Random Large Deformation Analysis of Unsaturated Slopes Using Data-Driven and Physics-Informed Method

Landslides would travel certain distances during the post-failure period. Once occur, the human's lives and properties along its runout route will be threatened, underlining that the accurate estimation of landslide consequence is of great necessity. In this study, a data-driven and physics-based random coupled Euler-Lagrange (RCEL) method is proposed to evaluate the unsaturated slope horizontal runout distance. During the slope initial failure stage, both the site investigation data and field monitoring data are adopted to conduct the Bayesian updating and then the conditional random fields can be generated. Both the spatial variability and the epistemic uncertainty can be considered. During the subsequent post-failure stage, the RCEL analysis is performed and the influence of the soil strain softening on the landslide runout behaviours is also discussed. A case study of Baishuihe landslide in the Three Gorges Reservoir area of China is selected as an example to demonstrate the application of the developed method. It is found that both Bayesian updating and soil strain softening effect cannot be overlooked, otherwise the landslide runout distance would be underestimated.

Keywords: Random large deformation analysis, Spatial variability, Unsaturated slopes, Coupled Euler-Lagrange method, Strain softening, Data-driven and physics-informed.