Characterising 3D spherical packings by principal component analysis

Abstract

Purpose

The purpose of this paper is to extend the previous study [Computer Methods in Applied Mechanics and Engineering 340: 70-89, 2018] on the development of a novel packing characterising system based on principal component analysis (PCA) to quantitatively reveal some fundamental features of spherical particle packings in three-dimensional.

Design/methodology/approach

Gaussian quadrature is adopted to obtain the volume matrix representation of a particle packing. Then, the digitalised image of the packing is obtained by converting cross-sectional images along one direction to column vectors of the packing image. Both a principal variance (PV) function and a dissimilarity coefficient (DC) are proposed to characterise differences between different packings (or images).

Findings

Differences between two packings with different packing features can be revealed by the PVs and DC. Furthermore, the values of PV and DC can indicate different levels of effects on packing caused by configuration randomness, particle distribution, packing density and particle size distribution. The uniformity and isotropy of a packing can also be investigated by this PCA based approach.

Originality/value

Develop an alternative novel approach to quantitatively characterise sphere packings, particularly their differences.

Citation

Zhao, T., Feng, Y.T. and Tan, Y. (2020), "Characterising 3D spherical packings by principal component analysis", Engineering Computations, Vol. 37 No. 3, pp. 1023-1041. https://doi.org/10.1108/EC-05-2019-0225