Prediction Of Tunnel Squeezing By SVM And Study On Its Support Measures

Tunnel squeezing is one of the major geological disasters that often occur during the construction of tunnels in weak rock masses subjected to high in situ stresses.It could cause shield jamming,budget overruns,and construction delays,and could even lead to tunnel instability and casualties.Therefore,accurate prediction or identification of tunnel squeezing and adoption of cost-effective support measures extremely important in the design and construction of tunnels.This study presents a novel application of a multiclass support vector machine(SVM)to predict tunnel squeezing based on four parameters,i.e.,diameter(D),buried depth(H),support stiffness(K),and rock tunneling quality index(Q).Firstly,we compiled a database from the literature,including 117 case histories obtained from different countries such as India,Nepal,and Bhutan,to train the multiclass SVM model;secondly,in order to improve the accuracy of model prediction,the commonly-used SVM model parameter optimization algorithms such as grid search algorithm,particle swarm optimization algorithm,and genetic algorithm are compared and analyzed,and the best grid search is performed.The algorithm is applied to the parameter optimization of SVM,and a tunnel squeezing prediction model based on multiclass SVM is established;then,the predictability of the established SVM model was evaluated by the 8-fold cross validation and the leave-one-out cross validation method.The average accuracy rate obtained by the two verification methods was 88.13% and 90.60% respectively.Compared with the existing domestic and foreign classical empirical formulas approaches and binary SVM models,the proposed multiclass SVM model not only yields a better performance in predictive accuracy,but also can estimate the severity of potential squeezing problems based on the predicted squeezing categories/classes;in addition,the model proposed in this paper can also predict the probability of occurrence of tunnel squeezing,which is of great significance for the tunnel engineering based on reliability design;finally,a visualization program for the prediction system of tunnel squeezing was developed using MATLAB programming.The system includes two parts of a single tunnel squeezing prediction system and multiple tunnel squeezing prediction systems,which can conveniently classify the tunnel squeezing,predict the probability of occurrence of tunnel squeezing,and visually display the classification result diagram.In addition,this article combines the tunnel squeezing example of the Shilibei tunnel.In the view of the optimization of support measures for soft rock tunnel squeezing,the mechanical mechanism and failure characteristics of tunnel squeezing are introduced.The reinforcement mechanism of supporting measures for soft rock tunnels is analyzed,and the role of two types of steel arch supports and anchor bolts in soft rock tunnels is simulated through FLAC3 D software,the effect of using different setting parameters of steel arches and bolt on the tunnel squeezing was obtained.Finally,the optimal steel arch type,spacing,and the optimal length of the anchor rod were obtained in this project case,which has good practical value and reference significance for similar projects in the future.