Impact of Random Field Simulations on FEM-Based Earth Slope Reliability

This paper compares the probabilistic analysis results of earth slopes using Random Variable (RV) and Random Field (RF) approaches, with a focus on potential differences in the mobilized sliding mass. The Finite Element Method (FEM) with Shear Strength Reduction (SSR) is utilized to determine Pore Water Pressure (PWP) and the Factor of Safety (FoS). In the RF approach, random FEM is conducted using "crude" Monte Carlo Simulation (MCS) and the Karhunen-Lo & egrave;ve expansion, while the RV approach employs the First-Order Reliability method (FORM) and Importance Sampling Monte Carlo Simulation (IMCS). Sensitivity analysis was performed to reveal the most important parameters. The RV results may either underestimate or overestimate the probability of failure (Pf) compared to those obtained by the RF approach. Smaller Pf values are observed in RF with smaller correlation lengths (L) of soil properties. The mean value of the factor of safety (mu FoS) closely matches the deterministic FoS when L is largest, while the coefficient of variation of FoS (COVFoS) increases. Finally, in the RF analysis, it was found that intermediate sliding volumes carry higher risks, with smaller volumes having higher occurrence probabilities and larger volumes having much lower probabilities.