A Reliability Analysis Approach for Embankment Stability Considering Spatial Variability of Soil Properties

ABSTRACT Embankment stability analysis is a complex coupled problem influenced by multiple sources of uncertainty. Due to diverse deposition histories and environmental conditions, geotechnical parameters within the soil body exhibit significant spatial variability, which leads to uncertainty in the location and geometry of the critical failure surface. To investigate the impact of this spatial variability on embankment stability reliability, this study proposes a comprehensive simulation framework that integrates the non‐intrusive stochastic finite element method, neural network modeling, and Monte Carlo simulation. Based on random field theory, a spatially variable soil model is established to simulate the mechanical heterogeneity of embankment materials. The proposed approach allows for the efficient estimation of the probability of failure and provides a robust tool for reliability analysis. Numerical examples demonstrate that spatial variability plays a key role in determining the overall stability of embankments, emphasizing the necessity of incorporating realistic soil property distributions in stability evaluations.