A Comparison of Mesh Simplification Strategies for Realistic Grains

Resumo

In recent years, the Discrete Element Method (DEM) has emerged as a powerful tool for studying granular materials. In a DEM simulation, the objective is to reproduce the grain-to-grain interactions, and by working at this scale, the method can provide new insights into granular behavior. In the early stages of DEM, the particles were represented by disks and spheres. Currently, many codes have been developed to incorporate new ways to represent the grain shape more accurately. However, this process is not simple, taking into account that more realistic shapes bring the necessity of more complex algorithms for contact detection. Furthermore, to represent realistic grains through some ways like polyhedrons, a large number of entities (i.e., points, faces, and edges) are necessary, which increases the computational costs, making the simulation infeasible in some situations. To circumvent this problem, an alternative commonly used with polyhedrons is to reduce the number of entities in the mesh, decreasing the computational costs of simulation and the level of fidelity to the real grain shape. Due to the importance of this pre-processing stage, this paper compares three strategies for mesh simplification, analyzing the impact in shape descriptors, and the computational costs involved entities in DEM simulations, when varying their level of representation of 30 realistic grains.