Optimal Regression Model Based on Statistical Method for Predicting Tunnel Deformation: A Case Study of Large-span Mudstone Tunnels

Abstract

The prediction of tunnel time-dependent deformation data is of crucial significance to ensure the safety of tunnel construction. In order to improve the accuracy and reliability of tunnel deformation prediction, this research presents a procedure for building an optimal regression model, which is applied to a large-span mudstone tunnel. Our procedure performs a large number of regression calculations using sixty-four data sets and multiple types of regression models. The data characteristics of the calculation results are analyzed using box plots and grey decision-making method. In addition, two individual regression models are combined into regression ensemble model, for improving the accuracy of deformation prediction. In this procedure, the building of the model is divided into two stages: early stage and late stage of initial support construction, taking into account the importance of deformation prediction throughout the construction process. Finally, we compared four machine learning models and analyzed their differences in predicting tunnel deformation. The results show that the optimal regression model developed by this procedure have great performance for the early and later stages of tunnel deformation.