Probabilistic Slope Stability Analysis Using Novel Response Surface-Paper

Paper

Pdf Summary

The paper presents a response-surface approach to make probabilistic slope stability analysis faster while retaining useful accuracy. Direct Monte Carlo simulation (MCS) can be computationally expensive when each trial requires nonlinear finite-element analysis, particularly for complex soil behavior or low failure probabilities. The proposed method fits a second-order polynomial to a limited set of finite-element results, then uses the approximation with MCS or the First-Order Reliability Method (FORM).<br /><br />The approach was demonstrated on an idealized submarine slope modeled in PLAXIS. The analysis treated undrained shear strengths in three soil layers as independent, normally distributed variables, with no spatial variability. The fitted response surface closely reproduced the finite-element results used to build it.<br /><br />For the case study, the estimated failure probability was about 1.0–1.2%, with reliability indices of approximately 2.25–2.32 across the methods. These estimates were broadly consistent with direct finite-element MCS. The response-surface calculations took about 0.35% of the time required for direct MCS using the finite-element model. The authors note that estimates can vary with simulation count and that differences among methods are relatively small for this example.<br /><br />The study concludes that response surfaces combined with MCS or FORM can provide an efficient alternative to repeated finite-element simulations, especially when computational costs are high. The method could help make reliability-based slope assessments more practical for engineering applications.

Keywords

response-surface method

probabilistic slope stability

Monte Carlo simulation

finite-element analysis

FORM

reliability index

failure probability

PLAXIS

submarine slope

undrained shear strength