Simulation of 3D effect of excavation face advancement using a neural network trained by numerical models

Abstract

By the use of convergence-confinement method, three-dimensional problem of tunnel excavation is simulated by an equivalent two-dimensional plane strain analysis. In this method, the evaluation of the convergence occurred before the time support starts interacting with ground is the critical point. The aim of this paper is to assess this convergence for deep tunnels excavated in elastoplastic continuum and anisotropic stress conditions with the aid of a neural network approach. Numerical 3D FE models supply data sets required for the training process of the network. About 170 circular tunnels between 100 and 1000 meters deep, excavated in fair to good rock masses (according to RMR classification), are analyzed. The trained network will be capable to evaluate the convergence values for different distances to the excavation face with regard to rock specifications and stress conditions and is used for a sensitivity analysis of the parameters involved. (A) This paper was presented at Safety in the underground space - Proceedings of the ITA-AITES 2006 World Tunnel Congress and the 32nd ITA General Assembly, Seoul, Korea, 22-27 April 2006. For the covering abstract see ITRD E129148. Reprinted with permission from Elsevier.