Probabilistic risk assessment of underground rock caverns

Abstract

There have been significant advances in the use of computational methods for the design of underground rock caverns over the last thirty to forty years. One limitation of the conventional deterministic design approach is that the stochastic nature of the rock mass properties, the in situ stress fields and the geometrical complexities cannot be taken into account explicitly. In contrast, these uncertainties can be accounted for by adopting probabilistic methods. This thesis is intended to systematically carry out probabilistic analysis on underground rock cavern(s), with five aims. The first one is to develop a deterministic framework of combining numerical and empirical design approaches for analyzing the performance of the cavern(s). The second aim is to build predictive models for cavern(s) performance estimation. The third is to develop a probabilistic framework incorporating the limit state functions into reliability assessment methods to perform reliability analysis. The fourth aim is to investigate different approaches of developing surrogate models for implicit performance functions, and demonstrate the use of these surrogate models coupled with reliability methods. The final aim is to illustrate a rational and practical approach to assess the system reliability for cavern(s) design.