A contact detection algorithm for triangle boundary in GPU-based DEM and its application in a large-scale landslide

Abstract

Triangle boundaries are crucial geometric objects in the discrete element method (DEM). In this study, an improvement for the uniform grid method is proposed and implemented on the GPU. The improved algorithm includes the spatial discretization of triangles in which triangles are classified by their size and motion. The large and relative static triangles are separated into grid cells such that it relaxes the restriction of cell size in the uniform grid method. To verify the accuracy and efficiency of the algorithm, the improved grid method is compared with the normal uniform grid method and the linear bounding volume hierarchy (LBVH) method using a simulation of free-falling grains in a mixer. The results show that the improved grid method does significantly increase the simulation efficiency compared with the normal uniform grid method. Then, taking a large-scale landslide as an example, the improved grid method and the LBVH method are used to simulate the failure process of the landslide. The results indicate that the improved grid algorithm has higher efficiency for simulations of large-scale landslides than the LBVH method but at the cost of more memory usage.