Investigation of efficient geometric shape algorithms for numerical simulation of discrete particle systems

Abstract

The efficiency of a discrete particle simulation implementation relies on several factors, including the geometric representation, contact resolution, and neighbor-sorting algorithm used. The focus here is on the first two of this list: the geometric representation and the corresponding contact resolution. An argument is developed to advocate the use of geometric representations with inconstant radius based on results of a numerical study of angles of repose for deposited grains. The equivalent spheres method is then developed as a potential cure for the problems arising from constant radius geometric representations. A coherent approach to utilizing this geometry in discrete element modeling is developed. This includes the proposal of an accurate, robust contact resolution algorithm and explicit functions to describe moments of inertia. Considerations for numerical modeling are also addressed, including numerical integration, formulation of rotation transformations, and resolution of forces and motions in the context of rigid body motions. The details of a generalized computational implementation of the representation are also given, and empirical convergence properties are compared with a different method for detecting contact between ellipsoidal approximations.