Slope stability evaluation by clustering analysis

Study on searching methods of slip surface plays a crucial role in slope stability analysis to assess the safety factor of slopes. The potential failure surface under ultimate stress state can be found by limit equilibrium method (LEM) and strength reduction method (SRM). The objective of this paper is to present a new slip surface searching method, which considers the distribution of local safety factor and displacement under current stress state. The statistical concept, K-means clustering, is introduced to distinguish the points with similar characters in slope into potential failure set and stable set. Then the boundary of the potential failure set is regarded as the potential slip surface under current stress state and the safety factor of potential slip surface is calculated using vector sum method (VSM). Afterward, three examples were calculated using the presented method to compare with results from LEM and SRM. Results show that 1) Slope stability evaluation by clustering analysis has been proved to be reliable and effective; 2) If the slope is near the limit state, the potential surface obtained from the proposed method is consistent with the slip surfaces obtained from LEM and SRM. If the slope is stable, the potential slip surface under current stress state is deeper and larger but is with higher safety factor. If there is a weak layer in the slope, the potential slip surface is steeper, and the safety factor calculated from VSM is lower than that calculated from LEM and SRM.