on a plurality of seismic sample feature vectors extracted from the plurality of seismic sample data to obtain a training data set, wherein the training data set comprises a plurality of training sample data, each training sample data comprises target seismic sample data, the water depth parameter and seismic response data of a plurality of monitoring points corresponding to the target seismic sample data, and the target seismic sample data is data which is determined from the plurality of seismic sample data and serves as a training sample; and training an initial earthquake response prediction model by taking the target earthquake motion sample data and the water depth parameters in the training sample data as input data and taking the earthquake response data of a plurality of monitoring points in the training sample data as labels of the input data to obtain a wading underground structure earthquake response prediction model, wherein the initial earthquake response prediction model is a prediction model constructed based on an attention mechanism, a long-short-term memory neural network and a characteristic modulation module. In a possible implementation manner of the first aspect, the generating a seismic response database based on simulation results of the structure-soil-fluid coupled finite element model includes: Constructing the structure-soil-fluid coupling finite element model based on a two-dimensional plane strain model; Normalizing and time-aligning a plurality of actually measured earthquake motion data in a preset actually measured earthquake motion database to obtain a plurality of earthquake motion sample data; Inputting a plurality of seismic sample data into the structure-soil-fluid coupling finite element model, and performing simulation under the condition of different water depth parameters to obtain a corresponding seismic response simulation result; and extracting a plurality of seismic response data of a plurality of monitoring points in the wading underground structure from the seismic response simulation result to obtain the seismic response database. In a possible implementation manner of the first aspect, the constructing the structure-so