Reliability analysis of rock slopes involving correlated nonnormals

A two-dimensional rock slope in Hong Kong and a three-dimensional hypothetical tetrahedral wedge are analyzed probabilistically using an intuitive and transparent constrained optimization approach for the first-order reliability method (FORM). The results are compared with Monte Carlo simulations. Simple procedures for incorporating truncated non-normal distributions in reliability analysis are described. The effects of statistical correlations on the computed reliability index are studied. The difference between probabilities of failure inferred from reliability index and from Monte Carlo simulations are investigated via the response surface method. It is shown that the efficiency of reliability-based approach can be combined with the robustness of Monte Carlo simulation. The meanings of reliability index and probability of failure are also discussed.