Nonlinear deformation prediction of tunnel surrounding rock with computational intelligence approaches

Abstract

Deformation of surrounding rock is widely monitored to discover surrounding rock behaviors for purpose of event forecasting. This article aims to present a comparative study on surrounding rock nonlinear deformation prediction using computational intelligence techniques. The Gaussian process (GP), the support vector machine (SVM), and the wavelet neural network (WNN), are analyzed comparatively for predicting the surrounding rock deformation series. Two representative tunnels, the Wangdeng tunnel on Chenglan railway in China and the Ureshino tunnel line I on Nagasaki expressway in Japan, are illustrated. The results prove that the computational intelligence approaches are capable of predicting surrounding rock nonlinear deformation. The GP, on the whole, performs best. The SVM shows better ability than the WNN and GM (1, 1) not only in predicting the settlement and convergence deformation values but also in tracking the trends of surrounding rock deformation curves.