A new GPU-based DEM simulator for polydispersed granular systems with wide size distribution and its application in the silo discharge

Abstract

While discrete element method (DEM) has been successfully established in many studies, serious problems have limited its industrial-scale applications. Considerably long runtime is one of the most critical bottlenecks of the DEM simulation applicability. Despite extensive efforts on the parallelization of DEM on the CPU or GPU, DEM runtime on the current generation of computers is so long that further improvements are demanded. Moreover, while many real-world granular systems consist of polydispersed particles with a relatively wide size distribution, the majority of DEM simulation studies have assumed monodispersed particle assemblies. Few have studied the parallelization of polydispersed systems, and fewer have developed GPU-based codes for these systems. The main purpose of this study is to provide a novel solver, NP-DEM, which is optimized for GPU-based simulation of polydispersed particle systems with a wide size distribution. Silo discharge is chosen as the case study to examine the applicability of the code.