Abstract: A non-intrusive stochastic finite element method (NISFEM) based on Latin hypercube sampling (LHS) for reliability analysis of unsaturated slope considering the spatial variability of multiple soil parameters is proposed. The safety factor is explicitly expressed as a function of uncertain input parameters using the Hermite polynomial chaos expansion, the Latin hypercube sample points are selected as the collocation points for calculating the unknown coefficients of polynomial chaos expansion. The Karhunen-Loève (K-L) expansion method is used to discretize the random fields of soil hydraulic conductivity, effective cohesion and internal friction angle. A computer program named NISFEM-KL-LHS is developed. An example of reliability analysis of unsaturated slope stability under the steady-state seepage condition is presented to demonstrate the validity and capability of the proposed method. The results indicate that the proposed NISFEM can effectively evaluate the reliability of unsaturated slope considering the spatial variability of multiple soil parameters. Both the spatial variability of the soil hydraulic conductivity and the rainfall intensity have significant effects on the location of the groundwater table and the critical slip surface of slope. The probability of slope failure increases obviously when the ratio of the rainfall intensity to the saturated hydraulic conductivity is more than 0.01. In addition, if the spatial variability of soil properties is ignored, the probability of slope failure will be overestimated significantly when the coefficients of variation or the negative cross-correlation of soil parameters become larger.

Key words: slope, unsaturated seepage, spatial variability, reliability, Latin hypercube sampling