Large deformation slope failure — A perspective from multiscale modelling

Abstract

Slope failure analysis is a traditional topic in geotechnical engineering. Current continuum modelling with phenomenological constitutive models fails to capture the physical mechanism at the particle scale. Discrete numerical methods may well describe particle scale features but are limited by the computational cost to handle practical engineering problems. In this paper, a coupled MPM/DEM hierarchical multiscale model is developed and implemented for slope failure analysis, aiming to obtain a comprehensive understanding of the slope failure and post-failure behaviours. For improving the numerical stability, the standard MPM version is modified by incorporating the Affine-particle-in-cell (APIC) velocity update format and B-spline basic function. A represent volume element (RVE) replacement approach is proposed in solving the excessive deformation problem during the slope failure simulation. Both the dynamic failure processes of cohesive and non-cohesive soil slopes are investigated. It is found that for non-cohesive soil slopes, the stability is mainly controlled by the microscopic frictional property and gradation, while the post-failure stage is dominated by particle level friction. For cohesive soil slopes, Johnson–Kendall–Robert (JKR) cohesive model is adopted in representing adhesive effect between particles. The results show that the slope failure pattern transmits from collapse to shear failure with increasing surface energy density. By statistical analysis on the attractive force between particles, the strength of the shear band is found to be controlled by cohesive effect for the top part and frictional property for medium and bottom parts.