A novel update criterion of Verlet list for accelerating DEM computation when modeling geotechnical dense granular materials under periodic loading

Abstract

Discrete element method (DEM) based on graphic processing unit (GPU) is widely utilized for studying the responses of geotechnical dense granular materials under periodic or traffic loading. However, limited computational efficiency of DEM hinders its further application. Conventionally, the criterion for updating particles’ potential contact list (i.e., Verlet list) is assessing whether the maximum particle displacement in global coordinate system exceeds the threshold. Although geotechnical dense particles exhibit considerable quasi-periodic displacement under periodic or traffic loading, the potential contact targets for most particles do not change during quite a few loading cycles. Therefore, there are numerous redundant updates of Verlet list induced by quasi-periodic displacement, restricting computational efficiency. In this study, we propose a novel criterion for Verlet list updating, in which the displacement of particles in local particle coordinate system is considered. Then, the proposed criterion is plugged in the MUSEN software. By simulating previous laboratory full-scale half-sleeper model tests, the accuracy and performance of the proposed criterion are testified based on GPU computing. The results show that compared to conventional criterion, the proposed criterion reduces the updates of Verlet list by 43–68% and improves the computational efficiency by 14–47%. This study indicates a potential way to improve computational efficiency of the GPU-based DEM for geotechnical dense granular materials under periodic loading.