Slope stability dynamic prediction and geological parameter inversion method

Abstract

The invention provides a slope stability dynamic prediction and geological parameter inversion method, and the method comprises the steps: building a slope stability simulation model through a COMSOL multi-physics field simulation platform, and generating simulation data in combination with an orthogonal experiment design and automatic parameter optimization technology; then, training and optimizing the simulation data based on a variational auto-encoder and a generative adversarial network to obtain enhanced simulation data; and performing geological parameter inversion based on SBAS-InSAR ground deformation data and the enhanced simulation data, generating a safety coefficient distribution map in the region by using geological parameters obtained by inversion, and fusing the safety coefficient distribution map with a geological disaster potential map to evaluate disaster risks of different regions. According to the method, the timeliness and the real-time performance of a prediction result are ensured, and a scientific basis is provided for engineering design, disaster monitoring and decision making.

Description

Technical Field

[0001] The invention belongs to the fields of geological engineering, numerical simulation and machine learning, and specifically relates to a method for dynamic prediction of slope stability and geological parameter inversion. Background Art

[0002] With the development of infrastructure construction in mountainous areas, slope instability caused by rainfall threatens the safety of transportation such as railways and highways. Traditional slope stability analysis methods such as limit equilibrium method and finite element method are based on static assumptions and rely on field data. They are difficult to reflect the impact of dynamic factors such as rainfall and the results are delayed. Although SBAS-InSAR technology can monitor surface deformation, it cannot analyze underground and complex geological hazards, lacks response to dynamic factors, and needs to be combined with other data models. Existing geological parameter inversion methods rely on field measurement experiments, which are time-consuming, narrow in scope, low in accuracy, and difficult to handle dynamic environmental factors. GIS platforms are often used for disaster assessment, but traditional platforms rely on static geological data and cannot update slope stability and disaster risk changes in real time, and their application value is limited in dynamic environments. Summary of the invention

[0003] The purpose of the invention is to provide a method for dynamic prediction of slope stability and inversion of geological parameters.

[0004] The technical solution of the present invention is as follows: A method for dynamic prediction of slope stability and inversion of geological parameters, the method comprising: A slope stability simulation model was established through the COMSOL multi-physics simulation platform, and simulation data was generated by combining orthogonal experimental design and automated parameter optimization technology; Then, the simulated data is trained and optimized based on a variational autoencoder and a generative adversarial network to obtain enhanced simulated data; Based on the SBAS-InSAR ground deformation data and the enhanced simulation data, geological parameter inversion is performed, and the inverted geological parameters are used to generate a safety factor distribution map in the region, which is then integrated with the geological disaster potential map to evaluate the disaster risks in different regions.

[0005] Furthermore, the slope stability simulation model is established through the COMSOL multi-physics field simulation platform, and the simulation data is generated by combining orthogonal experimental design and automated parameter optimization technology as follows: Use COMSOL multi-physics simulation platform to establish a slope stability simulation model; The slope stability simulation model is optimized by using automated parameter optimization technology and adaptive grid encryption technology; The orthogonal experimental design method was used to systematically change the geological parameters and rainfall, and the slope safety factor under different conditions was obtained through the optimized slope stability simulation model. By calculating the slope safety factor under different conditions, the simulation data of slope stability is obtained.

[0006] Furthermore, the geological parameters include soil permeability and geotechnical properties.

[0007] Furthermore, the simulation data is trained and optimized based on the variational autoencoder and the generative adversarial network to obtain the enhanced simulation data: The potential spatial distribution of slope stability data is learned through variational autoencoders to generate simulated data under different geological conditions and rainfall conditions to supplement the deficiencies in the original dataset. Then, the quality of the generated simulation data is optimized through the adversarial training mechanism through the generative adversarial network.

[0008] Furthermore, the geological parameter inversion based on the SBAS-InSAR ground deformation data and the enhanced simulation data is specifically as follows: The SBAS-InSAR ground deformation data are converted into the safety factor of the slope according to Newton's second law; Integrate the enhanced simulated data set and the safety factor obtained from SBAS-InSAR conversion; The geological parameter inversion model was constructed by integrating deep learning algorithms using Bayesian hyperparameter optimization convolutional neural networks and bidirectional long short-term memory neural networks; The integrated data are input into the geological parameter inversion model trained by Bayesian hyperparameter optimization, and the geological parameters are obtained by output inversion.

[0009] Furthermore, the safety factor distribution map in the region generated by inverting the geological parameters is specifically as follows: The geological parameters obtained by inversion are combined with real-time rainfall data and brought into the COMSOL multi-physics simulation platform to calculate the safety factor in the area; The safety factor data points obtained by simulation are converted into raster data to generate a safety factor distribution map in the area.

[0010] Furthermore, the simulated safety factor data points are converted into raster data using the Kriging interpolation method. Compared with the prior art, the present invention has the following advantages: The present invention uses dynamic analysis based on a multi-physics field simulation platform, which not only takes into account dynamic factors such as rainfall and soil infiltration, but also combines adaptive grid division and multi-scale simulation technology to further improve the accuracy and reliability of slope stability prediction; the introduction of the VAE-GAN model improves the quality of training data by enhancing the diversity and authenticity of the simulation data set, thereby ensuring the training effect of the deep learning model; the combination of SBAS-InSAR technology and geological parameter inversion method not only realizes accurate prediction under dynamic monitoring, but also optimizes the inversion results of geological parameters according to real-time environmental changes; the dynamic safety factor distribution is updated in combination with the GIS platform to ensure the timeliness and real-time nature of the prediction results, providing a scientific basis for engineering design, disaster monitoring and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0012] Figure 1 A schematic diagram of the method flow of the present invention is shown; Figure 2 A schematic diagram of the GAN-VAE process of the present invention is shown; Figure 3 A schematic diagram of the geological parameter inversion model prediction process of the present invention is shown. DETAILED DESCRIPTION

[0013] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present 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.

[0014] like Figure 1-Figure 3 As shown, an embodiment of the present invention provides a method for dynamic prediction of slope stability and inversion of geological parameters, which specifically includes: 1. Slope model establishment and high-precision simulation based on COMSOL

[0015] (1) Establishing slope stability simulation model:

[0016] Firstly, a slope stability model including rainfall infiltration, soil hydraulic characteristics, and mechanical properties was established using the COMSOL multi-physics simulation platform. This model can simulate the influence of soil permeability, rock and soil mechanical properties, and rainfall on slope stability under different geological conditions.

[0017] Especially for the geological conditions along the railway, the orthogonal experimental design method is used to systematically change the geological parameters (such as soil permeability, soil mechanical properties, etc.) and environmental factors such as rainfall, and calculate the slope safety factor under different conditions.

[0018] Through these experimental data, slope stability values ​​under different conditions can be generated, further providing a rich simulation data set for subsequent machine learning models.

[0019] (2) Automatic parameter optimization and adaptive mesh refinement:

[0020] The automated parameter optimization technology is used to optimize the model under different geological and rainfall conditions to obtain the optimal slope stability prediction model.

[0021] Adaptive mesh encryption technology is introduced to automatically adjust the density and accuracy of the mesh according to the geological characteristics and changes in different areas during the slope stability simulation process. In this way, the stress, permeability and deformation characteristics of the slope can be captured more finely, thereby improving the accuracy of the slope safety factor calculation. 2. VAE-GAN generated dataset enhancement

[0022] (1) Data generation mechanism combining VAE and GAN

[0023] In order to enhance the diversity and quality of the simulated data set, this paper adopts a generative model that combines a variational autoencoder (VAE) with a generative adversarial network (GAN). VAE is used to learn the potential spatial distribution of slope stability data and generate simulated data under different geological conditions and rainfall conditions to supplement the deficiencies in the original data set.

[0024] Specific steps for VAE-GAN to generate dataset enhancement Data preprocessing: ① Organize slope stability simulation data to ensure that the data covers different geological conditions, rainfall environments and slope states. ② Perform data normalization to ensure the uniformity of input data and improve the stability and generalization ability of model training.

[0025] Variational Autoencoder (VAE) generates data: ① Learn the potential spatial distribution of slope stability data through VAE and extract key features. ② Generate simulated data under different geological conditions and rainfall conditions to supplement the deficiencies in the original data set. ③ Use reconstruction loss function to optimize the VAE model and improve the quality of generated data.

[0026] Generative Adversarial Network (GAN) optimizes data: ① GAN is used for data enhancement, and the data generated by VAE is optimized through adversarial training mechanism. ② The generator is responsible for generating realistic slope stability data, and the discriminator is used to distinguish between real data and generated data. ③ Through continuous iterative training, the authenticity of the generated data is improved to make it more consistent with the actual distribution of slope stability.

[0027] Data fusion and enhancement: ① Integrate the data generated by VAE and the data optimized by GAN to form a high-quality enhanced data set. Combined with real SBAS-InSAR deformation data, the diversity and reliability of the data are further improved. Data quality assessment is performed to ensure that the enhanced data conforms to the actual law of slope stability changes and can be used for subsequent geological parameter inversion and stability prediction.

[0028] Application to geological parameter inversion and stability prediction: Use the enhanced data set to train the Bayesian hyperparameter optimized convolutional neural network (CNN) and bidirectional long short-term memory network (BiLSTM) models to improve the accuracy of geological parameter inversion. Combined with deep learning methods, dynamic prediction of slope stability is achieved, safety factor calculation is optimized, and the real-time and reliability of prediction are improved.

[0029] This method generates latent distribution data through variational autoencoders (VAEs) and optimizes data quality through adversarial training mechanisms of generative adversarial networks (GANs), significantly improving the diversity and authenticity of data sets. GANs generate more realistic data that meets actual slope stability conditions through adversarial learning between generators and discriminators, thereby enhancing the reliability and diversity of data. This provides high-quality input data support for subsequent slope stability analysis and model training. 3. SBAS-InSAR deformation data conversion and geological parameter inversion

[0030] (1) SBAS-InSAR deformation data conversion:

[0031] In the process of slope stability prediction, the present invention uses the ground deformation data acquired by SBAS-InSAR in the historical remote sensing image as input data.

[0032] Based on Newton's second law, a slope motion model is established to convert the deformation rate into safety factor data. These safety factor data can be used as observation data and combined with other data sources to improve the accuracy of slope stability prediction.

[0033] (2) Inversion of geological parameters:

[0034] Combining the dataset generated by VAE-GAN optimization and SBAS-InSAR deformation data, the Bayesian hyperparameter optimization convolutional neural network (CNN) and bidirectional long short-term memory neural network (BiLSTM) integrated deep learning algorithm are used. The model structure process is as follows Figure 3 As shown, the inversion of geological parameters is performed.

[0035] ① Generate enhanced simulated data through VAE-GAN for training CNN-BiLSTM deep learning model; ②After training is completed, input SBAS-InSAR deformation data points (converted into safety factors) and use the trained model to invert geological parameters (such as soil permeability, rock and soil mechanical properties, etc.) Through this method, key information is extracted from the geological parameters obtained by inversion (such as soil permeability, mechanical properties, etc.), and these parameters are brought into the slope stability prediction model to further optimize the model prediction effect.

[0036] This process improves the precision and accuracy of geological parameter inversion, thereby enhancing the reliability of the slope stability prediction model. 4. Spatial and temporal distribution of regional safety factor

[0037] (1) Regional safety factor calculation and spatiotemporal update:

[0038] The present invention utilizes a geographic information system (GIS) platform such as ArcGIS to combine real-time rainfall dynamic parameters with updated geological parameter data points, and brings them into a COMSOL physical model to calculate the safety factor.

[0039] The calculated safety factor data points are converted into raster data through the Kriging interpolation method to generate a safety factor distribution map in the region. This distribution map can intuitively reflect the slope stability status of different regions and can be updated in real time with the change of time and rainfall conditions. 5. Safety factor distribution and disaster risk assessment:

[0040] The generated safety factor distribution map not only shows the slope stability of each area, but can also be integrated with the geological disaster potential map to assess the disaster risks in different areas. In this way, it provides reliable data support for subsequent engineering design, disaster warning, monitoring and decision-making.

[0041] Based on these spatial and temporal distribution of safety factor data, we can further analyze the dynamic changes in slope stability, and make timely disaster warnings in combination with real-time weather data, providing more accurate support for engineering design and disaster prevention.

[0042] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for dynamic prediction of slope stability and inversion of geological parameters, characterized in that: The method comprises: A slope stability simulation model was established through the COMSOL multi-physics simulation platform, and simulation data was generated by combining orthogonal experimental design and automated parameter optimization technology; Then, the simulated data is trained and optimized based on a variational autoencoder and a generative adversarial network to obtain enhanced simulated data; Based on the SBAS-InSAR ground deformation data and the enhanced simulation data, geological parameter inversion is performed, and the inverted geological parameters are used to generate a safety factor distribution map in the region, which is then integrated with the geological disaster potential map to evaluate the disaster risks in different regions.

2. The method for dynamic prediction of slope stability and geological parameter inversion according to claim 1 is characterized in that: The slope stability simulation model is established through the COMSOL multi-physics field simulation platform, and the simulation data is generated by combining orthogonal experimental design and automatic parameter optimization technology as follows: Use COMSOL multi-physics simulation platform to establish a slope stability simulation model; The slope stability simulation model is optimized by using automated parameter optimization technology and adaptive grid encryption technology; The orthogonal experimental design method was used to systematically change the geological parameters and rainfall, and the slope safety factor under different conditions was obtained through the optimized slope stability simulation model. By calculating the slope safety factor under different conditions, the simulation data of slope stability is obtained.

3. The method for dynamic prediction of slope stability and inversion of geological parameters according to claim 2 is characterized in that: The geological parameters include soil permeability and geomechanical properties.

4. The method for dynamic prediction of slope stability and inversion of geological parameters according to claim 1 is characterized in that: The simulation data is trained and optimized based on the variational autoencoder and the generative adversarial network to obtain the enhanced simulation data: The potential spatial distribution of slope stability data is learned through variational autoencoders to generate simulated data under different geological conditions and rainfall conditions to supplement the deficiencies in the original dataset. Then, the quality of the generated simulation data is optimized through the adversarial training mechanism through the generative adversarial network.

5. The method for dynamic prediction of slope stability and inversion of geological parameters according to claim 1, characterized in that: The geological parameter inversion based on the SBAS-InSAR ground deformation data and the enhanced simulation data is specifically as follows: The SBAS-InSAR ground deformation data are converted into the safety factor of the slope according to Newton's second law; Integrate the enhanced simulated data set and the safety factor obtained from SBAS-InSAR conversion; The geological parameter inversion model was constructed by integrating deep learning algorithms using Bayesian hyperparameter optimization convolutional neural networks and bidirectional long short-term memory neural networks; The integrated data are input into the geological parameter inversion model trained by Bayesian hyperparameter optimization, and the geological parameters are obtained by output inversion.

6. The method for dynamic prediction of slope stability and inversion of geological parameters according to claim 1, characterized in that: The safety factor distribution map of the geological parameters obtained by inversion within the region is specifically as follows: The geological parameters obtained by inversion are combined with real-time rainfall data and brought into the COMSOL multi-physics simulation platform to calculate the safety factor in the area; The safety factor data points obtained by simulation are converted into raster data to generate a safety factor distribution map in the area.

7. The method for dynamic prediction of slope stability and inversion of geological parameters according to claim 1, characterized in that: The safety factor data points obtained by simulation are converted into raster data through Kriging interpolation method.

Images