Material point method for large deformation analyses of coseismic landslides

Abstract

Estimation of coseismic landslides is of particular importance for seismic hazards assessment and mitigation strategies. To realitically and efficiently simulate the complete landslide process during earthquake from triggering to post-failure large movements, a robust computational tool is necessary. In this thesis, a physics-based numerical framework is developed based on the Material Point Method (MPM) for estimating coseismic landslides under the influence of wave motion, topographic amplification and hydrogeological conditions. Various numerical examples are simulated with an emphasis on the failure mechanisms and the large deformation behavior. The major findings from the thesis are summarized below: (i) A two-layer hydro-mechanically coupled Material Point Method has been formulated in the framework of unsaturated soil mechanics. The method employs two layers of material points to represent coupling effects between the solid and fluid phases. Several numerical examples demonstrate unique features of the method, including free water infiltration into the unsaturated soil and dynamic soil-water interaction in an example of underwater soil collapse. Furthermore, rainfall-induced failure in unsaturated soil slopes is analyzed. The proposed method can well capture change in matric suction, development of shallow and deep-seated shear bands, and finally, large-deformation post-failure behavior. Parametric studies also demonstrate soil cohesion, dilatancy and friction angle play significant roles on the slope failure mechanisms. (ii) The Material Point Method is further developed to simulate dynamic slope stability and liquefaction-induced embankment failure under earthquake loading. First, by using elastic or elastoplastic models, topographic amplification and different slope failure modes are analyzed considering the effects of slope geometry, soil properties and excitation frequencies etc. The MPM model is then applied to predict a cascading slope failure process, including triggering, shear band formation, runoff and final deposition. Finally, a fully nonlinear bounding surface soil model is implemented in the two-phase soil-water coupled MPM framework to investigate the liquefaction mechanism and associated dam failure, using Success Dam and Lower San Fernando Dam as two examples. The numerical results are generally comparable with the post-failure profiles obtained from field investigation, which highlight the advantage of MPM in handling liquefaction-induced large deformation. The MPM shows great promise to quantitatively assess risk and consequence associated with seismic slope failure and soil liquefaction, thereby, advance the performance-based engineering design and analysis. (iii) A computational framework based on Spectral Element Method (SEM) and Material Point Method (MPM) has been developed for multiscale, large-deformation analysis of coseismic landslides. At a regional scale, SEM is used to model elastic wave propagation from seismic source to a local site, such that topographic amplification, soil response and near-field characteristics of earthquake shaking can be simulated. At a local scale, the progressive landslide process and large deformation behavior are simulated by a fine-scale nonlinear MPM model. The SEM and MPM models are coupled through a domain reduction method, which is validated through a benchmark example. In this study, the coupled SEM-MPM method is used to simulate the massive Hongshiyan landslide triggered by the 2014 M_s 6.5 Ludian earthquake in China, in which a whole picture from fault rupture to regional-scale seismic wave propagation to landslides triggering, runoff and deposition is simulated. The post-failure morphologies of the landslide simulation are generally in agreement with those from field investigations, showing the SEM-MPM method is a promising physics-based numerical tool for a multiscale analyses of regional coseismic landslides.

Date of Award	2021
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Gang WANG (Supervisor)