Assessment of neighbor particles searching methods for discrete element method (DEM) based simulations

Abstract

The performance of discrete element method (DEM) simulations is highly dependent on the requirements of the associated algorithms in terms of computer memory usage and CPU time. In particular, computer CPU time heavily depends on the identification of neighbor particles and the computation of particle-particle interactions. Over the years several neighbor particles searching methods have been developed. Accordingly, in this work the performance of two of well-known searching methods, linked cell and Verlet tables algorithms, are assessed in the context of the development of a new DEM-based tool. More specifically, the neighbor searching methods performance and related computational costs are parametrically analyzed and an assessment of their suitability for carrying the intended numerical simulations is provided. The referred numerical simulations are performed accounting for a canonical configuration used for the verification of the algorithms included in the new computational tool under development. The referred tool incorporating state of the art physical and numerical models will be used for modelling following a CFD-DEM approach mineral transport and grinding processes present in concentrator plants.