Efficient Probabilistic Stability Analysis of Geosynthetic Reinforced Slopes Using Collocation-Based Stochastic Response Surface

Abstract

This paper presents an efficient and robust probabilistic approach to analyze the geosynthetic reinforced slopes (GRSs). The deterministic analysis will be performed that employs the rigorous (5N−1) formulation of the horizontal slice method (HSM), which is made efficient using the nonlinear constrained optimization (N = number of slices). The probabilistic analysis will be performed using a surrogated assisted Monte Carlo simulation (MCS). Collocation based stochastic response surface (SRS) will be employed to build the surrogate model using a third-order multidimensional polynomial chaos expansion (PCE). The random variables include the internal friction angle of soil (ϕ), soil unit weight (γ), and tensile strength of the reinforcement (Tu). Dependence between the random variables will be established using the Gaussian copula. A comparative analysis of the results with the First-Order Reliability Method (FORM) will be presented. The performance function will be evaluated 125 times using the SRS method in contrast to the direct MCS where it will be evaluated ≥50,000 times. This will reduce the computation time from approximately 2 days to approximately 20 min. In addition, the influence of correlation between the random variables will be highlighted comprehensively by adopting a wide range of correlation coefficients. This study concludes that the SRS method that incorporates an accurate deterministic model is a highly efficient and powerful approach to analyze GRSs probabilistically.