Abstract:A reasonable representation of stochastic uncertainties is a key part in the reliability assessment of rock slope. Normal To Anything (NORTA) is a rapid generation algorithm of multi-dimensional random vector with any marginal and feasible correlation matrix. Taking the reliability of the Sau Mau Ping rock slope stability as an example, NORTA was applied to generate correlated random numbers with specified marginal and correlation coefficients. Then, the instability probability of the Sau Mau Ping rock slope was estimated for typical cases using the Monte Carlo Simulation (MCS) procedure. The limitations of the approach, the randomness of the simulation results and the estimation approach for sample size based on the control of variation coefficient were also discussed. Results showed a remarkable difference of the instability probability between the natural case considering the dependency and the case ignoring the dependency for the Sau Mau Ping rock slope. In addition, with the increase of the algebraic value of correlation coefficients between cohesion and frication angle or/and between tension crack depth and degree of filling water, the instability probability of the Sau Mau Ping rock slope increased gradually as well.
