Regional seismic liquefaction space-time field construction method fusing physical simulation and machine learning

Abstract

The invention discloses a regional seismic liquefaction space-time field construction method fusing physical simulation and machine learning, and the method comprises the following steps: selecting an evaluation region, and dividing the region through a kilometer grid; constructing a seismic oscillation space-time field of the evaluation area to obtain seismic oscillation data of each grid point; collecting seismic liquefaction site data to construct a training set; performing seismic liquefaction grade evaluation training by using the training set to generate a seismic liquefaction grade model; performing liquefaction grade evaluation on each grid point of the evaluation area to obtain a liquefaction grade evaluation result; and drawing a stratigraphic liquefaction grade spatial distribution diagram of the evaluation area based on the site soil liquefaction grade evaluation result. According to the regional seismic liquefaction time-space field construction method integrating physical simulation and machine learning, seismic liquefaction grade evaluation of a site soil body under an earthquake can be set, and the spatial change condition of the site seismic liquefaction grade is reflected while the soil layer liquefaction danger degree is evaluated.

Description

Technical Field

[0001] The present invention relates to the field of geotechnical engineering seismic technology, and in particular to a method for constructing a regional earthquake liquefaction space-time field by integrating physical simulation and machine learning. Background Art

[0002] In the past two decades, catastrophic earthquakes around the world have been frequently accompanied by large-scale liquefaction and the uneven subsidence of building foundations and serious damage to underground facilities. In view of the complex engineering geological conditions and potential earthquake dangers in my country, it is an important topic to ensure the earthquake safety of engineering by studying the spatial distribution of the earthquake liquefaction level of the site and scientifically and quantitatively evaluating the liquefaction risk of the soil layer.

[0003] There are many factors that cause earthquake liquefaction, and there is a strong nonlinear relationship between the seismic motion and soil parameters that affect liquefaction and the conclusion of the site earthquake liquefaction level evaluation. The machine learning method is a practical tool for learning data based on a probabilistic model. It has the ability to extract complex patterns and effective features from data streams, and has the potential to summarize and deduce new characteristics and new mechanisms, which can provide new ideas for earthquake liquefaction level evaluation. Commonly used machine learning methods include linear regression and logistic regression, naive Bayes algorithm, support vector machine, random forest and artificial neural network. Among them, artificial neural network (ANN) is a type of structure that simulates biological nervous system. It has a strong nonlinear mapping function. It mainly relies on learning past experience from sample data. It does not need to design any mathematical model. It is suitable for dealing with nonlinear problems such as site earthquake liquefaction evaluation.

[0004] Earthquake motion is an important factor affecting earthquake liquefaction. The space-time field of earthquake motion in the evaluation area can be calculated through physical simulation methods, providing earthquake motion parameters for evaluating the liquefaction level of the soil in the site. At present, the physical simulation methods of earthquake motion can be roughly divided into three categories according to the theories adopted, namely deterministic methods, random methods and hybrid methods. Among them, the deterministic method generally takes the medium constitutive relationship and Newton's second law as the basis, calculates the wave equation in the process of seismic wave propagation, and realizes numerical simulation through numerical solutions such as finite difference method, finite element method, spectral element method, etc. The spectral element method (SEM) is a numerical method that can simulate three-dimensional wave fields in large-scale complex geological structures. Based on the principle of this method, Komatitsch et al. developed the SPECFEM3D software package, which can simulate the seismic wave field excited by the source in acoustic, elastic, and viscoelastic media.

[0005] It is difficult to reflect the spatial distribution information of earthquake liquefaction at the site by only evaluating the liquefaction level of each sampling point, and it is difficult to make a quick and effective evaluation of the earthquake liquefaction situation of the entire site. Geographic Information Systems (GIS) is a science that collects, stores, calculates, manages, models, and displays geographic data (spatial location, attribute characteristics, and time characteristics) in space with the support of computer software and hardware. It not only has strong spatial analysis capabilities, but also can visualize spatial data, which can be used to display the spatial distribution of earthquake liquefaction levels at the site.

[0006] In summary, the liquefaction phenomenon of the site during an earthquake is characterized by high randomness and uncertainty, and is affected by multiple factors. There are highly nonlinear characteristics between each factor and the earthquake liquefaction grade evaluation conclusion, and only evaluating the liquefaction grade of each sampling point cannot reflect the earthquake liquefaction situation of the entire site. In view of the above situation, how to accurately evaluate the liquefaction grade of the site during an earthquake and to intuitively and clearly display the spatial distribution of the site's earthquake liquefaction grade is a problem that needs to be solved urgently. Summary of the invention

[0007] The purpose of the present invention is to provide a method for constructing a regional earthquake liquefaction space-time field by integrating physical simulation and machine learning, which can evaluate the earthquake liquefaction level of the site soil under a set earthquake, and reflect the spatial variation of the site earthquake liquefaction level while assessing the degree of soil liquefaction hazard.

[0008] To achieve the above objectives, the present invention provides a method for constructing a regional earthquake liquefaction space-time field by integrating physical simulation and machine learning, and the steps are as follows: S1. Select the evaluation area and divide it into kilometer grids; S2. Use physical simulation methods to construct the spatiotemporal field of earthquake motion in the evaluation area and obtain earthquake motion data for each grid point; S3, collecting earthquake liquefaction site data and constructing a training set, wherein the earthquake liquefaction site data includes liquefaction grade evaluation factors; S4. Conduct earthquake liquefaction grade training through machine learning methods, establish a nonlinear relationship between the liquefaction grade evaluation factors in the training set and the earthquake liquefaction grade evaluation conclusions of the site, and generate an earthquake liquefaction grade evaluation model; S5. Evaluate the liquefaction grade of each grid point in the evaluation area according to the earthquake liquefaction grade model to obtain a liquefaction grade evaluation result; S6. Use GIS technology to spatially composite the evaluation results of the site soil liquefaction level and provide a spatial distribution map of the stratum liquefaction level in the evaluation area.

[0009] Preferably, S1 specifically includes: S1.1. Select an evaluation area and determine the scope of the evaluation area; S1.2, using kilometer grids to divide the evaluation area into a number of 1km×1km grids according to a unified standard; S1.3. Take the soil layer parameters at the center of each grid as the geological data within the grid range to provide geological parameters for evaluating the liquefaction level of the site soil.

[0010] Preferably, S2 specifically includes: S2.1. Select the focal parameters of the set earthquake in the evaluation area. The focal parameters include global focal parameters and local focal parameters. The global focal parameters include the location and size of the fault, magnitude, focal mechanism, focal depth, average slip of the fault surface, etc. The local focal parameters include the location and size of the asperity, slip of the asperity, etc. S2.2. Construct an overall physical model and use physical simulation methods to calculate the temporal and spatial fields of seismic motion in the evaluation area; S2.3. Based on the space-time field of seismic motion, the seismic motion data of each grid point in the evaluation area are obtained to provide seismic motion parameters for evaluating the liquefaction grade of the soil at the site.

[0011] Preferably, the physical simulation method used in S2.2 includes: a deterministic spectral element method, a finite difference method, a random method or a hybrid method.

[0012] Preferably, the machine learning method in S4 includes linear regression and logistic regression, naive Bayes algorithm, support vector machine algorithm, random forest algorithm, artificial neural network algorithm, etc.

[0013] Preferably, in S4, a BP neural network based on the Levenber-Marquardt algorithm is used to perform earthquake liquefaction grade evaluation training to generate an earthquake liquefaction grade evaluation model, wherein the earthquake liquefaction grade evaluation model includes three layers of neural networks, namely, one input layer, one hidden layer and one output layer, the number of neurons in the input layer is 6, and the number of neurons in the hidden layer is 10. The specific steps are as follows; S4.1, information forward propagation, the six liquefaction level evaluation factors in the training set are used as neurons in the input layer, and are transmitted to the output layer after being processed by the hidden layer and nonlinearly transformed; S4.2, the error is transmitted backward layer by layer. After the information is transmitted to the output layer, the prediction error is calculated using the loss function, and then the error term is transmitted from the output layer to the hidden layer and finally to the input layer; S4.3. Iterative update of weights and thresholds. The error is calculated once in each iteration. The weights and threshold parameters are updated by calculating the gradient of the loss function with respect to the output result. After obtaining the updated weights and threshold parameters, the previous information forward propagation process is repeated until the error is less than a certain value. The iteration is terminated, the final weight and threshold results are determined, the model training is completed, and an earthquake liquefaction grade evaluation model is generated.

[0014] Preferably, S5 specifically includes: using the seismic peak acceleration, seismic duration and relative density, groundwater level, soil layer depth and overlying non-liquefied soil layer thickness of each grid point in the evaluation area obtained by simulating the space-time field of seismic motion under a set earthquake as input parameters of the BP neural network, and evaluating the liquefaction grade of each grid point according to the earthquake liquefaction grade model to obtain the liquefaction grade evaluation result.

[0015] Preferably, S6 specifically includes: S6.1. The quantitative value of the three-dimensional site soil liquefaction grade evaluation result is used as the input data source of the GIS software; S6.2. Use the spatial interpolation inversion algorithm of GIS to perform interpolation calculation on the quantitative value of the liquefaction grade evaluation result; S6.3. Use the generated three-dimensional data to perform three-dimensional volume imaging of the liquefaction level and draw a spatial distribution map of the liquefaction level of the assessment area.

[0016] Therefore, the present invention adopts the regional earthquake liquefaction space-time field construction method that integrates physical simulation and machine learning in the above steps, and realizes the earthquake liquefaction level evaluation of the site soil under a set earthquake based on physical simulation and machine learning. By combining spatial three-dimensional interpolation and volume visualization technology, it realizes the establishment of a corresponding three-dimensional stratigraphic model based on discrete site liquefaction level data, and intuitively displays the liquefaction level evaluation results in a graphical manner. Compared with traditional earthquake liquefaction potential assessment, it has the significant advantage of being able to intuitively reflect the spatial changes of the site earthquake liquefaction level and the liquefaction risk level of the site soil.

[0017] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] 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 will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 Flowchart of the method for constructing the spatiotemporal field of regional earthquake liquefaction by integrating physical simulation and machine learning; Figure 2 A schematic diagram of gridding the evaluation area using kilometer grids; Figure 3 This is the structural diagram of the earthquake liquefaction grade evaluation model; Figure 4 A flow chart for modeling three-dimensional visualization of regional earthquake liquefaction level evaluation; Figure 5 A schematic diagram of the spatial distribution of formation liquefaction levels in the evaluation area; Figure 6 A two-dimensional cross-sectional diagram of the liquefaction level of the assessment area. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example The embodiment of the present invention provides a method for constructing a regional earthquake liquefaction space-time field by integrating physical simulation and machine learning. Figure 1 As shown, the method comprises the following steps: S1. Select the evaluation area and divide it into kilometer grids; S2. Use physical simulation methods to construct the temporal and spatial field of earthquake motion in the evaluation area, and obtain earthquake motion data such as the peak value and duration of earthquake motion at each grid point; S3, collecting earthquake liquefaction site data and constructing a training set, wherein the earthquake liquefaction site data includes liquefaction grade evaluation factors; S4. Conduct earthquake liquefaction grade training through BP neural network based on Levenber-Marquardt algorithm, establish nonlinear relationship between factors such as ground motion and soil parameters affecting liquefaction in training set and conclusion of earthquake liquefaction grade evaluation of site, and generate earthquake liquefaction grade evaluation model; S5. Evaluate the liquefaction grade of each grid point in the evaluation area according to the earthquake liquefaction grade model to obtain a liquefaction grade evaluation result; S6. Use GIS technology to spatially composite the evaluation results, i.e., liquefaction levels, and provide a spatial distribution map of the liquefaction levels in the evaluation area.

[0023] The following is combined with Figure 2-6 The embodiments of the present invention are further described.

[0024] S1. Select the evaluation area and divide it into kilometer grids, such as Figure 2 As shown, the specific method is: Select the evaluation area and determine the scope of the area; then use the kilometer grid to divide the area into several 1km×1km grids according to a unified standard; take the soil layer parameters at the center of each grid as the geological data within the grid range to provide geological parameters for evaluating the liquefaction level of the soil at the site.

[0025] S2. The temporal and spatial field of earthquake motion in the evaluation area is constructed by physical simulation methods to obtain earthquake motion data such as the peak value and duration of earthquake motion at each grid point. The specific method is as follows: S2.1. Select the earthquake source parameters of the evaluation area; construct an overall physical model including finite fault earthquake source, crustal velocity structure and complex site, where the finite fault earthquake source adopts the GP14.3 hybrid earthquake source model, in which the global earthquake source parameters such as the position and size of the fault, magnitude, earthquake source mechanism, earthquake source depth, average slip of the fault plane, and the local earthquake source parameters such as the position and size of the asperity and slip of the asperity are set, and generate the earthquake source spatiotemporal distribution file based on the SPECFEM3D format; S2.2, import the above-mentioned earthquake source spatiotemporal distribution file into SPECFEM3D, and calculate the spatiotemporal field of earthquake motion in the evaluation area; S2.3. Based on the above-mentioned seismic space-time field, the seismic peak value, duration and other seismic data of each grid point in the evaluation area are obtained and stored in a table form to provide seismic parameters for the liquefaction grade of the soil in the evaluation site.

[0026] S3. Collect earthquake liquefaction site data and construct a training set. The earthquake liquefaction site data includes liquefaction grade evaluation factors. The specific method is as follows: Collect earthquake liquefaction site data. The original data used in this embodiment includes the measured data of many earthquakes in the world, especially several major earthquakes that occurred in recent years, and introduces the measured data of liquefaction properties of several major earthquake sites in China, including Tonghai, Haicheng and Tangshan. A total of 297 site liquefaction measured data, including 184 liquefied sites and 113 non-liquefied sites, and the collected earthquake liquefaction site data are used as training sets. Based on the above data, 6 liquefaction level evaluation factors are considered, namely, peak acceleration of seismic motion, duration of seismic motion, relative density, groundwater level, soil layer burial depth and thickness of overlying non-liquefied soil layer.

[0027] S4. Conduct earthquake liquefaction grade training to generate an earthquake liquefaction grade evaluation model: BP (Back Propagation) neural network is a multi-layer feedforward network trained by error back propagation, and is a type of artificial neural network. The BP neural network is used to train the earthquake liquefaction level evaluation using the Levenber-Marquardt algorithm to learn the liquefaction level evaluation factors in the training set. The generated earthquake liquefaction level evaluation model structure is as follows: Figure 3 As shown, the specific method is: S4.1, information forward propagation, that is, the six liquefaction level evaluation factors in the training set are used as neurons in the input layer, and the data of the input layer is transmitted to the output layer after being processed and nonlinearly transformed by the hidden layer. The number of layers of the input layer, output layer and hidden layer is 1, and the number of neurons in the hidden layer is 10. The output layer outputs the test data, which is divided into 4 categories: no liquefaction, mild liquefaction, moderate liquefaction and severe liquefaction. The activation functions of the neurons in the hidden layer and the output layer are both Sigmoid functions (S-type functions). In the process of forward propagation from the input layer to the hidden layer, the sample parameters of the input liquefaction level evaluation factors are transmitted to the output layer through the input layer, layer-by-layer processing and nonlinear transformation of the hidden layer; S4.2, the error is transmitted in reverse layer by layer, that is, after the information is transmitted to the output layer, the prediction error is calculated using the loss function, and then the error term is transmitted from the output layer to the hidden layer, and finally to the input layer; S4.3, during which the weights and thresholds are iteratively updated, that is, the error is apportioned to each layer during the propagation process, and then the weights of the neurons in each layer are updated. The error is reduced along the gradient direction by adjusting the connection strength and threshold of the neuron nodes between the hidden layer and the input layer, and between the hidden layer and the output layer. After obtaining the updated weight and threshold parameters, the previous information forward propagation process is repeated. In a large number of repeated learning and training processes, the evaluation accuracy of the model is improved accordingly. When the error is less than a certain value, the iteration is terminated, the final weight and threshold results are determined, the model training is completed, and the earthquake liquefaction grade evaluation model is generated. Through the transformation of the initial input data represented by the neurons in the hidden layer of the neural network, the last layer of neurons can complete the evaluation of the liquefaction grade. The earthquake liquefaction grade evaluation method proposed in this application can comprehensively reflect the nonlinear relationship between various influencing factors and the site liquefaction grade, and has a high data recognition ability and liquefaction grade evaluation accuracy. The model can be used as a new multi-parameter comprehensive evaluation model to correctly predict the liquefaction grade of the earthquake site.

[0028] S5. According to the earthquake liquefaction grade model, the liquefaction grade of each grid point in the evaluation area is evaluated to obtain the liquefaction grade evaluation result. The specific method is as follows: The peak acceleration, duration and relative density of seismic motion, groundwater level, soil depth and thickness of overlying non-liquefied soil layer of each grid point in the evaluation area obtained by simulating the space-time field of seismic motion under the set earthquake are used as the input parameters of the BP neural network. The liquefaction grade of each grid point is evaluated according to the earthquake liquefaction grade model to obtain the liquefaction grade evaluation results.

[0029] S6. Use GIS technology to spatially composite the evaluation results, i.e., liquefaction levels, and provide a spatial distribution map of the liquefaction level of the evaluation area, such as Figure 4 As shown, the specific method is: S6.1. Based on the concept of Digital Elevation Model (DEM), the quantified value of the evaluation result of the site seismic liquefaction level is equivalent to the elevation data in the DEM model, and the three-dimensional information of the site liquefaction level is converted from GIS into a data format supported by the visualization software - 3D Surfer, so as to realize the conversion of the interpolation data format; 3D Surfer is a three-dimensional map drawing software developed by Golden Software Company of the United States; S6.2, call 3D Surfer, based on the interpolation principle of the Kriging method, and perform internal difference calculation on the quantitative value of the evaluation result of the liquefaction grade of the three-dimensional site soil through the variation function, so as to estimate the liquefaction grade of the soil layer at the unpredicted point through the information (liquefaction grade) of the known predicted point; S6.3. Based on the three-dimensional volume visualization technology, the three-dimensional data generated by the above interpolation is used to assign different color values ​​and opacity values ​​according to the different attributes of the volume elements to generate a three-dimensional model of the site liquefaction level, such as Figure 5 As shown; Based on the three-dimensional volume imaging of the liquefaction level, the volume cutting and slicing functions can be used to realize the three-dimensional cutting volume and two-dimensional cross-section of the site liquefaction level (such as Figure 6 By combining spatial three-dimensional interpolation and volume visualization technology, it is possible to establish a corresponding three-dimensional stratigraphic model based on discrete site liquefaction grade data, and to intuitively display the site liquefaction grade evaluation results in a graphical manner, which better reflects the spatial changes in the site's earthquake liquefaction grade and the liquefaction risk of the site's soil layer.

[0030] The remaining technical features in the above embodiments can be flexibly selected by those skilled in the art according to actual conditions to meet different specific practical needs. However, it is obvious to those skilled in the art that it is not necessary to adopt these specific details to implement the present invention. In other examples, in order to avoid confusing the present invention, the well-known components, structures or parts are not specifically described, which are all within the technical protection scope defined by the technical solution claimed for protection in the claims of the present invention.

[0031] Modifications and changes made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the scope of protection of the claims attached to the present invention. In the above description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details are not necessary to practice the present invention. In other examples, in order to avoid confusing the present invention, well-known technologies, such as specific construction details, operating conditions and other technical conditions, are not specifically described.

[0032] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. 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 methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for constructing a regional earthquake liquefaction space-time field by integrating physical simulation and machine learning, characterized in that: Here are the steps: S1. Select the evaluation area and divide it into kilometer grids; S2. Use physical simulation methods to construct the spatiotemporal field of earthquake motion in the evaluation area and obtain earthquake motion data for each grid point; S3, collecting earthquake liquefaction site data and constructing a training set, wherein the earthquake liquefaction site data includes liquefaction grade evaluation factors; S4. Conduct earthquake liquefaction grade training through machine learning methods, establish a nonlinear relationship between the liquefaction grade evaluation factors in the training set and the earthquake liquefaction grade evaluation conclusions of the site, and generate an earthquake liquefaction grade evaluation model; S5. Evaluate the liquefaction grade of each grid point in the evaluation area according to the earthquake liquefaction grade model to obtain a liquefaction grade evaluation result; S6. Use GIS technology to spatially composite the evaluation results of the site soil liquefaction level and provide a spatial distribution map of the stratum liquefaction level in the evaluation area.

2. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 1, characterized in that: S1 specifically includes: S1.

1. Select an evaluation area and determine the scope of the evaluation area; S1.2, using kilometer grids to divide the evaluation area into a number of 1km×1km grids according to a unified standard; S1.

3. Take the soil layer parameters at the center of each grid as the geological data within the grid range to provide geological parameters for evaluating the liquefaction level of the site soil.

3. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 2, characterized in that: S2 specifically includes: S2.

1. Select the earthquake source parameters of the set earthquake in the evaluation area. The earthquake source parameters include global earthquake source parameters and local earthquake source parameters. The global earthquake source parameters include the location and size of the fault, magnitude, rupture direction, depth of the earthquake source, and average slip of the fault surface. The local earthquake source parameters include the location and size of the asperity and the slip of the asperity. S2.2, constructing an overall physical model including finite fault sources, crustal velocity structure and complex sites, and calculating the spatiotemporal field of seismic motion in the evaluation area using a physical simulation method based on the overall physical model; S2.

3. Based on the space-time field of seismic motion, the seismic motion data of each grid point in the evaluation area are obtained to provide seismic motion parameters for evaluating the liquefaction grade of the soil at the site.

4. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 3 is characterized in that: The physical simulation methods used in S2.2 include: deterministic spectral element method, finite difference method, stochastic method or hybrid method.

5. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 1, characterized in that: The machine learning methods in S4 include linear regression and logistic regression, naive Bayes algorithm, support vector machine algorithm, random forest algorithm, and artificial neural network algorithm.

6. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 3, characterized in that: In S4, a BP neural network based on the Levenber-Marquardt algorithm is used to perform earthquake liquefaction grade evaluation training to generate an earthquake liquefaction grade evaluation model, wherein the earthquake liquefaction grade evaluation model includes three layers of neural networks, namely, one input layer, one hidden layer and one output layer, the number of neurons in the input layer is 6, and the number of neurons in the hidden layer is 10. The specific steps are as follows; S4.1, information forward propagation, the six liquefaction level evaluation factors in the training set are used as neurons in the input layer, and are transmitted to the output layer after being processed by the hidden layer and nonlinearly transformed; S4.2, the error is transmitted backward layer by layer. After the information is transmitted to the output layer, the prediction error is calculated using the loss function, and then the error term is transmitted from the output layer to the hidden layer and finally to the input layer; S4.

3. Iterative update of weights and thresholds. The error is calculated once in each iteration. The weights and threshold parameters are updated by calculating the gradient of the loss function with respect to the output result. After obtaining the updated weights and threshold parameters, the previous information forward propagation process is repeated until the error is less than a certain value. The iteration is terminated, the final weight and threshold results are determined, the model training is completed, and an earthquake liquefaction grade evaluation model is generated.

7. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 6, characterized in that: S5 specifically includes: using the seismic peak acceleration, seismic duration and relative density, groundwater level, soil burial depth and overlying non-liquefied soil layer thickness of each grid point in the evaluation area obtained by simulating the space-time field of seismic motion under a set earthquake as input parameters of the BP neural network, evaluating the liquefaction grade of each grid point according to the earthquake liquefaction grade model, and obtaining the liquefaction grade evaluation result.

8. The method for constructing a regional earthquake liquefaction space-time field integrating physical simulation and machine learning according to claim 7, characterized in that: S6 specifically includes: S6.

1. The quantitative value of the three-dimensional site soil liquefaction grade evaluation result is used as the input data source of the GIS software; S6.

2. Use the spatial interpolation inversion algorithm of GIS to perform interpolation calculation on the quantitative value of the liquefaction grade evaluation result; S6.

3. Use the generated three-dimensional data to perform three-dimensional volume imaging of the liquefaction level and draw a spatial distribution map of the liquefaction level of the assessment area.

Images