An image based clump library for DEM simulations

Abstract

To create realistic virtual soil specimens for Discrete Element Method simulations, a library containing nearly 100,000 “clumps” was developed. A clump essentially models a soil particle. It consists of numerous overlapping spheres in 3D, or circles in 2D, that are tangent to the particle perimeter. By using a unique corner-preserving algorithm based on the classic 2D definition of particle roundness, the clump generation requires many fewer circles than by previous algorithms. In this paper, the clumps are based on 2D images of real soil particles and they are indexed in the library by their roundness R and sphericity S values. A real soil can be simulated by choosing particles from the library to match the soil’s actual distributions of R and S. The clumps are also enlarged or reduced to match a desired particle size distribution. The utility of the clump library in parametric studies was demonstrated by direct shear tests on five very different virtual materials created from clumps in the library.