Landslides are a serious natural hazard which causes loss of lives and damages the infrastructure all around the world. Therefore, to protect the infrastructure and reduce the fatalities, one needs to have a proper numerical tool that can predict a completed process of landslides from the triggering phases, the onset of the failure and post-failure to the final deposition. This research investigates the landslides from the geotechnical engineering point of view and by means of recent advanced numerical method, so-call Material Point Method (MPM).
Indeed, the MPM has been recently used for back-analyses of many landslides in the past ten years. However, there is very little research to access the reliability of the MPM for the landslide modelling even when there are some concerns on the numerical stability and accuracy of the MPM which reported in literature. This poses a challenge to improve the algorithmic performance of the MPM to make this method more attractive and reliable for landslide modelling.
Motivated by this challenge, this work aims to validate the capability of the MPM with various benchmarks and replicate a spread failure of a sensitive clay landslide in Sainte-Monique, Quebec in 1994. The extensive validations can reveal the advantages and limitations of the MPM in order to improve the MPM formulation. The contribution of this research is to enhance the stability and the accuracy of the MPM. The former is done by mitigating non-physical velocity/pressure oscillations using a temporal and a null-space filter for the MPM. The latter is to improve the spatial convergence rate of the MPM by using an improved moving least squares method and gradient velocity enhancement. Overall, these proposed developments improve the MPM algorithm and potentially apply to the multi-phase MPM in the future.
|Publication status||Published - 2019|
|MoE publication type||G5 Doctoral dissertation (article)|
- material point method, landslides, fallcone test, strain rate effects, progressive failure, large deformation modelling