Hybrid Reliability Analysis Framework for Reliability Analysis of Tunnels

Abstract

Numerical modelling of tunnels, generally carried out by using some numerical analysis tools [e.g., finite element (FE) method, discrete element method], is often computationally expensive. Hence, it becomes extremely difficult to perform reliability analysis of such models using conventional reliability analysis tools. This necessitates development of efficient techniques for reliability analysis. In this work, a novel framework, referred to as a hybrid reliability analysis framework (HRAF), is developed for reliability analysis of tunnels. The proposed approach utilizes a hybrid polynomial correlated function expansion (H-PCFE), a distribution adaptive sequential experimental design (DA-SED), and an adaptive algorithm for further refining the estimates of DA-SED based H-PCFE. The primary idea of HRAF is to use DA-SED based H-PCFE in zones where the probability of misclassification is less and to use actual simulation for zones with a higher probability of misclassification. As a consequence, results obtained are highly accurate and, at that too, from a reasonably lower number of training points and/or actual simulations. Two tunnel problems have been presented to illustrate the performance of the proposed approach. Results obtained have been benchmarked against results obtained using a full-scale Monte Carlo simulation (MCS) with 105 sample points. Results obtained indicate the excellent performance (both in accuracy and efficiency) of the proposed framework.