A Deep Learning-Based Intelligent Recognition Method and System for Geotechnical Mechanics Parameters

Abstract

This invention discloses a deep learning-based intelligent identification method and system for geotechnical mechanics parameters, aiming to accurately identify the elastic modulus, strength parameters, and permeability coefficient of soil and rock, supporting decision-making and design for engineering projects under complex geological conditions. Starting with data requirements analysis, the method clarifies parameter identification objectives and plans the layout for field testing. A basic dataset is constructed by comprehensively collecting and integrating multi-source data, including laboratory and field test data, and remote sensing information. The data undergoes preprocessing and analysis to remove noise and explore correlations between data points. The core steps involve using deep learning algorithms to construct a geotechnical mechanics model, defining and optimizing an objective function to achieve high-precision parameter inversion. The inversion results are verified and uncertainty analyzed. This method feeds the analysis results back to engineering design adjustments and decision support, optimizing the solution. By continuously monitoring and utilizing real-time data to dynamically adjust the model and strategies, the safe and efficient implementation of the project is ensured.

Description

Technical Field

[0001] This invention relates to the field of geotechnical mechanical parameter analysis technology, specifically to a deep learning-based intelligent identification method and system for geotechnical mechanical parameters. Background Technology

[0002] Geotechnical parameters, such as elastic modulus, strength parameters, and permeability coefficient, are influenced by various factors, including soil structure, particle size distribution, water content, and stress state. These factors exhibit complex nonlinear relationships. Traditional empirical formulas or simple statistical methods struggle to accurately describe this complexity. Furthermore, acquiring high-quality geotechnical test data is costly and complex, resulting in relatively limited data volume, and the data may contain noise and inconsistencies, affecting the accuracy and generalization ability of models. Geotechnical engineering design and safety assessment heavily rely on accurate estimation of geotechnical parameters. Traditional methods often rely on simplistic assumptions, making high-precision predictions difficult, especially under complex geological conditions. Uncertainties in geotechnical engineering stem from multiple sources, including the uncertainty of test data and the approximation of model assumptions. Effective methods are needed to quantify and manage these uncertainties to improve the reliability of engineering decisions. This method, through the application of deep learning technology, can automatically learn complex nonlinear patterns from a large amount of multi-source data, improving the accuracy and efficiency of parameter identification. At the same time, by combining optimization algorithms and uncertainty analysis, it makes engineering design and decision-making more scientific and flexible, which helps to reduce engineering risks, improve construction efficiency and safety, and meet the needs of modern geotechnical engineering projects for intelligent and refined management. Summary of the Invention

[0003] A deep learning-based intelligent identification method for geotechnical mechanics parameters includes the following steps;

[0004] S1. Project Planning and Data Requirements Analysis: Clarify the requirements for engineering objectives and parameter identification: elastic modulus, strength parameters, and permeability coefficient of soil and rock, and plan the layout of field testing and monitoring, including drilling, in-situ testing: static cone penetration test, standard penetration test and installation of remote monitoring sensors;

[0005] S2. Data Collection and Integration: Collect various types of field test data, including laboratory test data, in-situ test data, remote sensing and ground-penetrating radar data, integrate historical engineering data and regional geological data, and establish a basic dataset;

[0006] S3. Preprocessing and Data Analysis: The collected dataset is cleaned, outliers are removed, data is transformed and standardized, and statistical analysis is used to explore the correlation between data in advance, in preparation for subsequent modeling.

[0007] S4. Constructing the Inversion Model and Objective Function: A geotechnical mechanics model is established using deep learning algorithms, and the formulas for the geotechnical mechanics prediction model are integrated. ,in s Let represent stress, σ be the viscosity coefficient, reflecting the intensity of material flow or viscous effects and related to the material's flow resistance, and σ be the elastic modulus, describing the material's stiffness within its elastic range, expressed by the formula σ = σ / σ. The calculation is performed, where ν is Poisson's ratio, representing the negative ratio of transverse strain to longitudinal strain within the elastic range of the material, reflecting the degree of transverse expansion or contraction of the material under stress; G is the shear modulus, measuring the material's ability to resist shear deformation under tangential loads; and σ is the strain. , These represent the rates of change of stress and strain over time, i.e., stress rate and strain rate, respectively. F is the yield condition, which defines the critical point at which a material transitions from an elastic state to a plastic state. Deformation prediction is performed using a comprehensive model formula, taking into account the mechanical response under various working conditions. This provides support for geotechnical engineering design and safety assessment. An objective function is defined based on the actual engineering situation. Minimizing the objective function reflects the error between the model's predicted value and the measured value.

[0008] S5. Parameter Inversion and Optimization: Apply the selected intelligent algorithm to perform parameter inversion, iteratively optimize the objective function to obtain the optimal parameter set, and perform multiple iterations and algorithm parameter adjustments to ensure convergence and parameter accuracy;

[0009] S6. Result Validation and Uncertainty Analysis: Apply the parameters obtained from the inversion to the model, verify the predictive ability of the model by simulation and comparison with independent observation data, conduct sensitivity analysis and uncertainty assessment, understand the impact of parameter changes on model prediction, and evaluate the reliability of the results;

[0010] S7. Feedback Design and Decision Support: Adjust engineering design schemes based on inversion results, optimize structural design and construction schemes, provide risk assessment reports and recommendations, and provide scientific basis for decision-makers;

[0011] S8. Continuous monitoring and dynamic adjustment: During the project implementation process, key parameters are continuously monitored, real-time data is used for dynamic correction and optimization of the model, intelligent management and control are implemented, and design and construction strategies are adjusted based on the latest data feedback to ensure the safe and efficient progress of the project.

[0012] Furthermore, a deep learning-based intelligent identification method for geotechnical mechanics parameters is proposed.

[0013] The specific steps for establishing the geotechnical mechanics model using deep learning algorithms in step S4 are as follows:

[0014] S41. Data Preparation and Preprocessing: Data Collection: Collect historical data and experimental data including stress, strain, time variables, temperature, and moisture content. Data Preprocessing: Clean the data, handle missing and outlier values, standardize or normalize input features, and ensure data quality.

[0015] S42. Feature Engineering: Based on the mechanical model, features including strain rate, stress rate, elastic modulus, Poisson's ratio, shear modulus, and viscosity coefficient, as well as external environmental factors, are constructed. Key features for model prediction are selected through correlation analysis and principal component analysis.

[0016] S43. Constructing a Deep Learning Model: Divide the dataset into training, validation, and test sets. Use the training set to train the geotechnical mechanics model using a multilayer perceptron algorithm. Adjust the model parameters using a gradient descent optimization algorithm and integrate them to obtain the geotechnical mechanics model formula. ,in s Let represent stress, σ be the viscosity coefficient, reflecting the intensity of material flow or viscous effects and related to the material's flow resistance, and σ be the elastic modulus, describing the material's stiffness within its elastic range, expressed by the formula σ = σ / σ. The calculation is performed, where ν is Poisson's ratio, representing the negative ratio of transverse strain to longitudinal strain within the elastic range of the material, G is the shear modulus, and σ is the strain. , These represent the rates of change of stress and strain over time, respectively, and F is the yield condition. The model performance is evaluated on the validation set, and hyperparameters such as learning rate and batch size are adjusted by methods such as cross-validation to avoid overfitting and improve generalization ability.

[0017] S44. Testing and Evaluation: Use the test set to evaluate the final performance of the model, focusing on evaluation metrics such as prediction accuracy, RMSE, and R² score. Analyze the model's sensitivity to different input variables and understand which factors have the greatest impact on the prediction results.

[0018] S45. Prediction and Application of Working Condition Simulation: By inputting parameters under different working conditions such as loading conditions and changes in environmental factors into the model formula, the mechanical response of soil and rock is predicted: stress and strain. Based on the prediction results, the stability and safety of geotechnical engineering under specific conditions are evaluated. Based on new data from actual engineering applications, the model is continuously optimized and adjusted to ensure its long-term effectiveness and accuracy, providing a scientific basis for design optimization, risk management, and construction schemes.

[0019] Furthermore, a deep learning-based intelligent identification method for geotechnical mechanics parameters is proposed.

[0020] In step S5, the PSO algorithm is used to invert the parameters of the geotechnical mechanics model. The specific steps are as follows:

[0021] S51. Parameter space definition and constraint setting: Determine the reasonable range of each geotechnical parameter, set the upper and lower bounds of the parameters based on field test data, and ensure that the parameter combinations in the search space are physically feasible.

[0022] S52. Objective Function Construction: Define an objective function to measure the degree of agreement between the model's prediction results and the measured data, including mean squared error, mean absolute error, and correlation coefficient;

[0023] S53. PSO Parameters and Initial Population Settings: Considering the characteristics of geotechnical problems, the control parameters of PSO are finely adjusted: inertia weight and acceleration factor, in order to balance global exploration and local development capabilities. The initial population is randomly distributed in the parameter space. Based on known engineering experience and preliminary test results, biased initialization is performed to accelerate convergence.

[0024] S54. Model Solving and Fitness Evaluation Loop: Iterative calculation. For each particle, the predicted response under its parameters is calculated using the geotechnical mechanics model: displacement and stress distribution. The model prediction results are compared with field monitoring and experimental data. The objective function value is calculated as the fitness index. The speed and position are updated according to the PSO algorithm. At the same time, the potential correlation between geotechnical mechanics parameters is considered, and an adaptive weight adjustment strategy is adopted.

[0025] S55. Results Validation and Sensitivity Analysis: Conduct a detailed analysis of the optimal solution, including sensitivity analysis, examine the impact of small changes in key parameters on the model output, and use cross-validation or other statistical methods to evaluate the model's generalization ability.

[0026] S56. Application and feedback of parameter inversion results: Apply the optimized parameters back to the geotechnical mechanics model for detailed engineering analysis: stability assessment, seepage simulation, adjust the design scheme based on the model prediction results, assess construction risks, and propose improvement suggestions.

[0027] Furthermore, a deep learning-based intelligent identification system for geotechnical mechanics parameters is provided. This system is used to implement any of the deep learning-based intelligent identification methods for geotechnical mechanics parameters described above. The deep learning-based intelligent identification system for geotechnical mechanics parameters includes: a data requirement analysis module, a data acquisition and integration module, a model construction and optimization module, a result verification module, a decision support and feedback module, and a monitoring and dynamic adjustment module.

[0028] The data requirements analysis module involves identifying the geotechnical parameters (such as elastic modulus, strength parameters, permeability coefficient, etc.) that need to be identified based on the engineering objectives, designing field testing plans, including borehole layout, in-situ testing (such as static cone penetration test, standard penetration test), and deployment plans for remote monitoring sensors.

[0029] Data Acquisition and Integration Module: Performs field tests and laboratory analyses, collects various types of data, integrates historical data and geological data, constructs a basic database, and performs data cleaning, outlier removal, data format unification, standardization processing, preliminary statistical analysis, and explores the relationships between data.

[0030] Model building and optimization module: Based on deep learning algorithms, a geotechnical mechanics model is built, an objective function reflecting the prediction error is defined to provide a basis for model training, and an intelligent optimization algorithm is applied to invert the model parameters, iteratively optimize the objective function, and find the optimal parameter set;

[0031] Results Validation Module: Uses independent datasets to validate the model's predictive ability, performs sensitivity analysis and uncertainty assessment, and evaluates the model's reliability and parameter stability.

[0032] Decision support and feedback module: Adjust engineering design based on model results, provide risk assessment and decision suggestions, and formulate optimized construction plans;

[0033] Monitoring and Dynamic Adjustment Module: Continuous monitoring during the implementation of the project, dynamic correction of the model using real-time data, and adjustment of construction strategies based on the latest situation.

[0034] The beneficial effects of this invention are as follows: Deep learning algorithms can learn complex data features and nonlinear relationships, and compared with traditional methods, they can more accurately predict geotechnical parameters such as elastic modulus, strength, and permeability, thereby improving the accuracy of engineering design and analysis. The system can perform sensitivity analysis and uncertainty assessment, helping to identify key parameters and their impact on engineering stability, providing scientific support for risk management and decision-making, and reducing the risk of engineering failure. Real-time monitoring and dynamic model correction functions ensure the adaptability and flexibility of the project during implementation, enabling timely responses to changes in geological conditions, taking preventative measures, and avoiding potential problems.

[0035] Automated and intelligent analysis processes significantly shorten the time from data collection to decision-making, improve project management efficiency, and make decisions faster and more accurate. Attached Figure Description

[0036] Figure 1 This is a flowchart of a deep learning-based intelligent identification method for geotechnical mechanics parameters. Detailed Implementation

[0037] A deep learning-based intelligent identification method for geotechnical mechanics parameters includes the following steps;

[0038] S1. Project Planning and Data Requirements Analysis: Clarify the requirements for engineering objectives and parameter identification: elastic modulus, strength parameters, and permeability coefficient of soil and rock, and plan the layout of field testing and monitoring, including drilling, in-situ testing: static cone penetration test, standard penetration test and installation of remote monitoring sensors;

[0039] S2. Data Collection and Integration: Collect various types of field test data, including laboratory test data, in-situ test data, remote sensing and ground-penetrating radar data, integrate historical engineering data and regional geological data, and establish a basic dataset;

[0040] S3. Preprocessing and Data Analysis: The collected dataset is cleaned, outliers are removed, data is transformed and standardized, and statistical analysis is used to explore the correlation between data in advance, in preparation for subsequent modeling.

[0041] S4. Constructing the Inversion Model and Objective Function: A geotechnical mechanics model is established using deep learning algorithms, and the formulas for the geotechnical mechanics prediction model are integrated. ,in s Let represent stress, σ be the viscosity coefficient, reflecting the intensity of material flow or viscous effects and related to the material's flow resistance, and σ be the elastic modulus, describing the material's stiffness within its elastic range, expressed by the formula σ = σ / σ. The calculation is performed, where ν is Poisson's ratio, representing the negative ratio of transverse strain to longitudinal strain within the elastic range of the material, reflecting the degree of transverse expansion or contraction of the material under stress; G is the shear modulus, measuring the material's ability to resist shear deformation under tangential loads; and σ is the strain. , These represent the rates of change of stress and strain over time, i.e., stress rate and strain rate, respectively. F is the yield condition, which defines the critical point at which a material transitions from an elastic state to a plastic state. Deformation prediction is performed using a comprehensive model formula, taking into account the mechanical response under various working conditions. This provides support for geotechnical engineering design and safety assessment. An objective function is defined based on the actual engineering situation. Minimizing the objective function reflects the error between the model's predicted value and the measured value.

[0042] S5. Parameter Inversion and Optimization: Apply the selected intelligent algorithm to perform parameter inversion, iteratively optimize the objective function to obtain the optimal parameter set, and perform multiple iterations and algorithm parameter adjustments to ensure convergence and parameter accuracy;

[0043] S6. Result Validation and Uncertainty Analysis: Apply the parameters obtained from the inversion to the model, verify the predictive ability of the model by simulation and comparison with independent observation data, conduct sensitivity analysis and uncertainty assessment, understand the impact of parameter changes on model prediction, and evaluate the reliability of the results;

[0044] S7. Feedback Design and Decision Support: Adjust engineering design schemes based on inversion results, optimize structural design and construction schemes, provide risk assessment reports and recommendations, and provide scientific basis for decision-makers;

[0045] S8. Continuous monitoring and dynamic adjustment: During the project implementation process, key parameters are continuously monitored, real-time data is used for dynamic correction and optimization of the model, intelligent management and control are implemented, and design and construction strategies are adjusted based on the latest data feedback to ensure the safe and efficient progress of the project.

[0046] Furthermore, a deep learning-based intelligent identification method for geotechnical mechanics parameters is proposed.

[0047] The specific steps for establishing the geotechnical mechanics model using deep learning algorithms in step S4 are as follows:

[0048] S41. Data Preparation and Preprocessing: Data Collection: Collect historical data and experimental data including stress, strain, time variables, temperature, and moisture content. Data Preprocessing: Clean the data, handle missing and outlier values, standardize or normalize input features, and ensure data quality.

[0049] S42. Feature Engineering: Based on the mechanical model, features including strain rate, stress rate, elastic modulus, Poisson's ratio, shear modulus, and viscosity coefficient, as well as external environmental factors, are constructed. Key features for model prediction are selected through correlation analysis and principal component analysis.

[0050] S43. Constructing a Deep Learning Model: Divide the dataset into training, validation, and test sets. Use the training set to train the geotechnical mechanics model using a multilayer perceptron algorithm. Adjust the model parameters using a gradient descent optimization algorithm and integrate them to obtain the geotechnical mechanics model formula. ,in s Let represent stress, σ be the viscosity coefficient, reflecting the intensity of material flow or viscous effects and related to the material's flow resistance, and σ be the elastic modulus, describing the material's stiffness within its elastic range, expressed by the formula σ = σ / σ. The calculation is performed, where ν is Poisson's ratio, representing the negative ratio of transverse strain to longitudinal strain within the elastic range of the material, G is the shear modulus, and σ is the strain. , These represent the rates of change of stress and strain over time, respectively, and F is the yield condition. The model performance is evaluated on the validation set, and hyperparameters such as learning rate and batch size are adjusted by methods such as cross-validation to avoid overfitting and improve generalization ability.

[0051] S44. Testing and Evaluation: Use the test set to evaluate the final performance of the model, focusing on evaluation metrics such as prediction accuracy, RMSE, and R² score. Analyze the model's sensitivity to different input variables and understand which factors have the greatest impact on the prediction results.

[0052] S45. Prediction and Application of Working Condition Simulation: By inputting parameters under different working conditions such as loading conditions and changes in environmental factors into the model formula, the mechanical response of soil and rock is predicted: stress and strain. Based on the prediction results, the stability and safety of geotechnical engineering under specific conditions are evaluated. Based on new data from actual engineering applications, the model is continuously optimized and adjusted to ensure its long-term effectiveness and accuracy, providing a scientific basis for design optimization, risk management, and construction schemes.

[0053] Furthermore, a deep learning-based intelligent identification method for geotechnical mechanics parameters is proposed.

[0054] In step S5, the PSO algorithm is used to invert the parameters of the geotechnical mechanics model. The specific steps are as follows:

[0055] S51. Parameter space definition and constraint setting: Determine the reasonable range of each geotechnical parameter, set the upper and lower bounds of the parameters based on field test data, and ensure that the parameter combinations in the search space are physically feasible.

[0056] S52. Objective Function Construction: Define an objective function to measure the degree of agreement between the model's prediction results and the measured data, including mean squared error, mean absolute error, and correlation coefficient;

[0057] S53. PSO Parameters and Initial Population Settings: Considering the characteristics of geotechnical problems, the control parameters of PSO are finely adjusted: inertia weight and acceleration factor, in order to balance global exploration and local development capabilities. The initial population is randomly distributed in the parameter space. Based on known engineering experience and preliminary test results, biased initialization is performed to accelerate convergence.

[0058] S54. Model Solving and Fitness Evaluation Loop: Iterative calculation. For each particle, the predicted response under its parameters is calculated using the geotechnical mechanics model: displacement and stress distribution. The model prediction results are compared with field monitoring and experimental data. The objective function value is calculated as the fitness index. The speed and position are updated according to the PSO algorithm. At the same time, the potential correlation between geotechnical mechanics parameters is considered, and an adaptive weight adjustment strategy is adopted.

[0059] S55. Results Validation and Sensitivity Analysis: Conduct a detailed analysis of the optimal solution, including sensitivity analysis, examine the impact of small changes in key parameters on the model output, and use cross-validation or other statistical methods to evaluate the model's generalization ability.

[0060] S56. Application and feedback of parameter inversion results: Apply the optimized parameters back to the geotechnical mechanics model for detailed engineering analysis: stability assessment, seepage simulation, adjust the design scheme based on the model prediction results, assess construction risks, and propose improvement suggestions.

[0061] A deep learning-based intelligent identification system for geotechnical mechanics parameters is provided. This system is used to implement any of the deep learning-based intelligent identification methods for geotechnical mechanics parameters described above. The deep learning-based intelligent identification system for geotechnical mechanics parameters includes: a data requirement analysis module, a data acquisition and integration module, a model construction and optimization module, a result verification module, a decision support and feedback module, and a monitoring and dynamic adjustment module.

[0062] The data requirements analysis module involves identifying the geotechnical parameters (such as elastic modulus, strength parameters, permeability coefficient, etc.) that need to be identified based on the engineering objectives, designing field testing plans, including borehole layout, in-situ testing (such as static cone penetration test, standard penetration test), and deployment plans for remote monitoring sensors.

[0063] Data Acquisition and Integration Module: Performs field tests and laboratory analyses, collects various types of data, integrates historical data and geological data, constructs a basic database, and performs data cleaning, outlier removal, data format unification, standardization processing, preliminary statistical analysis, and explores the relationships between data.

[0064] Model building and optimization module: Based on deep learning algorithms, a geotechnical mechanics model is built, an objective function reflecting the prediction error is defined to provide a basis for model training, and an intelligent optimization algorithm is applied to invert the model parameters, iteratively optimize the objective function, and find the optimal parameter set;

[0065] Results Validation Module: Uses independent datasets to validate the model's predictive ability, performs sensitivity analysis and uncertainty assessment, and evaluates the model's reliability and parameter stability.

[0066] Decision support and feedback module: Adjust engineering design based on model results, provide risk assessment and decision suggestions, and formulate optimized construction plans;

[0067] Monitoring and Dynamic Adjustment Module: Continuous monitoring during the implementation of the project, dynamic correction of the model using real-time data, and adjustment of construction strategies based on the latest situation.

Claims

1. A deep learning-based intelligent identification method for geotechnical mechanics parameters, characterized in that, Comprising the following steps: S1. Project planning and data requirement analysis: Identify the engineering objectives and the need for parameter identification: elastic modulus of rock and soil, strength parameters, permeability coefficient, and plan the field test and monitoring arrangement, including drilling, in-situ test: static sounding, standard penetration test and installation of remote monitoring sensors; S2. Data collection and integration: Collect various types of field test data, including laboratory test data, in-situ test data, remote sensing and geological radar data, integrate historical engineering data and regional geological data, and establish a basic data set; S3. Preprocessing and data analysis: Clean the collected data set, remove outliers, perform data conversion and standardization, and use statistical analysis to preliminarily explore the correlation between data, preparing for subsequent modeling; S4. Constructing inversion model and objective function: Establishing geotechnical mechanics model using deep learning algorithm and integrating to obtain geotechnical mechanics model formula wherein S4. Model establishment: Apply deep learning algorithms to establish rock and soil mechanics models, including data preparation and preprocessing, feature engineering, model training and testing, and model evaluation; represents stress, η is the viscosity coefficient, which reflects the strength of material flow and viscosity effect, and is related to the flow resistance of the material, E is the elastic modulus, which describes the stiffness of the material within the elastic range, and is calculated by the formula wherein v is the Poisson's ratio, which represents the negative ratio of transverse strain to longitudinal strain of the material within the elastic range, G is the shear modulus, and ε is the strain, , respectively represent the rate of change of stress and strain with time, F is the yield condition, and the deformation prediction is carried out using the comprehensive model formula, considering the mechanical response under various working conditions, providing support for geotechnical engineering design and safety evaluation, and defining the objective function according to the actual engineering situation, minimizing the error between the model prediction value and the measured value. S5. Parameter inversion and optimization: Apply PSO algorithm for rock and soil mechanics model parameter inversion, iterative optimization of objective function to obtain optimal parameter set, multiple iterations and algorithm parameter adjustment; S6. Result verification and uncertainty analysis: Apply the inverted parameters to the model, verify the model's prediction ability by simulating and comparing with independent observation data, perform sensitivity analysis and uncertainty evaluation, understand the impact of parameter changes on model prediction, and evaluate the reliability of the results; S7. Feedback design and decision support: Adjust the engineering design scheme according to the inversion results, optimize the structure design and construction scheme, provide risk assessment report and suggestion, and provide scientific basis for decision makers; S8. Continuous monitoring and dynamic adjustment: Continuously monitor key parameters during the implementation of the project, use real-time data for dynamic correction and optimization of the model, implement intelligent control, adjust the design and construction strategy according to the latest data feedback, and ensure the safety and efficient progress of the project.

2. The intelligent rock and soil mechanics parameter identification method based on deep learning according to claim 1, characterized in that: The specific steps of establishing a rock and soil mechanics model using deep learning algorithms in step S4 are as follows: S41. Data preparation and preprocessing: Collect historical data containing stress, strain, time variable, temperature, moisture content, and experimentally obtained data, preprocess the data: clean the data, handle missing values and outliers, standardize or normalize the input features, and ensure data quality; S42. Feature engineering: According to the mechanics model, construct features including strain rate, stress rate, elastic modulus, Poisson's ratio, shear modulus and viscosity coefficient, as well as external environmental factors, and select key features for model prediction through correlation analysis and principal component analysis method; S43. Constructing a deep learning model: dividing the data set into training set, validation set and test set, using the training set to train the geotechnical mechanics model through the multilayer perceptron algorithm, adjusting the model parameters through the gradient descent optimization algorithm, and integrating the geotechnical mechanics model formula wherein S44. Test and evaluation: Evaluate the final performance of the model using the test set, focus on prediction accuracy, RMSE, R² score evaluation indicators, analyze the sensitivity of the model to different input variables, and understand which factors have the greatest impact on the prediction results; represents stress, η is the viscosity coefficient, which reflects the strength of material flow and viscous effect, and is related to the flow resistance of the material, E is the elastic modulus, which describes the stiffness of the material within the elastic range, and is calculated by the formula wherein v is the Poisson's ratio, which represents the negative ratio of transverse strain to longitudinal strain of the material within the elastic range, G is the shear modulus, and ε is the strain, , respectively represent the rate of change of stress and strain with time, F is the yield condition, and the model performance is evaluated on the validation set, and the hyperparameters such as learning rate and batch size are adjusted through cross-validation and other methods to avoid overfitting and improve generalization ability; ​ S45. Predicting and applying working condition simulation: input different working condition parameters into the model formula: loading conditions, environmental factor changes, predict the mechanical response of rock and soil: stress, strain, evaluate the stability and safety of geotechnical engineering under specific conditions according to the prediction results, and continuously optimize and adjust the model according to new data in actual engineering application, ensure the long-term effectiveness and accuracy of the model, and provide scientific basis for design optimization, risk management and construction scheme.

3. The intelligent identification method of geotechnical mechanics parameters based on deep learning according to claim 1, characterized in that, In step S5, the PSO algorithm is used for geotechnical mechanics model parameter inversion, and the specific steps are as follows: S51. Parameter space definition and constraint condition setting: determine the reasonable range of each geotechnical mechanics parameter, set the upper and lower boundaries of the parameters based on the field test data, and ensure that the parameter combinations in the search space are physically feasible; S52. Objective function construction: define the objective function to measure the degree of agreement between the model prediction results and the measured data, mean square error, mean absolute error, correlation coefficient; S53. PSO parameter and initial population setting: considering the characteristics of geotechnical mechanics problems, fine-tune the control parameters of PSO: inertia weight, acceleration factor, to balance the global exploration and local exploitation ability, the initial population is randomly distributed in the parameter space, and the biased initialization is carried out based on the known engineering experience and preliminary test results to speed up the convergence; S54. Model solution and fitness evaluation cycle: iterative calculation, for each particle, use the geotechnical mechanics model to calculate the predicted response under its parameters: displacement, stress distribution, compare the model prediction results with the field monitoring and experimental data, calculate the objective function value as the fitness index, update the velocity and position according to the PSO algorithm, and consider the potential correlation between geotechnical mechanics parameters, use adaptive weight adjustment strategy; S55. Result verification and sensitivity analysis: detailed analysis of the optimal solution, including sensitivity analysis, checking the influence of small changes in key parameters on model output, using cross-validation or other statistical methods to evaluate the generalization ability of the model; S56. Application and feedback of parameter inversion results: apply the optimized parameters back to the geotechnical mechanics model for detailed engineering analysis: stability evaluation, seepage simulation, adjust the design scheme according to the model prediction results, evaluate the construction risk, and propose improvement suggestions.

4. A deep learning-based intelligent identification system for geotechnical mechanics parameters, characterized in that, The intelligent identification method of geotechnical mechanics parameters based on deep learning is used to realize the intelligent identification method of geotechnical mechanics parameters based on deep learning according to any one of claims 1-3; the intelligent identification system of geotechnical mechanics parameters based on deep learning comprises: data requirement analysis module, data collection and integration module, model construction and optimization module, result verification module, decision support and feedback module, monitoring and dynamic adjustment module; The data requirement analysis module: according to the engineering target, it is clear that the geotechnical mechanics parameters to be identified are needed, the field test scheme is designed, including the layout of drill holes, in-situ test and remote monitoring sensor layout plan; Data Collection and Integration Module: Conduct field tests and laboratory analyses, collect various data, integrate historical data and geological information, build a basic database, and perform data cleaning, outlier removal, data format unification, standardization processing, preliminary statistical analysis, and exploration of relationships between data. Model Construction and Optimization Module: Construct geotechnical mechanics models based on deep learning algorithms, define objective functions reflecting prediction errors, provide basis for model training, apply intelligent optimization algorithms to inverse model parameters, iteratively optimize objective functions, and find optimal parameter sets. Result Verification Module: Verify model prediction ability using independent data sets, perform sensitivity analysis and uncertainty evaluation, and evaluate model reliability and parameter stability. Decision Support and Feedback Module: Adjust engineering design based on model results, provide risk assessment and decision recommendations, and develop optimized construction plans. Monitoring and Dynamic Adjustment Module: Implement continuous monitoring during the construction process, dynamically correct models using real-time data, and adjust construction strategies based on the latest information.

Images