Shape characteristics of granular materials through realistic particle avatars

Abstract

The micro- and macroscale mechanical behaviors of granular materials strongly depend on the morphological features of the grains. Imaging techniques, such as X-ray computed tomography, have been used to scan realistic particle geometries and obtain shape characteristics. However, these methods are expensive, laborious, and only applicable to limited grain sizes. A simpler alternative would be to estimate 3D shape features from 2D digital photographs in the field, but they might be affected by image resolution and particle orientation. This study aims to assess the effect of image quality for shape characterization and obtain relationships between widely used 2D and 3D shape descriptors through randomly generated computational particle geometries termed ‘avatars.’ Additionally, X-ray μ-CT scans are performed on beach sand grains with irregular shapes. Particle avatar-based 2D-3D form relationships are compared with μ-CT data of 38 beach sand grains and at least 2300 particles of other material types from literature. Changes in the resolution of the images affected both geometrical and morphological features of the particles. Volume and surface area measurements are affected by the 3D image quality, and low-quality 2D digital images influenced shape characterization. From the observed relationships between 2D and 3D shape descriptors, some 3D shape characteristics such as true sphericity and convexity can be reliably approximated from 2D digital images of randomly oriented particles.