Bridge earthquake vulnerability prediction method and system based on machine learning

Abstract

The invention discloses a bridge earthquake vulnerability prediction method and system based on machine learning, and the method comprises the steps: obtaining an earthquake parameter set, employing a deep neural network to construct a nonlinear mapping model, carrying out the vulnerability probability prediction based on the nonlinear mapping model, and obtaining initial vulnerability probability data; the method comprises the steps of obtaining bridge structure data to construct a bridge geometric model, analyzing segmented material heterogeneity of the bridge geometric model, and if the heterogeneity exceeds a preset threshold value, calculating segmented response data under the earthquake action; based on the segmented response data, analyzing spatial difference vulnerability probability distribution of the whole bridge, and obtaining a fine prediction result; and if the deviation between the fine prediction result and the initial vulnerability probability exceeds a preset threshold, adjusting the nonlinear mapping model, and outputting the updated vulnerability probability distribution to obtain a final vulnerability prediction result. According to the method, the accuracy and reliability of bridge earthquake vulnerability prediction are remarkably improved, and the method has important significance in improving the earthquake safety of a bridge structure.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge seismic vulnerability prediction, and in particular to a bridge seismic vulnerability prediction method and system based on machine learning. Background Art

[0002] Bridge earthquake engineering is an important branch of civil engineering, which is related to infrastructure safety and social stability. Its research is of key significance to protecting the safety of life and property and maintaining the function of transportation networks. With the frequent occurrence of earthquake disasters, accurate prediction of the seismic vulnerability of bridges has become an urgent need, which directly affects the efficiency of post-disaster rescue and recovery. However, existing prediction methods have obvious shortcomings in accuracy and efficiency. Many traditional methods rely on simplified models or empirical formulas, which make it difficult to fully capture the complex nonlinear relationship between seismic motion parameters and the probability of bridge failure, resulting in large deviations in prediction results. In addition, some methods do not fully consider the heterogeneity of bridge structures, ignore the unique response characteristics of different parts in earthquakes, and limit the degree of refinement of predictions.

[0003] In the prediction of bridge seismic vulnerability, the core challenges focus on how to effectively correlate seismic motion parameters with failure probability, and how to deal with the complex characteristics of bridge structures. Seismic motion parameters are diverse and mutually coupled, making it difficult to construct efficient mapping models with traditional methods, limiting prediction accuracy. At the same time, bridge structures exhibit significant spatial differences due to differences in geometry, material distribution, and connection methods. Overall analysis often obscures the vulnerability characteristics of local weak links. If feature extraction and prediction cannot be performed for structural segments, the overall results may be distorted. Therefore, how to construct an efficient prediction model by learning the correlation between seismic motion parameters and failure probability, and perform refined predictions based on the segmented characteristics of bridge structures, has become a key issue in improving the accuracy of bridge seismic vulnerability prediction. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a bridge seismic vulnerability prediction method based on machine learning to overcome the limitations of traditional methods that ignore the heterogeneity of bridge structures and provide technical support for bridge seismic design and reinforcement.

[0005] To achieve the above objectives, the present invention provides a bridge seismic vulnerability prediction method based on machine learning, comprising:

[0006] Obtaining a set of earthquake parameters, building a nonlinear mapping model using a deep neural network, and performing vulnerability probability prediction based on the nonlinear mapping model to obtain preliminary vulnerability probability data;

[0007] Acquiring bridge structural data to construct a bridge geometric model, analyzing segmented material heterogeneity of the bridge geometric model, and calculating segmented response data under earthquake action if the heterogeneity exceeds a preset threshold;

[0008] Based on the segmented response data, the spatially differentiated vulnerability probability distribution of the entire bridge is analyzed to obtain a refined prediction result;

[0009] If the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result.

[0010] Preferably, obtaining the earthquake parameter set includes:

[0011] Obtain multi-dimensional earthquake motion data from the earthquake motion parameter database and perform standardization on the data;

[0012] The principal component analysis algorithm is used to extract features from the standardized seismic motion dataset, separate the key feature acceleration values, frequency distribution and duration, and obtain the earthquake parameter set.

[0013] Preferably, acquiring bridge structure data and constructing a bridge geometric model includes:

[0014] The bridge structure data is obtained from the bridge structure database and the bridge structure data is cleaned;

[0015] Perform feature extraction on the cleaned bridge structure data to obtain geometric shape features, material distribution features, and connection mode features, and perform dimensionality reduction through principal component analysis to obtain an optimized feature set.

[0016] Extracting characteristic boundary conditions from the optimized feature set by a finite element analysis algorithm;

[0017] The mechanical parameters of the optimized feature set are calculated according to the characteristic boundary conditions, and a bridge geometric model is constructed according to the mechanical parameters.

[0018] Preferably, analyzing the segmented material heterogeneity of the bridge geometric model comprises:

[0019] Obtaining material data of each section from the bridge geometric model, unifying the format and filling in missing values ​​to obtain a standardized material data set;

[0020] Meshing the standardized material data set, decomposing the geometric boundary into calculation units, generating boundary conditions in combination with connection constraints, and obtaining meshed units;

[0021] Performing finite element calculation on the gridded unit to obtain segmented heterogeneity parameters;

[0022] The segmented heterogeneity parameters include material stiffness, damping characteristics, and mass distribution data.

[0023] Preferably, calculating the segmented response data under earthquake action includes:

[0024] Based on the segmented heterogeneity parameters, a long short-term memory network is used to calculate the displacement distribution data, velocity distribution data, and stress distribution data of each segment of the structure under earthquake action;

[0025] The segmented response data is obtained based on the displacement distribution data, velocity distribution data, and stress distribution data.

[0026] Preferably, obtaining the refined prediction result includes:

[0027] Using a machine learning classification algorithm to classify the segmented response data, and obtain a segmented vulnerability probability distribution set;

[0028] A spatial interpolation algorithm is used to perform spatial interpolation on the segmented vulnerability probability distribution set to generate the spatially differentiated vulnerability probability distribution of the entire bridge and obtain a refined prediction result.

[0029] Preferably, classifying the segmented response data includes:

[0030] Preprocessing the segmented response data to obtain a segmented structure feature set;

[0031] Classify the segmented structure feature set based on support vector machine, identify potential weak points, and obtain weak point distribution data;

[0032] Based on the weak point distribution data, the probability density estimation algorithm is used to calculate the vulnerability probability of each weak point;

[0033] The fragility probability of each weak point and the stress distribution data are integrated to obtain a segmented fragility probability distribution set.

[0034] Preferably, spatial interpolation is performed on the segmented fragility probability distribution set to generate a spatially differentiated fragility probability distribution of the entire bridge, thereby obtaining a refined prediction result, including:

[0035] Obtaining local probability data from a piecewise fragility probability distribution set;

[0036] Based on local probability data, a data screening algorithm is used to screen high-probability areas and obtain a local high-risk data set;

[0037] Based on the local high-risk data set, a spatial interpolation algorithm is used to calculate the probability transition value between each area to obtain the spatially differentiated vulnerability probability distribution of the entire bridge.

[0038] The spatially differentiated vulnerability probability distribution of the entire bridge is spatially smoothed and optimized to obtain precise prediction results.

[0039] Preferably, if the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result, including:

[0040] Calculating the difference between the refined prediction result and the preliminary vulnerability probability data using a deviation detection algorithm to obtain a difference value;

[0041] If the difference value exceeds the preset threshold, the high deviation area is filtered out through the data screening method to obtain the high deviation data set;

[0042] According to the high-deviation data set, adjusting the parameters of the nonlinear mapping model by a parameter optimization algorithm, updating the mapping relationship by iterative calculation, and obtaining an adjusted nonlinear mapping model;

[0043] The earthquake parameter set is re-input into the adjusted nonlinear mapping model to perform vulnerability probability distribution, and a final vulnerability prediction result is obtained.

[0044] A bridge seismic vulnerability prediction system based on machine learning, comprising:

[0045] A seismic parameter acquisition module is used to construct a nonlinear mapping model using a deep neural network, perform vulnerability probability prediction based on the nonlinear mapping model, and obtain preliminary vulnerability probability data;

[0046] a bridge model construction module, configured to acquire bridge structural data to construct a bridge geometric model, analyze the segmented material heterogeneity of the bridge geometric model, and calculate segmented response data under earthquake action if the heterogeneity exceeds a preset threshold;

[0047] A refined result prediction module is used to analyze the spatial difference vulnerability probability distribution of the entire bridge based on the segmented response data to obtain a refined prediction result;

[0048] The update prediction module is used to determine whether the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold. If the deviation exceeds the threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result.

[0049] Compared with the prior art, the present invention has the following advantages and technical effects:

[0050] The present invention discloses a bridge seismic vulnerability prediction method based on machine learning. The method comprises the following steps: obtaining an earthquake parameter set, constructing a nonlinear mapping model using a deep neural network, performing vulnerability probability prediction based on the nonlinear mapping model, and obtaining preliminary vulnerability probability data; obtaining bridge structural data to construct a bridge geometric model, analyzing the segmented material heterogeneity of the bridge geometric model, and calculating the segmented response data under earthquake action; analyzing the spatially differentiated vulnerability probability distribution of the entire bridge, and obtaining a refined prediction result; and adjusting the nonlinear mapping model if the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, outputting an updated vulnerability probability distribution, and obtaining a final vulnerability prediction result. This method combines a nonlinear mapping model for vulnerability prediction. The key is to analyze the segmented material heterogeneity of the bridge geometric model, obtain segmented response data, and then perform fine prediction. The spatial interpolation algorithm used in the fine prediction adopts Kriging interpolation, which can generate the spatially differentiated vulnerability probability distribution of the entire bridge, providing fine prediction results for the seismic design and maintenance of the bridge. It overcomes the limitation of traditional methods that ignore the heterogeneity of bridge structures, and compares and analyzes the fine prediction results with the preliminary vulnerability probability, adjusts the nonlinear mapping model, and significantly improves the accuracy and reliability of bridge seismic vulnerability prediction. The present invention is of great significance to improving the seismic safety of bridge structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0052] Figure 1 The figure is a flow chart of a bridge seismic vulnerability prediction method based on machine learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0054] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0055] like Figure 1 As shown, this embodiment proposes a bridge seismic vulnerability prediction method based on machine learning, including:

[0056] Obtain a set of earthquake parameters, use a deep neural network to build a nonlinear mapping model, perform vulnerability probability prediction based on the nonlinear mapping model, and obtain preliminary vulnerability probability data;

[0057] Obtain bridge structural data to construct a bridge geometric model, analyze the segmented material heterogeneity of the bridge geometric model, and calculate the segmented response data under earthquake action if the heterogeneity exceeds a preset threshold;

[0058] Based on the segmented response data, the spatially differentiated vulnerability probability distribution of the entire bridge is analyzed to obtain detailed prediction results.

[0059] If the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted to output the updated vulnerability probability distribution to obtain the final vulnerability prediction result.

[0060] Furthermore, obtaining the earthquake parameter set includes: obtaining multidimensional earthquake motion data from a earthquake motion parameter database and standardizing the data; using a principal component analysis algorithm to extract features from the standardized earthquake motion data set, separating key feature acceleration values, frequency distribution and duration, and obtaining the earthquake parameter set.

[0061] Specifically, in this embodiment, a seismic monitoring station recorded 1,000 sets of earthquake events, each set containing triaxial acceleration data with a sampling frequency of 100 Hz and a duration of 60 seconds. Multidimensional seismic motion data was obtained from a seismic parameter database, and normalization processing techniques were used to clean the data to remove noise and outliers. The principal component analysis algorithm was used for feature extraction. Assuming that the data set contains features such as acceleration peak value, frequency distribution, and duration, principal component analysis projects high-dimensional data into a low-dimensional space through linear transformation, retaining the main information.

[0062] Furthermore, a deep neural network is used to construct a nonlinear mapping model;

[0063] Specifically, in this embodiment, the model architecture: it includes a stem module and a body module; the stem module uses multi-path depth-separable convolution and linear transformation to quickly reduce the time dimension of the earthquake record and reduce the subsequent computational overhead; the body module adopts a four-stage design, and focuses on local and global features at the same time through a multi-head self-attention mechanism; training and optimization: during the training process, data enhancement methods play an important role in model performance. The data set can be randomly split, such as 8:1:1 split into training set, test set and validation set. At the same time, attention should be paid to the balance between model depth and computational complexity and overfitting risk; the model can be used for a variety of earthquake monitoring tasks such as earthquake detection, earthquake phase picking, first motion polarity classification, magnitude estimation, back azimuth estimation and epicenter distance estimation;

[0064] Furthermore, obtaining bridge structure data to construct a bridge geometric model includes: obtaining bridge structure data from a bridge structure database, and performing data cleaning on the bridge structure data; performing feature extraction on the cleaned bridge structure data to obtain geometric shape features, material distribution features, and connection mode features, and performing dimensionality reduction through principal component analysis to obtain an optimized feature set; extracting characteristic boundary conditions from the optimized feature set through a finite element analysis algorithm; calculating the mechanical parameters of the optimized feature set based on the characteristic boundary conditions, and constructing a bridge geometric model based on the mechanical parameters.

[0065] Specifically, in this embodiment, data cleaning includes filtering out outliers and noise: ensuring the accuracy and reliability of data through filtering and data calibration; data synchronization and calibration: ensuring the time consistency and accuracy of data from different sensors; geometric shape features of feature extraction: including bridge span arrangement, main beam form, bridge deck width, etc.; material distribution characteristics: such as elastic modulus and Poisson's ratio of different component materials; connection method characteristics: such as pier type, abutment form, etc., and a geometric model of a bridge is simulated based on the extracted bridge geometric features.

[0066] Furthermore, the analysis of the segmented material heterogeneity of the bridge geometric model includes: obtaining the material data of each segment from the bridge geometric model, unifying the format and filling in missing values ​​to obtain a standardized material data set; meshing the standardized material data set, decomposing the geometric boundary into calculation units, generating boundary conditions based on connection constraints, and obtaining meshed units; performing finite element calculations on the meshed units to obtain segmented heterogeneity parameters; segmented heterogeneity parameters include material stiffness, damping characteristics, and mass distribution data.

[0067] Specifically, in this embodiment, material data for each segment is extracted from the bridge geometry model. These data include the elastic modulus, Poisson's ratio, density, etc. of the material. The standardized material data set is meshed to decompose the geometric boundary into computational units. Adaptive meshing methods, such as those based on material quadtrees or octrees, can be used to ensure that the mesh accurately reflects the heterogeneity of the material. Combined with connection constraints, corresponding boundary conditions are generated for each mesh unit. These boundary conditions include fixed supports, applied loads, etc. Finite element analysis software (such as ANSYS or ABAQUS) is used to perform calculations on the meshed units.

[0068] Furthermore, if the heterogeneity exceeds a preset threshold, the segmented response data under earthquake action is calculated; if the segmented material heterogeneity does not exceed the threshold, the preliminary vulnerability probability data is directly output;

[0069] Furthermore, the segmented response data under earthquake action is calculated, including: based on the segmented heterogeneity parameters, using the long short-term memory network to calculate the displacement distribution data, velocity distribution data, and stress distribution data of each segment structure under earthquake action; and obtaining the segmented response data based on the displacement distribution data, velocity distribution data, and stress distribution data.

[0070] Specifically, in this embodiment, a long short-term memory network is constructed and trained; network structure: an LSTM network is constructed, with the input being the acceleration time history of the earthquake motion and the heterogeneous parameters of the structure, and the output being the structural response data such as displacement, velocity, and stress. A stacked structure can be used, such as three unidirectional LSTM layers and one fully connected layer; training data: historical earthquake data and corresponding structural response data are used as the training set. The data needs to be divided into subsequences to reduce the difficulty and time consumption of training; training process: mean square error is used as the loss function, the Adam optimizer is used for training, and dropout operation is used to prevent overfitting;

[0071] Furthermore, obtaining refined prediction results includes: using a machine learning classification algorithm to classify the segmented response data to obtain a segmented vulnerability probability distribution set; using a spatial interpolation algorithm to spatially interpolate the segmented vulnerability probability distribution set to generate a spatially differentiated vulnerability probability distribution of the entire bridge to obtain refined prediction results.

[0072] Specifically, in this embodiment, the machine learning classification algorithm uses the random forest algorithm, which has good robustness to high-dimensional data and noise. Standardized segmented response data is used as input to train the selected classification model. The goal of the model is to classify the response data of each segment structure into different vulnerability levels (such as low, medium, and high). The vulnerability probability distribution of each segment is obtained through the model's prediction results. The spatial interpolation algorithm uses Kriging interpolation, which can consider spatial autocorrelation and is suitable for cases where data distribution is uneven. Specifically, the segmented vulnerability probability distribution set is used as known data points, and interpolation is performed using the selected spatial interpolation algorithm based on the spatial layout of the bridge. For example, when using Kriging interpolation, it is necessary to first model the data using the semivariogram function, and then interpolate according to the minimum variance estimation principle. Generating a continuous distribution: Through spatial interpolation, the discrete segmented vulnerability probability distribution is expanded to the entire space of the bridge, generating a spatially differentiated vulnerability probability distribution for the entire bridge.

[0073] Furthermore, classifying the segmented response data includes: pre-processing the segmented response data to obtain a segmented structural feature set; classifying the segmented structural feature set based on a support vector machine to identify potential weak points and obtain weak point distribution data; calculating the vulnerability probability of each weak point using a probability density estimation algorithm based on the weak point distribution data; integrating the vulnerability probability of each weak point with the stress distribution data to obtain a segmented vulnerability probability distribution set;

[0074] Specifically, in this embodiment, an appropriate kernel function is selected based on the data distribution. For linearly separable data, a linear kernel can be used; for nonlinearly separable data, a Gaussian kernel can be used. The segmented structural feature set is used as input to train an SVM classifier. Classifier parameters, such as the penalty parameter C and the kernel function parameters, are optimized through methods such as cross-validation. The trained SVM classifier is used to classify each segment and identify potential weak points. The vulnerability probability and stress distribution data of each weak point are then integrated. The stress distribution data can be used as a weight to perform a weighted average of the vulnerability probabilities of the weak points. Finally, the integrated vulnerability probabilities are distributed to each segment, forming a segmented vulnerability probability distribution set.

[0075] Furthermore, spatial interpolation is performed on the segmented vulnerability probability distribution set to generate a spatially differentiated vulnerability probability distribution for the entire bridge and obtain a refined prediction result, including: obtaining local probability data from the segmented vulnerability probability distribution set; based on the local probability data, using a data screening algorithm to screen high-probability areas to obtain a local high-risk data set; based on the local high-risk data set, using a spatial interpolation algorithm to calculate the probability transition values ​​between each area to obtain the spatially differentiated vulnerability probability distribution for the entire bridge; and performing spatial smoothing optimization on the spatially differentiated vulnerability probability distribution for the entire bridge to obtain a refined prediction result.

[0076] Specifically, in this embodiment, the vulnerability probability dataset of a bridge segment contains 100 units, with probability values ​​ranging from 0.1 to 0.5. The screening threshold is set to 0.3, and units with probabilities greater than 0.3 are marked as high-risk.

[0077] Furthermore, if the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted, and an updated vulnerability probability distribution is output to obtain a final vulnerability prediction result, including: using a deviation detection algorithm to calculate the difference between the refined prediction result and the preliminary vulnerability probability data to obtain a difference value; if the difference value exceeds a preset threshold, screening high-deviation areas through a data screening method to obtain a high-deviation data set; adjusting the parameters of the nonlinear mapping model through a parameter optimization algorithm based on the high-deviation data set, updating the mapping relationship through iterative calculation, and obtaining an adjusted nonlinear mapping model; re-inputting the seismic parameter set into the adjusted nonlinear mapping model for vulnerability probability distribution to obtain a final vulnerability prediction result;

[0078] Specifically, in this example, 50 high-deviation units were detected in a bridge segment, with deviation values ​​ranging from 0.15 to 0.3. The screening criteria were that the deviation value was greater than 0.2 and that the units were located near bridge joints. Ultimately, 10 high-deviation units were extracted. This high-deviation dataset composed of these units provided precise targets for subsequent model optimization.

[0079] Example 2

[0080] In this embodiment, a machine learning-based bridge seismic vulnerability prediction system is proposed, comprising: a seismic parameter acquisition module for constructing a nonlinear mapping model using a deep neural network, performing vulnerability probability prediction based on the nonlinear mapping model, and obtaining preliminary vulnerability probability data; a bridge model construction module for acquiring bridge structural data to construct a bridge geometric model, analyzing the segmented material heterogeneity of the bridge geometric model, and calculating segmented response data under earthquake action if the heterogeneity exceeds a preset threshold; a refined result prediction module for analyzing the spatially differentiated vulnerability probability distribution of the entire bridge based on the segmented response data to obtain refined prediction results; and an update prediction module for determining whether the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold. If so, the nonlinear mapping model is adjusted, and an updated vulnerability probability distribution is output to obtain a final vulnerability prediction result. The proposed system fully considers the complexity and dynamic nature of seismic vulnerability prediction and can provide strong support for bridge seismic performance assessment.

[0081] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A bridge seismic vulnerability prediction method based on machine learning, characterized in that: include: Obtaining a set of earthquake parameters, building a nonlinear mapping model using a deep neural network, and performing vulnerability probability prediction based on the nonlinear mapping model to obtain preliminary vulnerability probability data; Acquiring bridge structural data to construct a bridge geometric model, analyzing segmented material heterogeneity of the bridge geometric model, and calculating segmented response data under earthquake action if the heterogeneity exceeds a preset threshold; Based on the segmented response data, the spatially differentiated vulnerability probability distribution of the entire bridge is analyzed to obtain a refined prediction result; If the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result.

2. The bridge seismic vulnerability prediction method based on machine learning according to claim 1 is characterized in that: Obtaining earthquake parameter sets includes: Obtain multi-dimensional earthquake motion data from the earthquake motion parameter database and perform standardization on the data; The principal component analysis algorithm is used to extract features from the standardized seismic motion dataset, separate the key feature acceleration values, frequency distribution and duration, and obtain the earthquake parameter set.

3. The bridge seismic vulnerability prediction method based on machine learning according to claim 1 is characterized in that: Obtaining bridge structure data and building a bridge geometric model includes: The bridge structure data is obtained from the bridge structure database and the bridge structure data is cleaned; Perform feature extraction on the cleaned bridge structure data to obtain geometric shape features, material distribution features, and connection mode features, and perform dimensionality reduction through principal component analysis to obtain an optimized feature set. Extracting characteristic boundary conditions from the optimized feature set by a finite element analysis algorithm; The mechanical parameters of the optimized feature set are calculated according to the characteristic boundary conditions, and a bridge geometric model is constructed according to the mechanical parameters.

4. The bridge seismic vulnerability prediction method based on machine learning according to claim 1, characterized in that: Analyzing the segmented material heterogeneity of the bridge geometry includes: Obtaining material data of each section from the bridge geometric model, unifying the format and filling in missing values ​​to obtain a standardized material data set; Meshing the standardized material data set, decomposing the geometric boundary into calculation units, generating boundary conditions in combination with connection constraints, and obtaining meshed units; Performing finite element calculation on the gridded unit to obtain segmented heterogeneity parameters; The segmented heterogeneity parameters include material stiffness, damping characteristics, and mass distribution data.

5. The bridge seismic vulnerability prediction method based on machine learning according to claim 1, characterized in that: Calculate segmented response data under earthquake action, including: Based on the segmented heterogeneity parameters, a long short-term memory network is used to calculate the displacement distribution data, velocity distribution data, and stress distribution data of each segment of the structure under earthquake action; The segmented response data is obtained based on the displacement distribution data, velocity distribution data, and stress distribution data.

6. The bridge seismic vulnerability prediction method based on machine learning according to claim 5 is characterized in that: Obtaining the refined prediction result includes: Using a machine learning classification algorithm to classify the segmented response data, and obtain a segmented vulnerability probability distribution set; A spatial interpolation algorithm is used to perform spatial interpolation on the segmented vulnerability probability distribution set to generate the spatially differentiated vulnerability probability distribution of the entire bridge and obtain a refined prediction result.

7. The bridge seismic vulnerability prediction method based on machine learning according to claim 6, characterized in that: Classifying the segmented response data includes: Preprocessing the segmented response data to obtain a segmented structure feature set; Classify the segmented structure feature set based on support vector machine, identify potential weak points, and obtain weak point distribution data; Based on the weak point distribution data, the probability density estimation algorithm is used to calculate the vulnerability probability of each weak point; The fragility probability of each weak point and the stress distribution data are integrated to obtain a segmented fragility probability distribution set.

8. The bridge seismic vulnerability prediction method based on machine learning according to claim 6, characterized in that: Spatial interpolation is performed on the segmented vulnerability probability distribution set to generate the spatially differentiated vulnerability probability distribution of the entire bridge and obtain refined prediction results, including: Obtaining local probability data from a piecewise fragility probability distribution set; Based on local probability data, a data screening algorithm is used to screen high-probability areas and obtain a local high-risk data set; Based on the local high-risk data set, a spatial interpolation algorithm is used to calculate the probability transition value between each area to obtain the spatially differentiated vulnerability probability distribution of the entire bridge. The spatially differentiated vulnerability probability distribution of the entire bridge is spatially smoothed and optimized to obtain precise prediction results.

9. The bridge seismic vulnerability prediction method based on machine learning according to claim 1, characterized in that: If the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result, including: Calculating the difference between the refined prediction result and the preliminary vulnerability probability data using a deviation detection algorithm to obtain a difference value; If the difference value exceeds the preset threshold, the high deviation area is filtered out through the data screening method to obtain the high deviation data set; According to the high-deviation data set, adjusting the parameters of the nonlinear mapping model by a parameter optimization algorithm, updating the mapping relationship by iterative calculation, and obtaining an adjusted nonlinear mapping model; The earthquake parameter set is re-input into the adjusted nonlinear mapping model to perform vulnerability probability distribution, and a final vulnerability prediction result is obtained.

10. A bridge seismic vulnerability prediction system based on machine learning, characterized in that: include: A seismic parameter acquisition module is used to construct a nonlinear mapping model using a deep neural network, perform vulnerability probability prediction based on the nonlinear mapping model, and obtain preliminary vulnerability probability data; a bridge model construction module, configured to acquire bridge structural data to construct a bridge geometric model, analyze the segmented material heterogeneity of the bridge geometric model, and calculate segmented response data under earthquake action if the heterogeneity exceeds a preset threshold; A refined result prediction module is used to analyze the spatial difference vulnerability probability distribution of the entire bridge based on the segmented response data to obtain a refined prediction result; The update prediction module is used to determine whether the deviation between the refined prediction result and the preliminary vulnerability probability exceeds a preset threshold. If the deviation exceeds the threshold, the nonlinear mapping model is adjusted to output an updated vulnerability probability distribution to obtain a final vulnerability prediction result.

Images