Research On Intelligent Prediction Method Of Hazard Risk Of Water And Mud Inrush In Karst Tunnel Based On Machine Learning

As the nation’s investment in infrastructure continues to increase,the construction of tunnels and underground engineering has been developed by leaps and bounds.At present,most of the tunnels under construction and planning in China are located in karst areas and other poor geological areas in the western region.Due to its special engineering geological conditions and hydrogeological conditions,the karst area has caused a wide distribution of carbonate strata,extensive interlaced water systems,and complex underground structures.When tunnels are excavated in this area,they are often harmed by water and mud inrush.Water and mud inrush hazards will not only destroy construction equipment,but also seriously threaten the life and safety of construction personnel.Therefore,it is of great engineering practical significance to establish a scientific tunnel water and mud inrush hazard risk prediction model.This paper extensively collects the data and information of the water inrush and mud inrush hazard cases of tunnels all around the world,analyzes the water inrush hazard factors of these tunnels,proposes the important hazard factors of water inrush in different types of tunnels,and summarizes these water and mud inrush hazard cases to bulid a Tunnel water inrush and mud inrush hazard casebase.Based on the hazard casebase,this paper establishes an intelligent prediction model and system for the risk of water and mud inrush from a karst tunnel,and this model and system have been successfully applied in engineering practice.The main research results and conclusions of this paper are as follows:(1)Through extensive literature review,this paper collects and analyzes tunnel water and mud inrush hazards cases around the world,and proposes three types of tunnel water and mud inrush hazard cases,including karst type,fault type and other types.In addition,this paper focuses on 135 cases of water and mud inrush from karst tunnels,and determines six main factors,which are groundwater level,stratum lithology,bad geology,stratum dip,negative topographic area ratio,and karst developmental characteristics.A casebase of water and mud inrush hazards in karst tunnels has been established.(2)This paper performs data preprocessing on the karst tunnel water and mud hazard case casebase.First,the datasets in the casebase are quantified,and then the outlier detection and replacement of the casebase are performed using the box plot method,and the data is reused.The average price or median of the datasets from the same data label is collected to supplement the missing values in the casebase.After that,the complete casebase is standardized,and the principal component analysis is used to reduce the dimensionality of the casebase.Finally,a pure numerical karst tunnel water and mud inrush hazard casebase is obtained.(3)The K-means method in the clustering algorithm is used to reclassify the data in the pure numerical karst tunnel water and mud inrush hazard casebase,and divide it into four risk levels,which are weak risk,medium risk,strong risk and very strong risk.A karst tunnel water and mud inrush hazard casebase with risk labels is established.(4)Based on the SKlearn in Python,three intelligent prediction models of karst tunnel water and mud inrush hazard risk based on machine learning are established They are decision tree model,support vector machine model and random forest model The "stratified sampling" method is used to divide the casebase into a training set and a test set.The K-fold cross-validation method is used to input the training set into the machine learning model to find the optimal hyperparameters of the model.The test set is used to measure the predictive performance of the three machine learning models.The results show that the prediction performance of the support vector machine model is the best,followed by the random forest model and the decision tree model.(5)Based on the Python,developed a machine learning-based karst tunnel water inrush and mud outburst hazard risk intelligent prediction system,and applied this system to the water inrush risk prediction of the Telmer Tunnel of the Chengdu-Kunming Railway.The results showed that the actual situation the same.This system has been successfully applied to the water inrush risk prediction of the Te’ermo Tunnel of the Chengdu-Kunming Railway and four other typical karst tunnels.