SR-DEM: an efficient discrete element method for particles with surface of revolution

TL;DR

SR-DEM introduces an efficient discrete element method tailored for particles with surfaces of revolution, leveraging geometric symmetry for faster contact detection and accurate simulation of granular systems.

Contribution

The paper presents a novel SR-DEM framework that combines a node-to-cross-section contact algorithm with a hybrid 2D/3D formulation for improved efficiency and simplicity.

Findings

Accurately predicts post-impact velocities and packing properties.

Demonstrates efficiency in simulating particle interactions and granular packing.

Provides a method to optimize surface resolution for system property convergence.

Abstract

In this paper, the surface of revolution discrete element method (SR-DEM) is introduced to simulate systems of particles with closed surfaces of revolution. Due to the cylindrical symmetry of a surface of revolution, the geometry of any cross-section about the axis of rotation remains the same. Taking advantage of this geometric feature, a node-to-cross-section contact algorithm is proposed for efficient contact detection between particles with a surface of revolution. In our SR-DEM framework, the contact algorithm is realized in a master-slave fashion: the master particle is approximated by its surface nodes, while the slave particle is represented by a signed distance field (SDF) of the cross-section about the axis of rotation. This hybrid formulation in both 2D and 3D space allows a very efficient contact calculation yet relatively simple code implementation. We then apply SR-DEM to simulate particle-particle, particle-wall impact, granular packing in a cylindrical container, and tablets in a rotating drum, to demonstrate SR-DEM's ability to predict the post-impact velocities, packing porosity, and dynamic angle of repose, respectively. Finally, we suggest a simple approach to find an optimal surface resolution, by increasing the number of surface nodes until some of the bulk properties that could characterize the system converge.