Global sensitivity analysis of the reliability of the slope stability based on the moment-independent combine with the Latin hypercube sampling technique

In parameter uncertainty analysis, traditional Monte Carlo simulation (MC) requires double sampling to calculate the importance index based on the moment-independent method, which takes a long time, especially when the failure probability is small. The Latin hypercube sampling method is used to simulate the uncertainty of the random variables in this study, which can consider the correlation between random variables and guarantee the accuracy in the small sample size. The moment-independent method combined with orthogonal polynomial estimation and kernel density estimation is developed to estimate the unconditional and conditional probability density functions of the output response to calculate the importance index. The results of the proposed method are in good agreement with those of the verification example and MC, and the efficiency is higher than that of the MC. Subsequently, the influence of the correlation between the shear strength parameters is investigated, and the differences in the influence of the random variables on the safety factor (Fs) and failure probability (P-f) are compared. The results show that the correlation between the shear strength parameters will greatly affect the importance indices, and the cumulative impact of each random variable on the Fs and P-f is different. This difference indicates that the influence of the parameter uncertainty is different when different dependent variables are taken as the output responses. Moreover, the correlation should not be ignored.