Efficient slope reliability analysis based on representative slip surfaces: a comparative study

Abstract

Slope reliability analysis can be conducted based on representative slip surfaces (RSSs) more efficiently than the conventional analysis based on many potential slip surfaces (PSSs). Various methods for selecting RSSs are proposed to enhance the efficiency of slope reliability analysis. These methods, however, generally require a complex calculation procedure (e.g., evaluation of reliability index for each PSS and/or correlation coefficients among PSSs) that cannot adaptively single out the RSSs, and the selected RSSs by these methods are commonly related to the statistics of soil properties. This leads to the question of how to efficiently and adaptively identify the RSSs of a slope for a subsequent reliability analysis with many parametric studies. To answer this question, an adaptive K-means clustering-based RSSs (AKCBR) selection method has been recently developed that is able to select the RSSs adaptively and efficiently from many PSSs. The RSSs identified by AKCBR do not vary with the variation of soil statistics, such as the inherent spatial variability that is beneficial to slope reliability analysis involving many parametric studies. As such, limitations of the available methods are tackled in AKCBR. A comprehensive comparative study is conducted in this paper to explore in detail the strength and weaknesses of the AKCBR against the available methods. Four slope examples that represent four kinds of slope stability problems are considered. Results show that AKCBR provides reliability results comparable with the available methods in terms of probability of failure and the most dominant failure modes, and it is generally more efficient. The AKCBR can adaptively identify the RSSs of slopes belonging to different types, and the RSSs are statistically robust against the statistics of soil properties, which is beneficial to reliability analysis involving many parametric studies.