A practical algorithm of reliability analysis for slope stability is presented based on response surface method and secondorder reliability method (SORM). Independent random variables in U-space are chosen as basic variables and transformed into correlated non-normal variables in original X-space for computing the performance functions. The response surface is constructed using an iterative algorithm to ensure its accuracy with minimum computational expense. Then

the probability of failure is evaluated using the first order reliability method (FORM) and SORM. The accuracy and efficiency of the proposed method are illustrated for a rock slope with plane failure mechanism

via comparing with Monte Carlo simulations

pure FORM and stochastic response surface method. The influence of correlation among random variables and selection of sampling range around tentative design points on the inferred reliability results are investigated. The parametric sensitivity and physical property of random variables obtained from the reliability analysis are also discussed. The proposed method can be used for reliability analysis of practical slopes and also for reliability-based design of reinforcing structures.