Tunnel adverse geological identification method and system based on Bayesian optimization random forest

Abstract

The present invention belongs to the field of geological body identification, and provides a method and system for identifying bad geological conditions in tunnels based on Bayesian optimization random forests, which obtains geological data obtained by different detection methods and preprocesses the geological data; analyzes and screens the preprocessed geological data, extracts features related to bad geological conditions; constructs a random forest model, and preliminarily classifies bad geological bodies; uses a Bayesian optimization algorithm to optimize the hyperparameters of the random forest model, uses the Bayesian model averaging method to assign weights to the decision trees in the model, and trains the final random forest model; uses the final random forest model to predict the probability of the bad geological body type, and predicts the location and scale of the bad geological body. The present invention automatically adjusts the hyperparameters of the random forest model through Bayesian optimization, improves the real-time and accuracy of identification, and reduces the risks and economic losses in construction.

Description

Technical Field

[0001] The present invention belongs to the field of geological body identification, and specifically relates to a method and system for identifying poor tunnel geology based on Bayesian optimization random forest. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] During the construction process, geological conditions have a vital impact on the safety and stability of the project. Especially in tunnel engineering projects, the complexity and uncertainty of geological conditions may cause a series of engineering risks, such as foundation settlement, landslides, debris flows, fault activity and other adverse geological phenomena. Accurate classification and identification of adverse geology is a key step in engineering planning, design, construction and maintenance. The main task of adverse geology classification is to identify and distinguish different types of adverse geological phenomena, such as landslides, fault fracture zones, debris flows, collapses, etc. These classification information has important reference value for geological surveys, design optimization, construction monitoring and safety warnings of engineering construction. Therefore, adverse geology classification is an important means to ensure engineering safety and reduce economic losses.

[0004] At present, the commonly used methods for identifying bad geological bodies in tunnel engineering mainly include geological exploration, geological mapping and engineers' field experience. These methods can provide geological information to a certain extent, but they have obvious limitations. For example, when geological engineers identify geological bodies, they rely on personal experience and judgment, which is easily affected by subjective factors, resulting in poor accuracy and consistency of identification results. Traditional geological exploration methods are usually carried out before tunnel excavation. The exploration data can only reflect static geological conditions and it is difficult to cope with dynamic geological changes encountered during excavation. At the same time, data acquisition and processing take a long time and it is difficult to meet the rapidly changing construction needs. In modern tunnel excavation, with the advancement of detection technology, sensors and measuring equipment have generated a large amount of high-dimensional geological data. Traditional methods are difficult to effectively process and analyze such a large amount of data, and cannot make full use of existing geological information for accurate identification and prediction.

[0005] When faced with complex and ever-changing geological conditions, traditional methods are difficult to make accurate judgments. In particular, for highly concealed geological bodies (such as deep faults, groundwater systems, etc.), the recognition accuracy of traditional methods is low, which can easily lead to increased construction risks.

[0006] In order to overcome the limitations of traditional methods, machine learning technology has been gradually introduced into the field of geological body identification in recent years. By learning from historical geological data, machine learning algorithms can extract useful features from a large amount of data and build prediction models, thereby realizing the automatic classification and identification of unknown geological bodies. Among them, the random forest algorithm has become a widely used algorithm due to its ensemble learning characteristics and good processing ability for high-dimensional data. However, although the random forest algorithm has shown good results in the identification of poor geological bodies, its performance still depends on the reasonable setting of hyperparameters (such as the number of decision trees, maximum depth, etc.). Traditional hyperparameter tuning methods, such as grid search and random search, cannot efficiently find the optimal parameter combination, especially when processing large-scale data, the calculation is large and the efficiency is low. In this case, it is particularly necessary to further optimize the performance of the random forest algorithm. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a tunnel bad geology identification method and system based on Bayesian optimization random forest. The present invention automatically adjusts the hyperparameters of the random forest model through Bayesian optimization, so that the model has higher prediction accuracy and robustness in the geological body classification task. The present invention can predict the spatial range, depth and scale of bad geological bodies through the analysis of geological data, and output the occurrence probability of bad geological bodies at the same time, and quantify geological risks; it not only improves the real-time and accuracy of identification, but also reduces the risks and economic losses in construction.

[0008] According to some embodiments, the present invention adopts the following technical solutions:

[0009] A method for identifying poor tunnel geology based on Bayesian optimization random forest, comprising the following steps:

[0010] Obtain geological data obtained by different detection methods and pre-process the geological data;

[0011] Analyze and screen the preprocessed geological data to extract features related to adverse geology;

[0012] Construct a random forest model to preliminarily classify adverse geological bodies;

[0013] Use the Bayesian optimization algorithm to optimize the hyperparameters of the random forest model, use the Bayesian model averaging method to assign weights to the decision trees in the model, and train the final random forest model;

[0014] The final random forest model is used to predict the probability of the type of adverse geological body, and to predict the location and scale of the adverse geological body.

[0015] As an optional embodiment, the geological data include stratigraphic lithology, geological structure, joint spacing, porosity and groundwater level data obtained using geological analysis, geological drilling, seismic method, induced polarization method and geological radar method, as well as historical geological disaster data and environmental data.

[0016] As an optional implementation, the process of preprocessing geological data includes processing missing values ​​and outliers, using interpolation to fill missing data, and removing unreasonable data points; standardizing numerical features; and encoding categorical features into numerical features.

[0017] As an optional implementation method, the process of analyzing and screening the pre-processed geological data includes: using variance analysis, chi-square test and mutual information method to preliminarily screen the features in the geological data, selecting the features most relevant to the classification of bad geological bodies, and applying recursive feature elimination method to extract important features, wherein the important features include topographic and geomorphic features, rock and soil features, and hydrological and climatic features;

[0018] Among them, topographic and geomorphological characteristics include slope, aspect and elevation information; geotechnical characteristics include rock type, soil moisture content, porosity and density; hydrological and climate characteristics include groundwater level, precipitation and temperature.

[0019] As an optional implementation, the process of building a random forest model includes: implementing the classification task by building multiple decision trees, using a sampling method with replacement from the geological data set, randomly extracting multiple subsets, each subset is used to train a decision tree, and at each split of each decision tree, randomly selecting some features for node splitting to increase the diversity of the model, each decision tree is classified independently, and finally the prediction results of each model are combined through the Bayesian model averaging method, and different weights are given to different models.

[0020] As an optional implementation, the process of optimizing the hyperparameters of the random forest model using the Bayesian optimization algorithm includes:

[0021] Define the hyperparameters that need to be optimized;

[0022] Use Gaussian process to fit the surrogate model and construct the objective function expression;

[0023] Using the confidence upper bound as the acquisition function, each iteration updates the proxy model according to the results of the current hyperparameter combination, and finally finds the optimal hyperparameters.

[0024] As an optional implementation, the process of assigning weights to decision trees in the model using the Bayesian model averaging method includes:

[0025] Calculate the posterior probability of each decision tree according to Bayes' theorem;

[0026] Use the predicted probability and posterior probability of each decision tree to perform a weighted average to calculate the final classification probability;

[0027] Compare the weighted prediction probabilities of each category and select the category with the highest probability as the final classification result.

[0028] As an optional implementation, the process of training the final random forest model includes training the final random forest model using optimized hyperparameters, and using a portion of the preprocessed geological data as a test set to evaluate the performance of the trained model. The evaluation indicators include classification accuracy, recall rate and F1 score. The prediction effect of the model is analyzed by the ROC curve, and the area under the curve is calculated to evaluate the classification performance of the model.

[0029] As an optional implementation method, the process of predicting the location and scale of unfavorable geological bodies includes: using the trained Bayesian optimized random forest model to predict the spatial distribution data obtained by the preprocessed seismic method, induced polarization method and geological radar method, using the model's feature weights and prediction results to determine the specific location of the unfavorable geological body, analyzing the spatial overlap of low wave velocity areas, high polarization rate areas and strong geological radar reflection signal areas, and locating the specific coordinates of the unfavorable geological body or the precise position from the starting point of the tunnel.

[0030] A tunnel bad geological identification system based on Bayesian optimization random forest, comprising:

[0031] A data acquisition module is configured to acquire geological data obtained by different detection methods and pre-process the geological data;

[0032] A data preprocessing module is configured to analyze and screen the preprocessed geological data to extract features related to adverse geology;

[0033] The model training and optimization module is configured to build a random forest model to perform preliminary classification of adverse geological bodies; use the Bayesian optimization algorithm to optimize the hyperparameters of the random forest model, use the Bayesian model averaging method to assign weights to the decision trees in the model, and train the final random forest model;

[0034] The prediction and analysis module is configured to use the final random forest model to predict the probability of the bad geological body type and predict the location and scale of the bad geological body.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] (1) The present invention uses Bayesian optimization to effectively improve the classification accuracy of the random forest algorithm, so that the model shows higher accuracy and stability in the task of identifying adverse geological bodies. The hyperparameters of the random forest are optimized using the Bayesian optimization method to ensure that the model parameters can reach the global optimum. Compared with traditional grid search and random search, it can more efficiently find the best parameter combination and improve model performance.

[0037] (2) The random forest algorithm based on the Bayesian improvement can quickly process and analyze large-scale geological data, saving a lot of time and resources. Through reasonable feature extraction and encoding, it can make full use of the potential information in geological data, improve data utilization efficiency, and enhance the prediction ability of the model.

[0038] (3) The decision tree results use the BMA method to consider the performance of each tree (i.e., the posterior probability) and assign higher weights to better trees, thereby making the final prediction results more accurate and reasonable, reducing the noise introduced in the decision tree prediction results due to differences in tree quality, and making a more accurate prediction of the geological body in front of the tunnel.

[0039] (4) The multi-dimensional geological features extracted by the present invention in various ways enable the model to adapt to various complex geological environments, provide more comprehensive and accurate classification results, and be applicable to various engineering construction scenarios. The automated and intelligent method for identifying adverse geological bodies reduces the need for manual intervention, reduces the errors and subjectivity of manual operations, improves the objectivity and consistency of classification results, provides early warnings and decision-making proposals, and reduces the risk of geological disasters during construction.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0042] Figure 1 It is a flow chart of a method for identifying bad tunnel geology based on Bayesian improved random forest and Bayesian optimized random forest according to an embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of a tunnel poor geology identification system based on Bayesian improved random forest and Bayesian optimized random forest according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0046] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0047] In the absence of conflict, the embodiments in this application and the features in the embodiments may be combined with each other.

[0048] Embodiment 1

[0049] according to Figure 1 This embodiment provides a method for identifying poor tunnel geology based on Bayesian improved random forest and Bayesian optimized random forest, which includes:

[0050] In a large-scale tunnel construction project, the construction area is located in a mountainous area with complex geological conditions. Common unfavorable geological bodies include fault fracture zones, karst and weak interlayers. In order to improve construction safety and efficiency, it is decided to apply the random forest algorithm based on Bayesian optimization provided by the present invention to identify unfavorable geological bodies in front of the tunnel.

[0051] Step 1: Collect geological data through geological analysis, geophysical exploration, drilling and other methods, and pre-process the data;

[0052] Collect data covering different geological sections and their adverse geological types. Obtain geological data from geological analysis at the tunnel construction site, geological drilling, seismic method, induced polarization method and geological radar. The collected data include stratigraphic lithology, geological structure, joint spacing, porosity, groundwater level and other characteristics related to adverse geological bodies. Historical geological hazard data and environmental data (such as rainfall, temperature) will also be taken into account to more fully evaluate the geological conditions.

[0053] In this embodiment, during the tunnel excavation process, the seismic wave propagation velocity (wave velocity data) is obtained by the seismic method to identify low wave velocity areas, such as fault fracture zones and karst; the resistivity and polarizability data are obtained by the induced polarization method to detect underground aquifers, weak interlayers and mineralized areas; the electromagnetic wave reflection data is obtained by the geological radar (GPR) to identify cracks, joints and stratum thickness; the detailed geological information such as lithology, porosity, water content, density, mineral composition, etc. is obtained by geological drilling; the rock type, mineral composition, etc. are obtained by geological analysis.

[0054] Specifically, in this embodiment, the seismic method analysis shows that there is a low-speed wave band 500 meters in front of the tunnel, extending about 200 meters; the induced polarization method analysis shows that the resistivity is low and the polarizability is high, indicating that there may be an aquifer or a weak interlayer; the geological radar analysis shows that the reflection signal is strong, indicating the development of cracks; the drilling analysis shows that the porosity is high and the water content is large.

[0055] Step 2: Analyze and screen the geological data to extract features related to adverse geology;

[0056] Among them, firstly, the K nearest neighbor interpolation method is used to fill in the missing values ​​that may appear in the data collection process to ensure data integrity; secondly, due to the different dimensions of different data sources (such as wave velocity, resistivity, porosity, water content), all numerical features need to be standardized (Z-score standardization) to ensure the consistency of data processing and modeling; finally, recursive feature elimination (RFE) and analysis of variance (ANOVA) are used to select key features from multi-source geological data.

[0057] In this embodiment, key data are selected, such as: wave velocity data: low wave velocity areas correspond to the spatial distribution of fault fracture zones; resistivity / polarizability: areas with low resistivity and high polarizability may correspond to aquifers or weak interlayers; GPR reflection intensity: used to determine the thickness of the formation and the location of the fracture development zone; porosity and water content of drilling samples: used to confirm the existence of weak geological bodies.

[0058] In this embodiment, recursive feature elimination (RFE) is used to further extract important features to ensure that the model only uses the most representative information, reduce data dimensions, and avoid model overfitting. Features include, but are not limited to, topographic and geomorphic features: slope, aspect, elevation, etc.; geotechnical features: rock type, soil moisture content, porosity, density, etc.; hydrological and climatic features: groundwater level, precipitation, temperature, etc.

[0059] Step 3: Construct a random forest model that can perform preliminary classification of adverse geological bodies;

[0060] Among them, multiple subsets are randomly extracted with replacement from the preprocessed geological data to generate multiple decision trees. When each node of each decision tree is split, some features are randomly selected for splitting. This can increase the diversity of the model and ensure that different decision trees make decisions on different feature combinations.

[0061] Step 4: Adjust the hyperparameters of the random forest model (such as the number of decision trees and the maximum depth) through Bayesian optimization;

[0062] Specifically, the number of decision trees was adjusted by Bayesian optimization, and 100 trees were finally selected and the maximum depth was finally set to 15.

[0063] Among them, Bayesian optimization uses the Gaussian process proxy model to search for hyperparameters to ensure that the model achieves optimal performance. The Bayesian optimization process is as follows:

[0064] Define the hyperparameters that need to be optimized, including the number of decision trees, maximum depth, minimum number of leaf node samples, etc.

[0065] The Gaussian process is used to fit the surrogate model, and the objective function is expressed as:

[0066] ;

[0067] in, is the mean function, is the covariance function.

[0068] Use the upper confidence bound (UCB) as the acquisition function:

[0069] ;

[0070] in, represents the mean, represents the covariance, is the weight parameter.

[0071] Each iteration updates the proxy model according to the results of the current hyperparameter combination, and finally finds the optimal hyperparameters.

[0072] Step 5: Use the BMA method to assign weights to the decision trees in the model, train the final random forest model, and evaluate the model performance on the test set;

[0073] Among them, the implementation process of the BMA method is as follows:

[0074] Each decision tree The posterior probability According to Bayes' theorem:

[0075] ;

[0076] in, The training data D is in the i-th decision tree The likelihood function under is the prior probability of the i-th decision tree, which is usually uniformly distributed.

[0077] The predicted probability and posterior probability of each decision tree are used to perform a weighted average. For a given sample X, the final classification probability is the weighted sum of the predicted probability of each decision tree by the posterior probability:

[0078] ;

[0079] in, is the probability that the i-th tree predicts sample X as category Z, is the posterior weight of the i-th tree.

[0080] By comparing the weighted prediction probabilities of each category, the category Z with the largest probability is selected as the final classification result:

[0081] ;

[0082] In some embodiments, the final random forest model is trained using the optimized hyperparameters, and the model performance is evaluated on the test set; the evaluation indicators include classification accuracy, recall rate, F1 score, etc. The prediction effect of the model is analyzed by the ROC curve, and the area under the curve (AUC) is calculated to evaluate the classification performance of the model.

[0083] Step 6: Use the model to predict the probability of the type of unfavorable geological body, and predict the location and scale of the unfavorable geological body.

[0084] Among them, the preprocessed new data was input into the Bayesian optimized random forest model. The model predicted that the probability of occurrence of fault fracture zone was 85%, weak interlayer was 10%, and karst was 5%. Therefore, it was finally determined that the unfavorable geological body in front of the tunnel was a fault fracture zone.

[0085] Specifically, in this embodiment, the model analyzes the spatial information of the drilling data and GPR data, and based on the geographical distribution of the low wave velocity zone, low resistivity and high polarizability zone, predicts that the fault fracture zone is located 500 meters in front of the tunnel; by analyzing the spatial distribution range of wave velocity data, resistivity and polarizability, combined with the stratum thickness and lithology confirmed in the drilling data, the model predicts that the length of the fault fracture zone is 200 meters, the depth is 50 meters, and the width is about 20 meters.

[0086] Embodiment 2

[0087] according to Figure 2 This embodiment provides a tunnel bad geological identification system based on Bayesian improved random forest and Bayesian optimized random forest, which includes:

[0088] Function of data acquisition module: Data acquisition is performed by integrating various geological detection equipment (such as seismic method, induced polarization method, geological radar and drilling equipment). Each device collects different types of geological information, such as seismic wave velocity, resistivity, polarizability, porosity, water content, etc. Real-time data acquisition ensures data continuity and timeliness. The system equipment interface includes seismic sensors for collecting wave velocity data and identifying low-speed areas; induced polarization sensors for collecting resistivity and polarizability data to identify weak interlayers and aquifers; geological radar (GPR) for collecting reflection signals and identifying formation thickness and fracture distribution; drilling equipment for obtaining formation samples and extracting parameters such as lithology, porosity, and water content.

[0089] Functions of the data preprocessing module: For the missing parts in the collected data, use interpolation algorithms (such as K nearest neighbor interpolation) to fill in to ensure data integrity; standardize data from different sources and units (Z-score standardization) to unify the data scale; use recursive feature elimination (RFE) or variance analysis to select the features most relevant to the identification of poor geological bodies, such as wave velocity, polarizability, porosity, etc.

[0090] Functions of the model training and optimization module: Automatically adjust the hyperparameters of the random forest model, such as the number of decision trees, maximum depth, minimum number of leaf nodes, etc., through Bayesian optimization to ensure optimal model performance; train multiple decision trees based on multi-source geological data, and randomly select feature splitting nodes to ensure the diversity and robustness of the model; use the BMA method instead of the majority voting method, calculate the posterior probability of each decision tree, perform weighted prediction, and improve classification accuracy.

[0091] The Gaussian process proxy model is used for Bayesian optimization, and the BMA method is used to weighted average the prediction results of each decision tree.

[0092] Functions of prediction and analysis module: The system calculates the probability of occurrence of each type of bad geological body through the BMA method, for example, the probability of a fault fracture zone is 85%, the probability of a weak interlayer is 10%, etc., and finally determines it as a fault fracture zone. Combining multidimensional data such as wave velocity, polarizability, porosity, etc., the system predicts the specific location of the bad geological body, for example, the fault fracture zone is located 500 meters in front of the tunnel. The spatial distribution of the bad geological body is analyzed through geological data, and its scale (such as length, depth, width) is output. The system provides intuitive three-dimensional geological images to display the prediction results of the bad geological body and its probability, location and scale.

[0093] Some embodiments also include a decision support module function: based on the predicted geological body risk, the system automatically generates corresponding construction suggestions. For example, if the risk of the fault fracture zone is high, the system will recommend reducing the excavation speed, increasing support, and making preparations for drainage. The system can update the model prediction results in real time based on new data and adjust the construction suggestions in a timely manner.

[0094] Embodiment 3

[0095] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the tunnel poor geology identification method based on Bayesian improved random forest and Bayesian optimized random forest as provided in Embodiment 1 are implemented.

[0096] Embodiment 4

[0097] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for identifying poor tunnel geology based on the Bayesian improved random forest and the Bayesian optimized random forest as provided in Example 1 are implemented.

[0098] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products of the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of the processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present invention without creative labor shall be included in the protection scope of the present invention.

Claims

1. A tunnel adverse geology identification method based on Bayesian optimization random forest, characterized in that: The following steps are involved: Obtain geological data obtained by different detection methods and pre-process the geological data; Analyze and screen the preprocessed geological data to extract features related to adverse geology; The process of analyzing and screening the preprocessed geological data includes: using variance analysis, chi-square test and mutual information method to preliminarily screen the features in the geological data, selecting the features most relevant to the classification of bad geological bodies, and applying recursive feature elimination method to extract important features, which include topographic and geomorphic features, rock and soil features, and hydrological and climatic features; Among them, topographic and geomorphic characteristics include slope, aspect and elevation information; geotechnical characteristics include rock type, soil moisture content, porosity and density; hydrological and climatic characteristics include groundwater level, precipitation and temperature; Construct a random forest model to preliminarily classify adverse geological bodies; Use the Bayesian optimization algorithm to optimize the hyperparameters of the random forest model, use the Bayesian model averaging method to assign weights to the decision trees in the model, and train the final random forest model; The final random forest model is used to predict the probability of the type of bad geological bodies, and to predict the location and scale of bad geological bodies; The process of predicting the location and scale of unfavorable geological bodies includes: using the trained Bayesian optimized random forest model to predict the spatial distribution data obtained by the preprocessed seismic method, induced polarization method and geological radar method, using the model's feature weights and prediction results to determine the specific location of the unfavorable geological body, analyzing the spatial overlap of low wave velocity areas, high polarization rate areas and strong geological radar reflection signal areas, and locating the specific coordinates of the unfavorable geological body or the precise position from the starting point of the tunnel.

2. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The geological data include stratigraphic lithology, geological structure, joint spacing, porosity and groundwater level data obtained using geological analysis, geological drilling, seismic method, induced polarization method and geological radar method, as well as historical geological disaster data and environmental data.

3. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The process of analyzing and screening the preprocessed geological data includes: using variance analysis, chi-square test and mutual information method to preliminarily screen the features in the geological data, selecting the features most relevant to the classification of bad geological bodies, and applying recursive feature elimination method to extract important features, which include topographic and geomorphic features, rock and soil features, and hydrological and climatic features; Among them, topographic and geomorphological characteristics include slope, aspect and elevation information; geotechnical characteristics include rock type, soil moisture content, porosity and density; hydrological and climate characteristics include groundwater level, precipitation and temperature.

4. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The process of building a random forest model includes: implementing the classification task by building multiple decision trees, using a sampling method with replacement from the geological data set, randomly extracting multiple subsets, each subset is used to train a decision tree, and at each split of each decision tree, some features are randomly selected for node splitting to increase the diversity of the model. Each decision tree is classified independently, and finally the prediction results of each model are combined through the Bayesian model averaging method, and different models are given different weights.

5. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The process of optimizing the hyperparameters of a random forest model using the Bayesian optimization algorithm involves: Define the hyperparameters that need to be optimized; Use Gaussian process to fit the surrogate model and construct the objective function expression; Using the confidence upper bound as the acquisition function, each iteration updates the proxy model according to the results of the current hyperparameter combination, and finally finds the optimal hyperparameters.

6. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The process of assigning weights to decision trees in a model using Bayesian model averaging includes: Calculate the posterior probability of each decision tree according to Bayes' theorem; Use the predicted probability and posterior probability of each decision tree to perform a weighted average to calculate the final classification probability; Compare the weighted prediction probabilities of each category and select the category with the highest probability as the final classification result.

7. The method for identifying poor tunnel geology based on Bayesian optimization random forest as claimed in claim 1, characterized in that: The process of training the final random forest model includes training the final random forest model with the optimized hyperparameters and using a portion of the preprocessed geological data as a test set to evaluate the performance of the trained model. The evaluation indicators include classification accuracy, recall rate and F1 score. The prediction effect of the model is analyzed through the ROC curve, and the area under the curve is calculated to evaluate the classification performance of the model.

8. A tunnel bad geological identification system based on Bayesian optimization random forest, characterized by: include: A data acquisition module is configured to acquire geological data obtained by different detection methods and pre-process the geological data; A data preprocessing module is configured to analyze and screen the preprocessed geological data to extract features related to adverse geology; The process of analyzing and screening the preprocessed geological data includes: using variance analysis, chi-square test and mutual information method to preliminarily screen the features in the geological data, selecting the features most relevant to the classification of bad geological bodies, and applying recursive feature elimination method to extract important features, which include topographic and geomorphic features, rock and soil features, and hydrological and climatic features; Among them, topographic and geomorphic characteristics include slope, aspect and elevation information; geotechnical characteristics include rock type, soil moisture content, porosity and density; hydrological and climatic characteristics include groundwater level, precipitation and temperature; The model training and optimization module is configured to build a random forest model to perform preliminary classification of adverse geological bodies; use the Bayesian optimization algorithm to optimize the hyperparameters of the random forest model, use the Bayesian model averaging method to assign weights to the decision trees in the model, and train the final random forest model; A prediction and analysis module is configured to use the final random forest model to predict the probability of the type of bad geological body and predict the location and size of the bad geological body; The process of predicting the location and scale of unfavorable geological bodies includes: using the trained Bayesian optimized random forest model to predict the spatial distribution data obtained by the preprocessed seismic method, induced polarization method and geological radar method, using the model's feature weights and prediction results to determine the specific location of the unfavorable geological body, analyzing the spatial overlap of low wave velocity areas, high polarization rate areas and strong geological radar reflection signal areas, and locating the specific coordinates of the unfavorable geological body or the precise position from the starting point of the tunnel.

Images