Realistic particle shape representation using clump in DEM: A critical state soil mechanics study

Recent advances in discrete element modelling (DEM), aided by X-ray micro-computed tomography, enable realistic particle shape reconstruction using clumped spheres. While an increasing number of elemental spheres per clump N ele enhances shape fidelity, it significantly raises computational cost due to the increase in number of total spheres involved in each step of computation. At present, no clear consensus exists on how to optimise clump reconstruction for realistic shapes in DEM to balance efficiency and predictive accuracy. This study addresses the computational challenges arising from numerically representing complex particle geometries. Results show that key shape descriptors and macro and micro mechanical behaviours of a realistic sand particle can be effectively captured with N ele of 25, avoiding the need for excessive values. More importantly, under critical state theory, triaxial simulations revealed that deviations in elongation index (<6.48 %), flatness index (<4.74 %), volume (<9%), and sphericity (<6.40 %) were sufficient to reproduce critical state lines, stress ratios, and behavioural correlations involving the initial state parameter. At the micro scale, the framework of bimodal stress transmission demonstrated that increased N ele mainly affected the weak contact force subnetwork, i.e., fabric anisotropy. In contrast, the strong contact force subnetwork, dominant in stress transmission, remained unaffected. These findings support efficient DEM modelling of realistic particles without requiring excessive shape detail, thereby reducing computational resource demands and facilitating large-scale studies and practical applications.