A stochastic multiscale algorithm for modeling complex granular materials

Abstract

Modeling of granular materials in which the grains have irregular shapes and surface is a long-standing problem that has been studied for decades. Almost all the current models either represent the grains as particles with geometrically-regular shapes or attempt to infer some low-order statistical properties of the materials in order to describe granular media. We use an approach to modeling of granular materials that utilizes a two- or three-dimensional image of the material’s morphology. It reconstructs realizations of the image based on a Markov process, and uses a multiscale approach and graph-theoretical concepts to refine the realizations and make them free of artefacts. The method is applied to several complex 2D and 3D examples of granular materials. Various morphological properties of the models are computed and are compared with those of the original images; very good agreement is found for all the cases. Furthermore, the computational cost of the method is very low and, therefore, the method can generate large-size models for complex granular materials.