Deformation Prediction And Deterioration Risk Evaluation Of Gently Tilting Rock Tunnel Bottom Structure Based On Extension Cloud-Integrated Learning

Soft rock with gently dipping layers and developed interbedded joints is widely distributed in the mountainous areas of southwestern China.Due to poor cementation and the effects of tectonic stress,the rock at the bottom of tunnels is prone to bending and buckling,leading to deterioration of the lining.However,there is a lack of research on the mechanism and risk assessment of structural degradation of gently dipping rock mass in tunnels,which urgently needs to be addressed in the construction of complex geological tunnels in China.The research conclusions are as follows:（1）Using numerical simulations to study the coupling effects of rock strength,layer inclination,and stress,this study reveals the dynamic development and maximum depth of the plastic zone in gently dipping rock mass tunnels,as well as the deformation and lining deterioration mechanisms at the bottom.Results show that increasing the rock strength parameter proportionally reduces tunnel arch deformation and plastic zone depth.As the rock softens,arch deformation increases by 20.6%and lining safety factor decreases by 51.8%.Decreasing layer inclination and increasing layer dip lead to increasing bottom deformation and increasingly asymmetric arch deformation.Increasing lateral pressure coefficient and burial depth result in higher deformation and deformation rate at the tunnel arch than other positions.（2）Based on the tunnel bottom deformation grading standard and lining deterioration grading system,the tunnel bottom structure deterioration risk is divided into I～V grades,and finally the tunnel bottom structure deterioration risk grading system is determined.The direct impact factors,intermediate factors and deep factors of structural deterioration risk and the subjective weights of each factor were determined according to the FISM-ANP model,the objective weight of risk factors was determined according to the CRITIC model with improved coefficient of variation,and the combined weight of risk factors was determined according to game theory.（3）Based on the extension theory of matter,the membership degree and comprehensive certainty of risk factors of single index are calculated to determine the risk level of the tunnel to be evaluated,and the fusion algorithm code of extension-cloud theory is written based on MATLAB,which improves the accuracy of the risk assessment model by 33.3%and improves the visualization of the output results.Finally,the application of engineering examples and numerical simulations determined the rationality of the extendable cloud model for the risk assessment of structural deterioration at the bottom of the slowly tilting surrounding rock tunnel.At the same time,based on Matlab App Designer,a risk assessment system for structural deterioration at the bottom of the slowly tilting tunnel was constructed.（4）An integrated learning algorithm,including Decision Tree,Random Forest,Gradient Boost,and XGBoost,was used to establish a deformation prediction model for gently dipping tunnel bottoms.The prediction accuracy R₂ of each model was 0.87,0.89,0.91,and 0.93,respectively,with solution efficiency ranging from 0.16s to 3.71 s.XGBoost was selected as the base learner for its high prediction accuracy and efficiency.An XGBoost-LR fusion model was built to learn new sample features,optimize XGBoost parameters,and improve prediction accuracy to 0.982 by reducing the model’s generalization error var.