Geological risk prediction method, system, equipment, medium and product in tunnel construction

By combining data preprocessing and feature engineering with various machine learning models, a self-learning and adaptive geological risk prediction system was constructed, which solved the problems of real-time performance and accuracy in geological risk prediction during tunnel construction, and improved construction safety and efficiency.

CN118966784APending Publication Date: 2024-11-15CHINA RAILWAY 15TH BUREAU GROUP CORPORATION LIMITED +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411159275.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing methods for predicting geological risks in tunnel construction rely on the experience of geological experts, lack self-learning and adaptive capabilities, struggle to handle large-scale, multi-source, heterogeneous data, and are insufficient in real-time performance and accuracy, failing to meet the demands of rapidly changing construction environments.

Method used

Data preprocessing and feature engineering techniques are used to process and select features from geological data. Multiple machine learning models are then used for training to build a self-learning and adaptive geological risk prediction model. This model enables real-time acquisition and analysis of geological data, achieving dynamic risk assessment.

Benefits of technology

It improves the accuracy and efficiency of geological risk prediction during tunnel construction, enables real-time response to changes in geological risks, provides scientific risk management and decision support, and reduces construction risks and costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966784A_ABST
    Figure CN118966784A_ABST
Patent Text Reader

Abstract

The invention discloses a geological risk prediction method, system and device in tunnel construction, a medium and a product, and relates to the field of crossing of geological engineering and data science, and the method comprises the steps: employing a data preprocessing technology to carry out the preprocessing of a geological data sample of a tunnel construction region corresponding to each geological risk prediction result, performing feature selection on the preprocessed geological data samples by using a feature engineering technology to obtain key geological feature samples corresponding to different geological risk prediction results, and training various machine learning models by using the key geological feature samples; carrying out model verification and testing on the plurality of trained machine learning models, and taking the model with the optimal performance in the plurality of trained machine learning models as a geological risk prediction model according to a model verification and testing result; and utilizing the geological risk prediction model to predict the geological risk in the tunnel construction process in real time. The geological risk prediction in the tunnel construction process can be efficiently and accurately realized, and geological risk changes can be responded in real time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the interdisciplinary field of geological engineering and data science, and in particular to a method, system, equipment, medium and product for predicting geological risks in tunnel construction. Background Art

[0002] In the traditional tunnel construction process, the prediction of geological risks mainly relies on the experience and qualitative analysis of geological experts, which often has certain limitations. With the development of geological exploration technology, a large amount of geological data, including rock type, structure, groundwater level, soil moisture, etc., can be collected and recorded. However, due to the complexity and diversity of geological data and the variability of the geological environment, how to accurately identify and predict geological risks from massive data has become a technical challenge.

[0003] Some existing geological risk assessment methods attempt to apply statistical analysis or simple machine learning algorithms to assist prediction, but these methods are often limited to specific data types or geological environments, and when processing large-scale, multi-source heterogeneous data, the prediction accuracy and generalization ability are limited, that is, the prediction accuracy and generalization ability are insufficient. In addition, these methods also have deficiencies in real-time data processing and dynamic risk assessment, and it is difficult to meet the needs of rapidly changing construction environments. In addition, the existing technology also has the defect of lack of self-learning and self-adaptation capabilities, that is, the existing models generally do not have the ability to self-learn and adapt to new geological conditions. Based on this, the field urgently needs a geological risk prediction method during tunnel construction that is efficient, accurate, and can respond to changes in geological risks in real time to improve the safety and efficiency of tunnel construction. Summary of the invention

[0004] The purpose of this application is to provide a method, system, equipment, medium and product for predicting geological risks in tunnel construction, which can realize efficient and accurate geological risk prediction during tunnel construction and can respond to geological risk changes in real time.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for predicting geological risks during tunnel construction, the method comprising:

[0007] Obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during tunnel construction;

[0008] The data preprocessing technology is used to preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result, so as to obtain the preprocessed geological data samples corresponding to different geological risk prediction results;

[0009] Feature engineering technology is used to select features of the preprocessed geological data samples corresponding to each geological risk prediction result, and key geological feature samples corresponding to different geological risk prediction results are obtained;

[0010] Using key geological feature samples corresponding to different geological risk prediction results to train multiple machine learning models, multiple trained machine learning models are obtained;

[0011] Conduct model verification and testing on multiple trained machine learning models. Based on the results of model verification and testing, use the model with the best performance among multiple trained machine learning models as the geological risk prediction model.

[0012] Real-time acquisition of geological data of the tunnel construction area to be predicted;

[0013] The data preprocessing technology is used to preprocess the geological data of the tunnel construction area to be predicted to obtain preprocessed geological data;

[0014] Using the feature engineering technology to perform feature selection on the preprocessed geological data to obtain key geological features;

[0015] The key geological features are input into the geological risk prediction model, the geological risk prediction model is used to predict the geological risks in the tunnel construction process in real time, and the geological risk prediction results corresponding to the key features are output.

[0016] In a second aspect, the present application provides a system for predicting geological risks during tunnel construction, the system comprising:

[0017] A geological data collection module, used to obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during the tunnel construction process;

[0018] A data preprocessing module is used to preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result by using data preprocessing technology to obtain preprocessed geological data samples corresponding to different geological risk prediction results;

[0019] A feature selection module is used to select features of the preprocessed geological data samples corresponding to each geological risk prediction result by using feature engineering technology to obtain key geological feature samples corresponding to different geological risk prediction results;

[0020] A machine learning model training module is used to train multiple machine learning models using key geological feature samples corresponding to different geological risk prediction results to obtain multiple trained machine learning models;

[0021] Model verification and testing module, used to verify and test a variety of trained machine learning models. According to the results of model verification and testing, the model with the best performance among the multiple trained machine learning models is used as the geological risk prediction model;

[0022] A real-time data acquisition module is used to acquire the geological data of the tunnel construction area to be predicted in real time;

[0023] A real-time data preprocessing module, used to preprocess the geological data of the tunnel construction area to be predicted by using the data preprocessing technology to obtain preprocessed geological data;

[0024] A real-time feature selection module, used to select features of the pre-processed geological data using the feature engineering technology to obtain key geological features;

[0025] The real-time risk prediction module is used to input the key geological features into the geological risk prediction model, use the geological risk prediction model to predict the geological risks in the tunnel construction process in real time, and output the geological risk prediction results corresponding to the key features.

[0026] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for predicting geological risks in tunnel construction.

[0027] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for predicting geological risks in tunnel construction.

[0028] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for predicting geological risks in tunnel construction.

[0029] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0030] The present application provides a method, system, equipment, medium and product for geological risk prediction in tunnel construction, which uses data preprocessing technology to preprocess geological data, and uses feature engineering technology to select features of the preprocessed geological data to obtain the most critical features for geological risk prediction (i.e., key geological features), and uses the key geological features to train a machine learning model. Based on the complex model structure of the machine learning model, the deep-level relationships and features in the geological data are captured to achieve real-time geological risk prediction. Through advanced data preprocessing and feature engineering, the problem of how to effectively identify and extract key geological features from massive and complex geological data is solved, ensuring that the machine learning model used for geological risk prediction can focus on the most predictive data. By effectively applying these advanced data science and machine learning technologies to geological risk prediction, especially in the specific field of tunnel construction, data science and machine learning technologies can process and analyze large-scale complex data sets, automatically identify geological risk patterns and features, thereby improving the accuracy and efficiency of geological risk prediction during tunnel construction, and achieving efficient, accurate and real-time response to geological risk changes in tunnel construction. Geological risk prediction during tunnel construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 A schematic diagram of a flow chart of a method for predicting geological risks in tunnel construction provided in one embodiment of the present application;

[0033] Figure 2 This is the data preprocessing flow chart for this application;

[0034] Figure 3 A flow chart for feature selection for this application;

[0035] Figure 4 Model training and validation flow chart for this application;

[0036] Figure 5 This is a simplified diagram of the system modules for this application;

[0037] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0039] The purpose of this application is to provide a method, system, equipment, medium and product for predicting geological risks in tunnel construction, which can realize efficient and accurate geological risk prediction during tunnel construction and can respond to geological risk changes in real time.

[0040] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0041] like Figure 1 As shown, the present application provides a method for predicting geological risks in tunnel construction, comprising:

[0042] Step 101: Obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during the tunnel construction process.

[0043] Step 102: Preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result using data preprocessing technology to obtain preprocessed geological data samples corresponding to different geological risk prediction results.

[0044] In step 102, the data preprocessing technology includes:

[0045] Use time series analysis methods and spatial filtering techniques to remove noise; time series analysis methods are used to remove periodic noise from time series data; spatial filtering techniques are used to reduce noise caused by spatial autocorrelation of spatial data; time series analysis methods include Bessel filters, Fourier transforms, and wavelet transforms; spatial filtering techniques include Gaussian filters.

[0046] Outlier processing methods are used to remove outliers; outlier processing methods include box plot method, IQR method and Z-score method.

[0047] Missing values ​​were filled using the k-nearest neighbor algorithm.

[0048] The z-score method was used for data normalization.

[0049] Step 103: Use feature engineering technology to perform feature selection on the preprocessed geological data samples corresponding to each geological risk prediction result to obtain key geological feature samples corresponding to different geological risk prediction results.

[0050] In step 103, the feature engineering technology includes:

[0051] Characteristic statistical analysis was performed using statistical analysis methods.

[0052] By calculating the correlation between features and target variables, features that are highly correlated with geological risks are identified.

[0053] Feature selection techniques are applied to select key geological features from features that are highly correlated with geological risks; feature selection techniques include recursive feature elimination and principal component analysis; recursive feature elimination is based on the random forest model to evaluate feature importance.

[0054] Before step 103, the following steps are also included:

[0055] For the preprocessed geological data samples corresponding to each geological risk prediction result, data fusion technology is used to fuse multi-source data of the preprocessed geological data samples to obtain fused data; the fused data is used for feature selection using feature engineering technology to obtain key geological feature samples.

[0056] Step 104: Use key geological feature samples corresponding to different geological risk prediction results to train multiple machine learning models to obtain multiple trained machine learning models.

[0057] In step 104, the multiple machine learning models include a random forest model, a support vector machine model, a neural network model, and a deep learning network model.

[0058] Step 105: Perform model verification and testing on multiple trained machine learning models. According to the results of model verification and testing, the model with the best performance among the multiple trained machine learning models is used as the geological risk prediction model.

[0059] Step 106: Acquire geological data of the tunnel construction area to be predicted in real time.

[0060] Step 107: Preprocessing the geological data of the predicted tunnel construction area using data preprocessing technology to obtain preprocessed geological data.

[0061] Step 108: Use feature engineering technology to perform feature selection on the preprocessed geological data to obtain key geological features.

[0062] Step 109: input the key geological features into the geological risk prediction model, use the geological risk prediction model to predict the geological risks in the tunnel construction process in real time, and output the geological risk prediction results corresponding to the key features.

[0063] The step 109 also includes:

[0064] Model adaptation technology is applied to the geological risk prediction model to enable the geological risk prediction model to have self-learning and self-adaptive capabilities.

[0065] The following is a specific example to illustrate the technical solution of the method for predicting geological risks in tunnel construction in this application:

[0066] The geological risk prediction method in tunnel construction of this application is a geological risk prediction technology in the intersection of geological engineering and data science. Specifically, it is a method that uses data science and machine learning algorithms to predict and evaluate geological risks that may be encountered during tunnel construction. This technology combines multiple steps such as geological data collection, preprocessing, feature selection, machine learning model training, and risk prediction, aiming to provide an efficient and accurate geological risk assessment solution to enhance the safety and efficiency of tunnel construction. By automatically processing and analyzing large amounts of geological data, the model of this application can identify and predict potential geological problems, provide scientific decision-making support for the construction team, and thus reduce construction risks and costs.

[0067] In order to overcome the problems existing in the existing geological risk assessment methods, in recent years, the application of data science and machine learning techniques in the field of geology has gradually increased. These technologies can process and analyze large-scale complex data sets, automatically identify geological risk patterns and characteristics, and thus improve the accuracy and efficiency of predictions. Advanced machine learning algorithms such as deep learning can capture the deep relationships and characteristics in geological data by building complex model structures, providing new solutions for geological risk prediction. However, how to effectively apply these advanced data science techniques to geological risk prediction, especially in the specific field of tunnel construction, is still a technical problem that needs to be solved urgently. This application is proposed in this context, aiming to use data science and machine learning techniques to provide a more accurate and efficient geological risk prediction model, that is, a geological risk prediction model based on data science and machine learning (i.e., data science and machine learning techniques).

[0068] Compared with the prior art, this application provides a geological risk prediction model based on data science and machine learning technology, which aims to improve the accuracy and efficiency of geological risk prediction during tunnel construction. The model comprehensively collects geological, geochemical, geophysical and remote sensing data, and uses advanced data preprocessing and feature selection techniques to construct a machine learning algorithm that can automatically identify geological risk patterns and characteristics. Key geological features are screened using methods such as recursive feature elimination, and the model is trained and optimized using algorithms such as random forests, support vector machines or deep learning networks. The model can perform real-time risk assessment of new geological data and output specific risk levels, thereby providing a scientific risk management basis for tunnel construction. Compared with traditional methods, the model of this application has higher prediction accuracy, can handle complex multi-source heterogeneous data, and has the ability of self-learning and self-adaptation.

[0069] The geological risk prediction model based on data science and machine learning provided in this application includes the following steps:

[0070] The collection step is used to collect geological data of the tunnel construction area, including but not limited to rock type, rock structure, fault distribution, groundwater level, soil moisture, geochemical composition, geophysical parameters and remote sensing image data, and the data is collected through geological exploration equipment, geochemical analysis instruments, geophysical exploration instruments and remote sensing satellites.

[0071] Preprocessing steps such as Figure 2 As shown, it is used to clean the collected data, including removing outliers, filling missing values, and performing data standardization and normalization to facilitate subsequent machine learning algorithm processing. Specifically, Bessel filters are used to remove noise, the k-nearest neighbor algorithm is used to fill missing values, and the z-score method is used to standardize data.

[0072] The feature selection step is used to select the features that have the greatest impact on geological risk prediction from the preprocessed data through statistical analysis and machine learning algorithms, such as recursive feature elimination, principal component analysis, or adaptive gain attribute selection algorithm. Specifically, it includes building an evaluation model that can quantify the contribution of each feature to the prediction of geological risk, and sort them according to the size of the contribution, and select the top-ranked features for subsequent model training.

[0073] The model training step is used to use machine learning algorithms, such as random forests, support vector machines, neural networks or deep learning networks, to train the selected features to build a geological risk prediction model. Specifically, it includes dividing the data set into a training set and a test set, optimizing the model parameters using a cross-validation method, using a grid search or Bayesian optimization algorithm to tune the hyperparameters, and validating on the test set to evaluate the prediction performance of the model.

[0074] The risk prediction step is used to use the trained model to conduct risk assessment on new geological data and output the probability or level of geological risk through the model. Specifically, it includes real-time reception of geological monitoring data during the construction process, such as groundwater level changes and soil stress collected through the sensor network, using the trained model to conduct risk assessment on real-time data, and providing early warning information or recommended preventive measures to the construction team based on the risk level output by the model.

[0075] The collection step (i.e., data collection step) further includes: using geological radar to detect underground structures and obtain data on rock structures and fault distribution. Analyzing soil and groundwater samples through geochemical sampling to obtain geochemical composition data. Using geophysical methods, such as seismic wave tomography, to collect geophysical parameters of underground structures. Using remote sensing technology, such as satellite remote sensing or remote sensing equipment carried by drones, to obtain remote sensing image data of the surface and underground.

[0076] The preprocessing step (i.e., data preprocessing step) further includes: for time series data, using time series analysis methods, such as Fourier transform or wavelet transform, to remove periodic noise. For spatial data, using spatial filters, such as Gaussian filters, to reduce noise caused by spatial autocorrelation. For multi-source data, using data fusion techniques, such as Kalman filters or extended Kalman filters, to integrate data from different sensors.

[0077] The feature selection step further includes: building a mutual information-based evaluation model to assess the correlation between features and geological risk. Using model-based feature selection methods, such as LASSO (least absolute shrinkage and selection operator), to select important features. Using tree-based algorithms, such as XGBoost or LightGBM, to perform feature importance assessment.

[0078] The model training steps further include: for the random forest algorithm, multiple decision trees are constructed and voting mechanisms are performed to improve the generalization ability of the model. For the support vector machine, a kernel function is used to map the data to a high-dimensional space to find the optimal decision boundary. For the neural network, a multi-layer perceptron structure is designed, and the weights are updated using the back-propagation algorithm, and regularization techniques are used to prevent overfitting. For the deep learning network, a convolutional neural network (CNN) or recurrent neural network (RNN) structure is used to optimize for specific geological data characteristics.

[0079] The risk prediction step further includes: designing a real-time data stream processing system to process real-time geological data collected by the sensor network. Implementing a dynamic threshold determination mechanism to dynamically adjust the threshold of risk assessment based on historical data and statistical characteristics of geological risks. Developing a visual interface to present the risk prediction results to the construction team in the form of graphs or charts for easy understanding and decision-making.

[0080] The model can automatically adjust model parameters to adapt to new geological conditions by continuously learning new geological data, and has the ability to self-learn and adapt, including: implementing an online learning mechanism that enables the model to incrementally learn new geological data without retraining the entire model. Designing a model evaluation framework to regularly evaluate the performance of the model, such as accuracy, recall, and F1 score, and adjust the model based on the evaluation results. Using transfer learning technology, a model trained in one geological environment is applied to another similar geological environment to accelerate the model training process.

[0081] The present application also provides a method for using the above-mentioned geological risk prediction model based on data science and machine learning, comprising the following steps: S1. Deploy the above-mentioned model into the real-time monitoring system of tunnel construction, specifically including integration into existing construction monitoring software or hardware. S2. Regularly update the geological data required for model training, including an automated data collection system and a manual data verification process. S3. Based on the prediction results of the model, formulate a risk management plan and emergency response strategy, specifically including the correspondence between risk levels and response measures. S4. Regularly evaluate the prediction accuracy of the model, and optimize the model based on the evaluation results, including feature engineering, model structure adjustment, and parameter tuning.

[0082] The technical solution of this application is to build a comprehensive data science and machine learning framework for predicting geological risks in tunnel construction. The solution includes the following specific steps:

[0083] 1. Data collection and integration:

[0084] Comprehensively collect geological data of the tunnel construction area using geological exploration equipment (such as geological radar, seismic wave tomography equipment), geochemical analysis instruments, geophysical exploration instruments and remote sensing technology.

[0085] The types of data collected include, but are not limited to, geological structure, rock type, fault distribution, groundwater level, soil moisture, geochemical composition, geophysical parameters, and high-resolution remote sensing images.

[0086] 2. Data preprocessing and cleaning:

[0087] The collected raw data is cleaned to remove noise and outliers and fill in missing data to improve data quality.

[0088] Time series analysis methods such as Bessel filters are used to process time series data, and spatial filtering techniques such as Gaussian filters are used to process spatial data.

[0089] In geological risk prediction, time series data usually refers to data points that change over time. These data can reflect the dynamic changes of geological conditions over time. Groundwater level: Groundwater level may change with factors such as seasons, weather conditions or groundwater extraction. Soil moisture: Soil moisture may change over time due to activities such as rainfall, evaporation or irrigation. The results obtained after processing time series data include: 1. Denoising: Use methods such as Bessel filters to remove random noise in the data, making the data trend clearer. 2. Outlier detection and processing: Identify abnormal points in the data, which may be caused by measurement errors or other non-geological factors, and then appropriately process these outliers, such as deleting or replacing them. 3. Trend analysis: After filtering, trends such as groundwater level or soil moisture that change over time can be more accurately identified. 4. Periodicity analysis: If the data exhibits periodic changes, time series analysis can reveal these periodic patterns. 5. Data interpolation: In the case of missing data, time series analysis methods can be used to interpolate data and estimate the values ​​of missing time points.

[0090] Spatial data specifically include: 1. Geological structure: describes the distribution and morphology of underground rock layers. 2. Rock type: identifies the rock types in different regions. 3. Fault distribution: indicates the spatial distribution of faults. 4. Geochemical composition: the distribution of chemical components in soil or rocks at different locations. 5. Geophysical parameters: such as the spatial distribution of seismic wave velocity. 6. High-resolution remote sensing images: provide spatial distribution information of surface features. These data usually have spatial attributes and can be used to analyze the distribution of geological features in different geographical locations. The results obtained after processing spatial data include: (1) Reduction of spatial autocorrelation: using spatial filters, such as Gaussian filters, to reduce noise caused by spatial autocorrelation and improve the spatial consistency of data. (2) Feature extraction: by analyzing spatial data, extract features useful for geological risk prediction, such as fault density, rock type distribution pattern, etc. (3) Spatial data interpolation: for missing values ​​or sparse areas in spatial data, spatial interpolation technology can be used to estimate the data values ​​of these areas. (4) Spatial pattern recognition: identifying patterns in spatial data, such as the changing trend of groundwater level and the spatial distribution characteristics of soil moisture. (5) Data fusion: Use data fusion techniques, such as Kalman filters, to integrate spatial data from different sensors and data sources to obtain a more comprehensive and accurate representation of geological features. (6) Spatial data standardization: Convert spatial data from different sources to the same scale or range for easy comparison and analysis. (7) Improve the spatial prediction ability of the model: The processed spatial data can improve the accuracy of the machine learning model in predicting the spatial distribution of geological risks.

[0091] 3. Multi-source data fusion:

[0092] Data fusion techniques, such as Kalman filters, are used to integrate information from different sensors and data sources to obtain a more comprehensive and accurate representation of geological features.

[0093] The integration process is as follows: 1. Data preprocessing: First, the data from different sensors are preprocessed, including calibration, denoising, and standardization, to ensure the quality of the data. 2. State estimation: The Kalman filter represents the state of the system through a state vector. This state vector may include geological features such as groundwater level, soil moisture, and rock type. 3. Initial state and uncertainty: Set the initial state estimate and the uncertainty of the initial state (covariance matrix). This provides a starting point for the filtering process. 4. Prediction: Based on the dynamic model of the system, predict the state at the next moment. The dynamic model can be a mathematical model based on physical processes that describes the evolution of the state over time. 5. Measurement update: When new observation data (from different sensors) arrives, the Kalman filter combines these observations and predicted states to correct the predicted state estimate through an update step. 6. Uncertainty update: The uncertainty of the state estimate is updated at the same time, which reflects the confidence of the integrated data. 7. Iterative process: As time goes by and new data continues to arrive, the Kalman filter repeats the prediction and update steps to provide continuous state estimates. The integration results are as follows: (1) More accurate state estimation: By integrating multiple data sources, the Kalman filter provides a more accurate state estimation than a single data source. (2) Reduce noise and uncertainty: The filtering process helps to reduce random noise in the data and reduce the uncertainty of the estimate. (3) Dynamic tracking: The Kalman filter can dynamically track the changes in geological features over time, even when there are time delays or inconsistencies between data sources. (4) Stability: Even if the data of some sensors are erroneous or missing, the Kalman filter can use the data of other sensors to compensate and improve the stability of the overall data. (5) Comprehensive feature representation: The integrated data provides a comprehensive feature representation, which helps to better understand the complexity of the geological environment. (6) Support decision making: Accurate state estimation and feature representation provide a solid data foundation for geological risk prediction and construction decision-making. (7) Model training and validation: The integrated data can be used to train and validate machine learning models to improve the predictive ability and generalization of the models.

[0094] 4. Feature Engineering:

[0095] like Figure 3 As shown, features highly correlated with geological risks are selected through statistical analysis and domain knowledge.

[0096] Feature selection techniques such as principal component analysis and recursive feature elimination are applied to reduce feature dimensions and improve the generalization ability of the model.

[0097] Among them, statistical analysis and domain knowledge are used to select features that are highly correlated with geological risks. The following are the specific steps to determine these features: 1. Data exploration: First, exploratory data analysis is performed on the collected data to understand the basic statistical properties of the data, such as mean, median, standard deviation, etc. 2. Correlation analysis: Calculate the correlation between different geological parameters and geological risks. This can be achieved through statistical methods such as Pearson correlation coefficient, Spearman rank correlation or mutual information. 3. Feature importance assessment: Use machine learning algorithms such as random forest, gradient boosting machine (GBM) or XGBoost to evaluate the contribution of each feature to the model's predictive ability. 4. Recursive feature elimination (RFE): RFE is a feature selection method that builds multiple models and removes the features with the lowest weight in each step, repeating this process until the required number of features remain. 5. Principal component analysis (PCA): PCA is a dimensionality reduction technique that can identify the main components in the data, which explain the largest part of the data variability. 6. Domain expert knowledge: Combine expertise in geology and engineering to evaluate which features are most critical in geological risk prediction. 7. Model training and cross-validation: Use the selected features to train multiple machine learning models and evaluate the performance of the models through cross-validation. 8. Iterative process of feature selection: Based on the results of model training and validation, it may be necessary to go back and re-evaluate the feature selection to optimize the feature set. 9. Statistical testing: Apply statistical tests, such as t-test or ANOVA, to test whether the differences in feature values ​​under different geological risk categories are significant. 10. Visual analysis: Use visualization tools such as scatter plots and box plots to intuitively analyze the relationship between features and geological risks. Through the above steps, we can identify those features that are statistically significantly correlated with geological risks and have an important impact on model predictions. These features will be used in subsequent model training and risk prediction to improve the accuracy and efficiency of predictions.

[0098] 5. Machine Learning Model Training:

[0099] like Figure 4 As shown, a variety of machine learning algorithms, including random forest, support vector machine, neural network and deep learning network, are used to train the model on the selected features.

[0100] Use methods such as cross-validation and grid search to tune the model's hyperparameters to obtain the best model performance.

[0101] This application uses a variety of machine learning algorithms for model training. The connections between the four models of random forest, support vector machine, neural network and deep learning network are as follows: 1. All are supervised learning algorithms: These models are trained under the framework of supervised learning, that is, they all learn how to predict the output of unknown data from labeled training data. 2. Used to predict geological risks: All these models are applied to the same goal, that is, predicting geological risks during tunnel construction. 3. Feature selection: They are all trained using feature sets obtained through the same feature selection process, which means that they consider the same geological features when predicting geological risks. 4. Model performance comparison: During the training process, the performance of these models can be compared with each other to determine which model performs best on a specific data set. 5. Cross-validation: All models may use cross-validation methods to evaluate and optimize their performance to ensure the generalization ability of the model. 6. Hyperparameter tuning: Each model has its hyperparameters, which need to be adjusted through methods such as grid search, random search or Bayesian optimization to achieve optimal performance.

[0102] Among them, the training data sets for training random forest models, support vector machine models, neural network models, and deep learning network models are as follows: 1. Data set division: From the collected geological data, the data is divided into training set, validation set, and test set. The training set is used for model training, the validation set is used for model selection and hyperparameter tuning, and the test set is used to evaluate the final performance of the model. 2. Data consistency: The training data sets of all models should be consistent, that is, based on the same geological characteristics and risk labels to ensure the fairness of model training and the comparability of results. 3. Data preprocessing: The training data used by all models should undergo the same preprocessing steps, including data cleaning, standardization, denoising, etc., to ensure data quality.

[0103] The inputs of the random forest model, support vector machine model, neural network model, and deep learning network model are all preprocessed geological feature data. These features may include but are not limited to rock type, fault distribution, groundwater level, soil moisture, etc. The output of all models (i.e., random forest model, support vector machine model, neural network model, and deep learning network model) is the prediction result of geological risk, specifically: 1. Classification task: If geological risk prediction is a classification problem, the output of the model will be the probability or label of the risk category (such as low risk, medium risk, high risk). 2. Regression task: If geological risk prediction is a regression problem, the output of the model will be a continuous value of the risk level.

[0104] Model input and output when in use: 1. Model input: When in use, the input of the model will be new geological monitoring data, which also needs to go through the same preprocessing steps as the training data to ensure consistency in data format and quality. 2. Model output: The output of the model will be the risk prediction results for these new data, which can be the probability of risk occurrence, risk level or specific category label.

[0105] Whether in the training phase or the use phase, the input of all models is consistent pre-processed geological feature data, and the output is the risk prediction results based on these input data. Different models may differ in specific implementation details, training processes, and hyperparameter tuning, but their basic tasks and goals are the same, that is, to predict geological risks during tunnel construction.

[0106] 6. Model verification and testing:

[0107] The trained model is validated on an independent test set to evaluate the model’s performance indicators such as prediction accuracy, recall, and F1 score.

[0108] In the workflow of machine learning, after completing model training, it is usually necessary to verify the model on an independent test set to evaluate the model's predictive performance: 1. Independent test set: Select a dataset that the model has not seen before as the test set. This dataset should be representative of the data that the model will process in actual applications. 2. Performance evaluation: Use the test set to evaluate the model. Common performance indicators include: (1) Accuracy: The proportion of samples correctly predicted by the model to the total number of samples. (2) Recall: Also known as the true positive rate, the proportion of samples correctly identified as positive by the model to all actual positive samples. (3) F1 score: The harmonic mean of accuracy and recall, a performance indicator that comprehensively considers accuracy and recall. (4) ROC curve and AUC: The receiver operating characteristic curve (ROC) and the area under the curve (AUC) are also commonly used indicators for evaluating the performance of binary classification models. 3. Model selection: Based on the results of performance evaluation, select the model with the best performance. The selection of the optimal model does not only rely on a single performance indicator, but may also need to consider other factors, such as model complexity, training time, prediction time, etc. 4. Generalization ability: The model that performs best on the test set is generally considered to have the best generalization ability, that is, it can make accurate predictions on new, unseen data. 5. Final model: The optimal model determined after evaluation will be used for the actual task of geological risk prediction for tunnel construction. 6. Continuous monitoring and updating: Even if the model has been deployed, its performance needs to be continuously monitored and the model needs to be regularly updated based on new data to maintain its prediction accuracy. Through this process, it can be ensured that the selected model not only performs well on the training data, but also can provide reliable geological risk predictions in practical applications.

[0109] 7. Real-time risk prediction and assessment:

[0110] Develop a real-time data processing system to process geological monitoring data collected in real time during construction.

[0111] Use the trained model to conduct risk assessment on real-time data and output the probability or level of geological risk.

[0112] 8. Risk management and decision support:

[0113] Based on the model's prediction results, risk management plans and emergency response strategies are formulated to provide scientific decision-making support for the construction team.

[0114] 9. Model self-learning and adaptation:

[0115] Implementing the model’s online learning mechanism enables the model to incrementally learn new geological data and automatically adjust model parameters to adapt to changes in geological conditions.

[0116] Implementing an online learning mechanism for a model so that it can adapt to changing geological conditions usually involves the following key steps: S1. Data stream collection: Establish a system to continuously collect new geological data. This data may come from sensors, geological exploration activities, or other data sources, and is usually real-time or near real-time. S2. Data preprocessing: Perform the same preprocessing steps as in the training phase on the newly collected data, including cleaning, standardization, denoising, etc., to ensure data quality. S3. Incremental learning: Develop or select algorithms that support online learning or incremental learning. These algorithms are able to update model parameters using new data without retraining the entire model. For example: For some machine learning models, online gradient descent can be used to continuously update model weights. For neural networks, mini-batch stochastic gradient descent (mini-batch SGD) can be used for online training. S4. Model update strategy: Determine a strategy for updating model parameters, such as setting a fixed update cycle or updating based on specific trigger conditions (such as model performance degradation or significant changes in data distribution). S5. Model evaluation: Regularly evaluate model performance to ensure that the online learning process has not caused model performance degradation. Online cross-validation can be used or a portion of the data can be reserved as a validation set. S6. Adaptive adjustment: Implement an adaptive mechanism to adjust the learning rate or other hyperparameters to adapt to changes in data and feedback on model performance. S7. Model maintenance: Monitor the long-term performance of the model and perform maintenance when necessary, such as retraining or adjusting the model structure. S8. Feedback loop: Establish a feedback mechanism to use the results of model predictions to guide the data collection process so that data useful for model improvement can be collected more efficiently. S9. Transfer learning: In some cases, transfer learning techniques can be used to apply a model trained in one geological environment to another similar geological environment to accelerate the model adaptation process. S10. Multi-task learning: If changes in geological conditions affect multiple related tasks, consider using a multi-task learning framework to simultaneously update the performance of the model on these tasks. S11. Model version control: Maintain different versions of the model so that you can roll back to the previous stable version if problems occur after the model is updated.

[0117] 10. User Interface and Visualization:

[0118] Develop a user-friendly visual interface to present risk prediction results to the construction team in the form of graphs or charts to facilitate understanding and decision-making.

[0119] 11. System Integration and Deployment:

[0120] The entire prediction system can be integrated into existing tunnel construction monitoring software or hardware to ensure the practicality and ease of use of the system.

[0121] The application of data science and machine learning technology to the geological risk prediction of tunnel construction in this application requires highly specialized knowledge to overcome the following technical challenges: 1. Data complexity and diversity: Geological data usually contains multiple types, such as rock type, structure, groundwater level, soil moisture, etc. The complexity and diversity of these data make it a technical challenge to accurately identify and predict geological risks from massive data. 2. Data preprocessing: The collected raw data needs to be cleaned, noise and outliers need to be removed, and missing data needs to be filled to improve data quality. 3. Multi-source data fusion: Integrate information from different sensors and data sources to obtain a more comprehensive and accurate representation of geological features. 4. Feature engineering: Select features that are highly relevant to geological risks, and apply feature selection techniques to reduce feature dimensions and improve the generalization ability of the model. 5. Real-time data processing and dynamic risk assessment: It is necessary to develop a real-time data processing system and a dynamic risk assessment mechanism to quickly respond to changes in geological conditions. 6. Self-learning and self-adaptation of the model: The model needs to have the ability of self-learning and self-adaptation to adapt to changes in geological conditions and ensure the continued effectiveness of the model and the high accuracy of prediction. 7. User interface and visualization: Develop a user-friendly visualization interface to enable the construction team to intuitively understand the risk prediction results. 8. System integration and deployment: Integrate the prediction system into existing tunnel construction monitoring software or hardware to ensure the practicality and ease of use of the system. In addition, the implementation of data science and machine learning technologies requires a lot of computing resources and expertise, as well as a deep understanding of existing workflows and systems.

[0122] Compared with the prior art, this application has the following technical means: 1. Comprehensive application of data science and machine learning: By comprehensively collecting geological, geochemical, geophysical and remote sensing data and applying data science and machine learning techniques for analysis, the accuracy and efficiency of prediction are improved. 2. Advanced data preprocessing and feature selection technology: Bessel filter is used to remove noise, k-nearest neighbor algorithm is used to fill missing values, and z-score method is used for data standardization to improve data quality. 3. Self-learning and adaptive machine learning model: The online learning mechanism of the model is implemented, so that the model can incrementally learn new geological data, automatically adjust model parameters, and adapt to changes in geological conditions. 4. Real-time risk assessment and dynamic threshold determination: A real-time data stream processing system and a dynamic threshold determination mechanism are developed to quickly respond to changes in geological conditions. 5. Risk management and decision support: A risk assessment report generator and a decision support system are provided to provide scientific decision-making suggestions for the construction team. 6. User interface and visualization: A user-friendly visualization interface is developed to enable the construction team to intuitively understand the risk prediction results. 7. System integration and deployment: The entire prediction system is integrated into the existing tunnel construction monitoring software or hardware to ensure the practicality and ease of use of the system. Through these technical means, the present invention solves the shortcomings of the existing technology in terms of accuracy, real-time, adaptability and user interactivity, and provides a more accurate, efficient, real-time responsive and user-friendly geological risk prediction solution.

[0123] The more accurate and efficient geological risk prediction model provided by this application is based on the following technical means: 1. Multi-source geological data fusion: By integrating heterogeneous data from multiple sources such as geological exploration equipment, geochemical analysis instruments, geophysical exploration instruments and remote sensing technology, the comprehensiveness and accuracy of the data are improved. 2. Data preprocessing: Use time series analysis methods such as Bessel filters and spatial filtering techniques such as Gaussian filters to remove noise and outliers, fill in missing data, and improve data quality. 3. Feature engineering: Apply techniques such as recursive feature elimination and principal component analysis to select features that are highly related to geological risks, reduce feature dimensions, and improve the generalization ability of the model. 4. Machine learning model training: Use advanced algorithms such as random forests, support vector machines, neural networks, and deep learning networks to train the selected features, and use methods such as cross-validation and grid search to tune the hyperparameters of the model. 5. Model validation and testing: Validate on an independent test set to evaluate the model's performance indicators such as prediction accuracy, recall rate, and F1 score to ensure the generalization ability of the model. 6. Self-learning and self-adaptation: Implement the model's online learning mechanism so that the model can incrementally learn new geological data and automatically adjust model parameters to adapt to changes in geological conditions. Its efficiency is based on the following technical means: (1) Automated data collection: Through the automated data collection system, the need for manual operation is reduced and the speed and efficiency of data processing are improved. (2) Real-time data processing: Develop a real-time data processing system to quickly respond to changes in geological conditions and provide timely risk prediction information to the construction team. (3) Model optimization: Improve the efficiency of model training and prediction performance through model hyperparameter optimization techniques such as grid search or Bayesian optimization algorithm. (4) Risk prediction module: Design a real-time risk prediction module to update risk assessment at a high frequency to ensure rapid response. (5) Risk management and decision support: Automatically generate risk management plans and emergency response strategies to improve the risk response efficiency of the construction team. (6) User interface and visualization: Provide an intuitive and interactive risk prediction and management suggestion display platform to facilitate the construction team to quickly understand and operate. (7) System integration and deployment: Integrate the prediction system into existing tunnel construction monitoring software or hardware to ensure the practicality and ease of use of the system and reduce the time and resource consumption during the integration process. These technical means work together to make the geological risk prediction model not only more accurate in predicting geological risks, but also improve the efficiency of the entire prediction process.

[0124] This application solves several technical problems existing in traditional geological risk prediction methods. First, it overcomes the limitations of relying solely on the experience of geological experts and qualitative analysis, and improves the accuracy and objectivity of risk prediction through automated data collection and the application of machine learning models. Secondly, this application solves the problem of how to effectively identify and extract key geological risk features from massive and complex geological data, and ensures that the model can focus on the most predictive data through advanced data preprocessing and feature engineering. In addition, this application also responds to the challenges of real-time geological risk assessment, and achieves rapid response to rapidly changing geological conditions through real-time data processing systems and dynamic risk assessment mechanisms. Finally, this application solves the problems of model generalization ability and continuous adaptability through the self-learning and adaptive capabilities of the model, so that the model can be continuously optimized and adapted to different geological environments and construction conditions. Through these technical innovations, this application significantly improves the scientificity and practicality of geological risk prediction for tunnel construction.

[0125] The key technical points of this application include: 1. Multi-source geological data fusion and feature extraction technology: One of the core points of this application is its advanced data fusion technology, which can integrate heterogeneous data from multiple sources such as geological exploration equipment, geochemical analysis instruments, geophysical exploration instruments and remote sensing technology. In addition, feature extraction techniques including recursive feature elimination and principal component analysis can identify and extract the most critical features for geological risk prediction from the fused data. These features not only improve the prediction accuracy of the machine learning model, but also enhance the generalization ability of the model, enabling it to adapt to changing geological conditions. 2. Self-learning and adaptive machine learning model: Another key point is the self-learning and adaptive mechanism of the model. The machine learning model of this application can not only learn through traditional training sets, but also learn from new geological data in real time through online learning mechanisms, thereby continuously optimizing and adjusting model parameters. This mechanism significantly improves the response speed and adaptability of the model to changes in the geological environment, ensuring the continued effectiveness of the model during tunnel construction and the high accuracy of prediction. These two points are the innovative core of this application. Together, they constitute an efficient, accurate prediction system that can respond to changes in geological risks in real time, which is of great significance for improving the safety and efficiency of tunnel construction.

[0126] The geological risk prediction model based on data science and machine learning provided by this application has significant beneficial effects. First, the model significantly improves the accuracy and reliability of geological risk prediction for tunnel construction by comprehensively analyzing and processing multi-source geological data. Secondly, the advanced feature extraction technology and machine learning algorithm adopted by the model make the risk prediction process more refined and can identify potential risks that are difficult to find by traditional methods. In addition, the self-learning and adaptive capabilities of the model ensure its continued effectiveness in the face of complex and changeable geological conditions, providing a dynamic risk management tool for the construction team. At the same time, the real-time risk assessment function of the model provides the possibility for rapid decision-making and emergency response, thereby effectively reducing construction risks and potential costs. Finally, through a user-friendly visual interface, the construction team can intuitively understand the risk prediction results, improve the efficiency of decision-making and the transparency of the construction process. In general, the technical solution of this application provides an innovative, efficient and practical solution for geological risk management in tunnel construction.

[0127] like Figure 5 As shown, the present application also provides a geological risk prediction system in tunnel construction.

[0128] The geological risk prediction system for tunnel construction provided by this application includes:

[0129] The geological data collection module is used to obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during the tunnel construction process.

[0130] The data preprocessing module is used to preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result using data preprocessing technology to obtain preprocessed geological data samples corresponding to different geological risk prediction results.

[0131] The feature selection module is used to select features of the preprocessed geological data samples corresponding to each geological risk prediction result using feature engineering technology to obtain key geological feature samples corresponding to different geological risk prediction results.

[0132] The machine learning model training module is used to train multiple machine learning models using key geological feature samples corresponding to different geological risk prediction results to obtain multiple trained machine learning models.

[0133] The model verification and testing module is used to verify and test a variety of trained machine learning models. According to the results of model verification and testing, the model with the best performance among the multiple trained machine learning models is used as the geological risk prediction model.

[0134] The real-time data acquisition module is used to acquire the geological data of the tunnel construction area to be predicted in real time.

[0135] The real-time data preprocessing module is used to preprocess the geological data of the predicted tunnel construction area using data preprocessing technology to obtain preprocessed geological data.

[0136] The real-time feature selection module is used to select features from the preprocessed geological data using feature engineering techniques to obtain key geological features.

[0137] The real-time risk prediction module is used to input key geological features into the geological risk prediction model, use the geological risk prediction model to predict the geological risks in the tunnel construction process in real time, and output the geological risk prediction results corresponding to the key features.

[0138] The following is a specific example to illustrate the technical solution of the geological risk prediction system in tunnel construction of this application:

[0139] like Figure 5 As shown, the geological risk prediction system for tunnel construction provided by the present application includes: a geological data collection module 10, a data preprocessing module 20, a feature selection module 30, a machine learning model training module 40, a model verification and testing module 50, a real-time risk prediction module 60, a risk management and decision support module 70, a user interface and visualization module 80, and a system integration and deployment module 90. The following is a detailed introduction to each module:

[0140] 1. Geological Data Collection (Module 10)

[0141] During the geological data collection phase, this application will deploy a series of high-precision geological exploration equipment, including geological radar and seismic wave tomography equipment, to obtain underground structure information in the tunnel construction area. At the same time, geochemical analysis instruments are used to analyze the composition of soil and groundwater samples, and geophysical exploration instruments are used to measure geophysical parameters such as seismic wave velocity. In addition, high-resolution image data of the surface and underground are collected through satellite remote sensing or remote sensing equipment carried by drones. Parameters: The acquisition frequency of geological exploration equipment is set to once every 2 hours to ensure the real-time and continuity of data. The resolution of remote sensing image data is at least 0.5 meters / pixel to meet the needs of high-precision analysis. Results: The collected raw data set contains a variety of geological parameters, which provides a basis for subsequent data preprocessing and analysis.

[0142] 2. Data preprocessing (Module 20)

[0143] The data preprocessing module is responsible for converting the raw geological data into a format suitable for analysis. First, data cleaning is performed to remove duplicate records and inconsistencies, and then outliers are identified and processed through box plot analysis and Z-score method. For missing values, the k-nearest neighbor (k-NN) algorithm is used for filling, where the k value is set to 5 to ensure the accuracy of filling. Finally, the Z-score standardization method is applied to standardize all features to eliminate the influence of different dimensions and distribution characteristics. Parameters: Data cleaning includes removing duplicate records and correcting inconsistencies. The IQR method is used for outlier processing, and the Z-score threshold is set to 3. In the k-NN algorithm, the k value is fixed to 5 to obtain a stable filling effect. Results: The format of the preprocessed data set is unified, and the data quality is significantly improved, laying a solid foundation for feature selection and model training.

[0144] 3. Feature Selection (Module 30)

[0145] In the feature selection stage, the present application first performs feature statistical analysis on the preprocessed data, including calculating the mean, variance, quantile and other statistics of the features. Then, by calculating the correlation between the features and the target variable, such as the Pearson correlation coefficient, the features that are highly correlated with geological risk are identified. Next, recursive feature elimination and principal component analysis techniques are applied to select the most representative features from the data (i.e., the preprocessed geological data set, which includes a variety of features related to geological risk). RFE is based on the random forest model for feature importance assessment. The PCA goal is to retain 95% of the variability in the data set (i.e., the preprocessed geological data set, which contains various features used for geological risk prediction analysis, such as rock type, fault distribution, groundwater level, soil moisture, etc.). Parameters: The threshold of the correlation analysis is set to 0.6 to screen out features with strong correlation with the target variable. The random forest model in RFE is set to 100 trees, and the number of components retained by PCA is determined based on the cumulative contribution rate. Results: Finally, a streamlined set of features is obtained, which has an important impact on geological risk prediction while reducing the complexity of model training.

[0146] The target variable specifically refers to:

[0147] Risk categories: Geological risks may be classified into several categories such as "low risk", "medium risk" and "high risk".

[0148] Risk Level: A continuous risk score or rating that indicates the severity of the risk.

[0149] Probability of an event occurring: indicates the possibility of a certain geological problem occurring, such as landslide, collapse, etc.

[0150] Quantitative indicators: certain geological risk indicators that can be directly measured, such as the change in groundwater level, the range of change in soil moisture, etc.

[0151] The implementation process of RFE for feature importance evaluation based on the random forest model is as follows:

[0152] Random forests work by building multiple decision trees and aggregating their predictions, with each tree randomly selecting a subset of features to split during training. During this process, the average impurity reduction for each feature across all trees can be calculated, which reflects the role of the feature in reducing node impurity. The more impurity reduction a feature causes, the higher its importance. In addition, permutation feature importance methods can be used, which assesses the importance of each feature by randomly permuting its value and observing the change in model accuracy. If the model accuracy drops significantly, the feature is considered important. Through these methods, the random forest model is able to provide an importance score for each feature, helping to identify the features that have the greatest impact on predicting geological risks.

[0153] The number of components retained by PCA is determined based on the cumulative contribution rate. The implementation process is as follows:

[0154] Calculate the contribution of each principal component: PCA first calculates the variance of each principal component and then compares it to the sum of all principal component variances to get the contribution of each component. This is usually expressed as a percentage.

[0155] Cumulative contribution: Starting from the first principal component, the contribution of each principal component is added up until the desired retention threshold is reached. For example, if the goal is to retain 95% of the variability, then the contribution of the principal components is added up until the sum reaches or exceeds 95%.

[0156] Determine the number of components to retain: When the cumulative contribution rate exceeds or equals 95% for the first time, the last principal component to be included is the number of components that should be retained. This number ensures that most of the variability in the data set is retained.

[0157] Interpret the results: The selected components will be used to form a new, reduced-dimensionality representation of the data that retains most of the variability in the original dataset while removing redundancy and noise.

[0158] For example: If the first principal component explains 40% of the variability, the second principal component explains 30%, and the third principal component explains 20%, then these three components together explain 90% of the variability. In this case, if the goal is to retain 95% of the variability, you may need to include more principal components until the cumulative contribution reaches or exceeds 95%.

[0159] 4. Machine Learning Model Training (Module 40)

[0160] In the machine learning model training stage, this application selected four machine learning algorithms, random forest, support vector machine (SVM), neural network and deep learning network, for model training. The random forest model was initialized with 100 trees to enhance the generalization ability of the model. SVM uses radial basis function (RBF) as the kernel function, and its regularization parameter C and kernel function parameter γ are optimized by grid search. The neural network model is designed to contain 2 hidden layers, each with 50 neurons, and is trained using ReLU activation function and Adam optimizer. All models are trained using training set data (i.e., the random forest model, support vector machine model, neural network model and deep learning network model are trained using the same training data set), and the model parameters are optimized by cross-validation. Parameters: The number of trees in the random forest model is 100, the kernel function of the SVM is RBF, the number of hidden layers of the neural network is 2, and the number of neurons in each layer is 50. The model training uses the cross-validation method, and the number of cross-validation folds is 5. Results: The multiple models obtained by training can capture the relationship between geological characteristics and risks, providing candidate models for the next step of model validation.

[0161] 5. Model Validation and Testing (Module 50)

[0162] During the model validation and testing phase, this application will use the validation set data to evaluate the performance of the trained model. The 5-fold cross-validation method is used to evaluate the model's accuracy, recall, F1 score, and area under the ROC curve (AUC). By comparing the performance indicators of different models, the model with the best performance on the validation set is selected as the final model. In addition, the final selected model will be evaluated on an independent test set to verify its generalization ability. Parameters: The number of cross-validation folds is set to 5, and the performance evaluation indicators include accuracy, recall, F1 score, and AUC. The test set is independent of the training set / validation set to simulate real-world prediction scenarios. Results: After rigorous verification and testing, this embodiment obtained a random forest model with an accuracy of 92% on the validation set. The model also showed good generalization ability on the test set and was selected as the final model. When actually predicting geological risks, the final model obtained after model verification and testing may be a random forest model (for example, the final model obtained in this embodiment is a random forest model), or it may be a support vector machine model, a neural network model, or a deep learning network model. Which model is finally selected as the optimal model depends on multiple factors: performance indicators, generalization ability, computational efficiency, model complexity, domain knowledge, and data characteristics.

[0163] 6. Real-time risk prediction (Module 60)

[0164] The final selected random forest model is deployed in the real-time risk prediction module. This module receives real-time geological monitoring data from the sensor network every minute, including groundwater level changes, soil stress, etc. The trained model is used to conduct risk assessment on the real-time data and output the risk prediction results. In addition, the module also contains a dynamic threshold determination mechanism, which dynamically adjusts the threshold of risk assessment according to the historical data and statistical characteristics of geological risks to adapt to changes in the construction environment. Parameters: The real-time data is updated once a minute, and the risk assessment threshold is dynamically adjusted according to historical data. The real-time data input to the model needs to go through the same preprocessing steps as the training data to ensure the consistency of the data format. Results: The real-time risk prediction module can quickly respond to changes in geological conditions and provide timely risk prediction information for the construction team.

[0165] The process of dynamically adjusting the threshold of risk assessment based on the historical data and statistical characteristics of geological risks is as follows: (1) First, analyze the historical geological risk data and identify the characteristics of risk distribution, such as mean, median, standard deviation, quantile, etc. (2) Use statistical models to estimate the risk distribution, which may include normal distribution, Poisson distribution or other probability distributions, and fit the model parameters based on historical data. (3) Collect geological monitoring data in real time, such as groundwater level, soil stress, etc., and compare them with the statistical characteristics of historical data. (4) Develop algorithms to dynamically adjust the threshold of risk assessment based on real-time data streams and statistical models. For example, if the monitored data points continue to exceed a certain standard deviation range, the algorithm may increase the risk level. (5) Use machine learning models to predict the trend of risk changes and adjust the threshold based on the model's prediction results. (6) Establish a feedback mechanism to continuously optimize the threshold adjustment strategy based on risk assessment results and actual geological events. (7) Implement automated tools so that when specific statistical changes or risk patterns are detected, the system automatically adjusts the threshold of risk assessment. (8) The connection between dynamically adjusting the threshold of risk assessment and model self-learning and adaptation is as follows: Adaptability: Self-learning and adaptive models can automatically adjust their parameters, including the threshold of risk assessment, based on new geological data to adapt to changes in geological conditions. Continuous learning: The online learning mechanism allows the model to continuously learn from new data, which helps the model capture the latest changes in geological risk patterns and adjust the threshold more accurately. Performance monitoring: The model evaluation framework regularly evaluates model performance, including the accuracy of risk assessment, which can guide the adjustment of thresholds. Transfer learning: Transfer learning technology can apply knowledge learned in one geological environment to another environment, which helps to quickly adjust the model to new geological conditions, including the setting of thresholds. Model updating: The adaptive model can be updated based on the results of performance evaluation and new geological data, which may involve resetting or adjusting the threshold of risk assessment.

[0166] 7. Risk Management and Decision Support (Module 70)

[0167] The risk management and decision support module automatically generates risk management plans and emergency response strategies based on the risk assessment results provided by the real-time risk prediction module. The module contains a risk assessment report generator that can generate detailed risk management reports based on the risk level classification criteria. In addition, the module also provides a decision support system to provide scientific decision-making suggestions for the construction team, such as preventive measures and emergency evacuation routes. Parameters: Risk level classification is based on the risk probability output by the model, and the management plan includes preventive measures, emergency response strategies, and evacuation routes. The parameters of the report generator and decision support system can be adjusted according to the specific construction environment. Results: The construction team obtained specific management measures and suggestions for different risk levels, which improved the efficiency and effectiveness of risk response.

[0168] 8. User Interface and Visualization (Module 80)

[0169] The user interface and visualization module provides users with an intuitive and interactive risk prediction and management recommendation display platform. The module includes a web-based user interface that allows users to intuitively view risk prediction results and management recommendations through graphics or charts. The interface design includes risk maps, probability distribution diagrams, and other visualization tools, such as risk heat maps and time series analysis diagrams. In addition, the module provides an interactive risk assessment report viewer that allows users to adjust display parameters such as risk level thresholds and time ranges as needed. Parameters: The user interface is designed with user experience in mind, including intuitive layout, clear instructions, and real-time feedback. The selection and design of visualization tools are based on the characteristics of geological risk prediction and the specific needs of users. Results: The user interface and visualization module greatly improved the construction team's understanding and response speed to risk prediction results, and improved the efficiency of decision-making.

[0170] 9. System Integration and Deployment (Module 90)

[0171] The system integration and deployment module is responsible for integrating the entire geological risk prediction system into the existing tunnel construction monitoring software and hardware. The module first performs a compatibility analysis on the existing monitoring software to ensure that the new system can be seamlessly integrated. Then, according to the specific conditions of the construction site, the necessary hardware equipment, such as sensors, data collectors, servers, and communication equipment, is deployed. In addition, the module provides a system configuration tool that allows users to adjust the parameter settings of the system according to specific needs. Parameters: The system integration takes into account the API interface, data exchange format, and compatibility of the hardware equipment of the existing monitoring software. The configuration of the hardware equipment is selected and adjusted according to the specific needs and environmental conditions of the construction site. Results: The system integration and deployment module ensures the compatibility of the geological risk prediction system with the existing monitoring system and improves the practicality and ease of use of the system.

[0172] Through this series of detailed and orderly implementation steps, this application achieves the following goals: Real-time and accurate collection and processing of geological data to provide high-quality risk prediction. Rapid response to changes in geological risks to provide scientific decision-making support for the construction team. The user interface is intuitive and highly interactive, making it easy for the construction team to understand and operate. The system seamlessly integrates with existing monitoring systems to improve construction safety and efficiency.

[0173] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store geological risk prediction data in tunnel construction. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting geological risks in tunnel construction is implemented.

[0174] Those skilled in the art will understand that Figure 6The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0175] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor. Specifically, a computer program for implementing the above-mentioned geological risk prediction model based on data science and machine learning is stored on the computer-readable storage medium, and the computer program includes: a data collection module for automatically collecting and storing geological data, specifically including interfaces and communication protocols with various geological exploration equipment. A data preprocessing module for automatically performing data cleaning and standardization, specifically including the implementation of data cleaning algorithms and data conversion algorithms. A feature selection module for automatically performing feature selection algorithms, specifically including the implementation of various feature selection algorithms and a feature importance evaluation mechanism. A model training module for automatically performing machine learning algorithm training, specifically including the implementation of multiple machine learning algorithms and model hyperparameter optimization technology. A risk prediction module for automatically performing geological risk assessment and output, specifically including a model reasoning process and risk level classification criteria.

[0176] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0177] In an exemplary embodiment, an electronic device is provided, including a processor and a memory, a computer program is stored on the memory, the computer program includes an instruction set of the above-mentioned geological risk prediction model based on data science and machine learning, and is used to execute on the processor, and the electronic device includes: an input device, such as a keyboard, a touch screen or a sensor interface, for receiving geological data input. An output device, such as a display, a printer or an alarm system, is used to display risk prediction results. A communication interface, such as a wireless network module or a wired network interface, is used to exchange data with a construction monitoring system. A power supply unit, such as a battery or an external power adapter, is used to provide stable power support for the electronic device. A memory, such as a hard disk drive or a solid-state drive, is used to store a large amount of geological data and model parameters. A processor, such as a central processing unit (CPU) or a graphics processing unit (GPU), is used to execute complex data science and machine learning algorithms.

[0178] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0179] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0180] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0181] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0182] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, 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 this application.

Claims

1. A method for predicting geological risks in tunnel construction. It is characterized in that The method for predicting geological risks in tunnel construction includes: Obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during tunnel construction; The data preprocessing technology is used to preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result, so as to obtain the preprocessed geological data samples corresponding to different geological risk prediction results; Feature engineering technology is used to select features of the preprocessed geological data samples corresponding to each geological risk prediction result, and key geological feature samples corresponding to different geological risk prediction results are obtained; Using key geological feature samples corresponding to different geological risk prediction results to train multiple machine learning models, multiple trained machine learning models are obtained; Conduct model verification and testing on multiple trained machine learning models. Based on the results of model verification and testing, use the model with the best performance among multiple trained machine learning models as the geological risk prediction model. Real-time acquisition of geological data of the tunnel construction area to be predicted; The data preprocessing technology is used to preprocess the geological data of the tunnel construction area to be predicted to obtain preprocessed geological data; Using the feature engineering technology to perform feature selection on the preprocessed geological data to obtain key geological features; The key geological features are input into the geological risk prediction model, the geological risk prediction model is used to predict the geological risks in the tunnel construction process in real time, and the geological risk prediction results corresponding to the key features are output.

2. The method for predicting geological risks in tunnel construction according to claim 1, It is characterized in that After inputting the key geological features into the geological risk prediction model, using the geological risk prediction model to predict the geological risks in the tunnel construction process in real time, and outputting the geological risk prediction results corresponding to the key features, the method for predicting geological risks in tunnel construction further includes: Model adaptation technology is applied to the geological risk prediction model to enable the geological risk prediction model to have self-learning and self-adaptation capabilities.

3. The method for predicting geological risks in tunnel construction according to claim 1, It is characterized in that Before the feature engineering technology is used to perform feature selection on the preprocessed geological data samples corresponding to each geological risk prediction result to obtain key geological feature samples corresponding to different geological risk prediction results, the method for geological risk prediction in tunnel construction also includes: For the preprocessed geological data samples corresponding to each geological risk prediction result, data fusion technology is used to perform multi-source data fusion on the preprocessed geological data samples to obtain fused data; the fused data is used for feature selection using the feature engineering technology to obtain key geological feature samples.

4. The method for predicting geological risks in tunnel construction according to claim 1, It is characterized in that The data preprocessing technology includes: Using time series analysis method and spatial filtering technology to remove noise; the time series analysis method is used to remove periodic noise of time series data; the spatial filtering technology is used to reduce noise caused by spatial autocorrelation of spatial data; the time series analysis method includes Bessel filter, Fourier transform, wavelet transform; the spatial filtering technology includes Gaussian filter; An outlier processing method is used to remove outliers; the outlier processing method includes a box plot method, an IQR method, and a Z-score method; Fill missing values ​​using the k-nearest neighbor algorithm; The z-score method was used for data normalization.

5. The method for predicting geological risks in tunnel construction according to claim 1, It is characterized in that The feature engineering techniques include: Conduct characteristic statistical analysis through statistical analysis methods; By calculating the correlation between features and target variables, features with high correlation to geological risks are identified; A feature selection technique is applied to select key geological features from features that are highly correlated with geological risks; the feature selection technique includes recursive feature elimination and principal component analysis; the recursive feature elimination is based on a random forest model to perform feature importance assessment.

6. The method for predicting geological risks in tunnel construction according to claim 1, It is characterized in that Various machine learning models include random forest models, support vector machine models, neural network models, and deep learning network models.

7. A geological risk prediction system for tunnel construction, It is characterized in that The geological risk prediction system in tunnel construction includes: A geological data collection module, used to obtain geological data samples of the tunnel construction area corresponding to different geological risk prediction results during the tunnel construction process; A data preprocessing module is used to preprocess the geological data samples of the tunnel construction area corresponding to each geological risk prediction result by using data preprocessing technology to obtain preprocessed geological data samples corresponding to different geological risk prediction results; A feature selection module is used to select features of the preprocessed geological data samples corresponding to each geological risk prediction result by using feature engineering technology to obtain key geological feature samples corresponding to different geological risk prediction results; A machine learning model training module is used to train multiple machine learning models using key geological feature samples corresponding to different geological risk prediction results to obtain multiple trained machine learning models; Model verification and testing module, used to verify and test a variety of trained machine learning models. According to the results of model verification and testing, the model with the best performance among the multiple trained machine learning models is used as the geological risk prediction model; A real-time data acquisition module is used to acquire the geological data of the tunnel construction area to be predicted in real time; A real-time data preprocessing module, used to preprocess the geological data of the tunnel construction area to be predicted by using the data preprocessing technology to obtain preprocessed geological data; A real-time feature selection module, used to select features of the pre-processed geological data using the feature engineering technology to obtain key geological features; The real-time risk prediction module is used to input the key geological features into the geological risk prediction model, use the geological risk prediction model to predict the geological risks in the tunnel construction process in real time, and output the geological risk prediction results corresponding to the key features.

8. A computer device, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting geological risks in tunnel construction as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method for predicting geological risks in tunnel construction described in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the method for predicting geological risks in tunnel construction described in any one of claims 1 to 6 is implemented.

Citation Information

Cited By

  • Advanced prediction method and system for unfavorable geology of tunnel based on data fusion theory

    CN119537870A

  • Decision tree determination method, subsea tunnel surrounding rock impermeability grading method and product

    CN119862486A

  • Intelligent aging data analysis method and system for laser

    CN121327755A