A Novel Approach for Slope Reliability Analysis Considering the Stratigraphic Uncertainty and Property Uncertainty

Slope reliability analysis plays a crucial role in assessing the stability of slopes under varying loading conditions, aiding engineers in identifying potential instability risks and implementing appropriate mitigation measures against natural disasters such as landslides. The complex geological formations and heterogeneous properties of geomaterials present significant challenges in geotechnical engineering projects. Traditional methods often overlook these uncertainties, potentially leading to underestimation or overestimation of inherent risks.This study introduces a novel approach for slope reliability analysis by integrating a stochastic stratigraphic model, in-house developed using Bayesian machine learning, with random field modeling of soil properties, creating a digital twin of the slope site. This virtual representation captures stratigraphic and property uncertainties, enhancing risk management in geotechnical engineering. The in-house developed stratigraphic model amalgamates a Markov random field (MRF) model and a discriminant adaptive nearest neighbor-based k-harmonic mean distance (DANN-KHMD) classifier within a Bayesian framework. Therefore, the stratigraphic model can highly improve the interpretation performance of stratigraphic uncertainty especially in the situation that only sparse boreholes are available. The random fields generated by the Karhunen-Loeve expansion method (KLEM) are employed to model the spatial variability of the geomaterial properties (cohesion c and internal friction angle AE), which is used to describe the property uncertainty. The random fields generated by the KLEM are based on the assumed distribution of geomaterial properties, and can effectively describe correlations between neighboring grids and different geomaterial properties. The superiority of the proposed approach over traditional methods is evidenced through its comprehensive, accurate representation of the stratigraphic uncertainty and property uncertainty in geological bodies, thereby facilitating more reliable slope reliability analysis. The efficacy of this approach is validated through a synthetic case study, suggesting its potential for broader application in industry practices aimed at enhancing risk management in geotechnical engineering.

Keywords: Slope reliability analysis, Stratigraphic uncertainty, Property uncertainty, Bayesian machine learning, Karhunen-Loeve expansion method.