Research On Comprehensive Evaluation Of Large Deformation And Prediction Method Of Deformation Potential For Squeezed Surrounding Rock Tunnel

The large deformation of the extruded surrounding rock tunnel is a time-dependent viscoelastic-plastic deformation.During the tunnel excavation process,the stress exceeds the ultimate shear stress,resulting in a large-scale plastic failure zone.The broken surrounding rock squeezes out of the tunnel boundary and causes serious damage to the supporting structure,The deformation usually occurs in tunnels with weak surrounding rock and high in-situ stress.At present,the research on the evaluation and prediction method of large extrusion deformation is more and more abundant,from qualitative development to quantitative development,there are empirical discriminant,comprehensive evaluation method,artificial intelligence and numerical simulation.However,different types of forecasting methods have certain limitations,and there is no generally applicable forecasting method with high forecasting performance.In summary,this paper fully considers the shortcomings of different types of forecasting methods,proposes new forecasting methods,and analyzes and discusses the advantages,disadvantages and application scope of different types of forecasting methods.In order to preliminarily judge whether the surrounding rock of the tunnel will undergo large compressive deformation,a relatively rich database is established in this paper,and two empirical formulas H=350Q^1/3and H=424.2Q^0.32 to predict the known data;and further use the Kriging interpolation method to judge the large deformation level of the compressive surrounding rock tunnel.The research shows that the prediction accuracy rates of the two empirical formulas are:82.79%and79.87%respectively;the kriging interpolation method performs well,and the prediction accuracy rate of its test set reaches 88.24%.The empirical discriminant has limited influence parameters,strong subjectivity,and low reliability of prediction results.To sum up,in view of the comprehensive evaluation method that can consider more influencing factors and the advantages of the unascertained measure theory in solving the uncertainty problem,this paper constructs an unascertained measure comprehensive evaluation model for the large deformation of the extruded surrounding rock tunnel.The study found that the model performed well,the prediction accuracy of the first set of data were:100%,83.33%,50%and 83.33%,and the accuracy of the second set of data were:70%,77.5%,67.5%and 70%%;The more comprehensive the influencing factors considered in the comprehensive evaluation model,the higher the prediction accuracy of the model;the linear function,parabolic function and sine function can be preferentially used to construct a comprehensive evaluation model with uncertain measures.However,the more influencing factors are considered,the more difficult it is to establish the index level,and the practicability of the comprehensive evaluation model decreases accordingly.Machine learning algorithms can solve the above shortcomings,therefore,this paper constructs the WOA-SVM classifier model.Four input parameters are selected:burial depth（H）,support stiffness（K）,rock tunnel quality index（Q）,diameter（D）and relative deformation（ε）.Simultaneously build three single classifier models（SVM,ANN,GP）to participate in performance evaluation.The research results show that the WOA-SVM model has high prediction performance,and its prediction accuracy is 95.65%;the contribution of?,H and K to the prediction results of the WOA-SVM model decreases in turn.