1 A Statistically-based Approach on Reconstructing Sand Particles for Discrete Modeling

ABSTRACT: This paper presents an overview of a statistically-based Fourier-Voronoi approach recently proposed by the authors to generate packing of granular particles with key characteristics reproduced from real sands. The key sand properties identified in this approach include the grain size distribution, density, particle orientation and the shape characteristics. Four important shape factors are chosen to characterize the shape of a particle, including elongation, circularity, roundness and regularity. To generate a packing of particles, a 2D container of arbitrary shape is discretised by a Voronoi tessellation with prescribed cell size and orientation distributions. Each of the Voronoi cells is then filled by a particle with the prescribed shape characteristics. The proposed method is shown to be capable of realistically reproducing a variety of granular sands based on available experimental data. This method is further generalized to reproduce 3D sand shapes based on the Fourier descriptor concept, with demonstrative examples on two sands (Michigan Beach sand and Ottawa sand). The proposed methods provide a rational and systematic way to reconstruct real sand particles for discrete modelling