Generation of Realistic Sand Particles Preserving Marginal Distributions and Intrinsic Correlations of Size and Multilevel Morphology

The generation of virtual particles with realistic size and morphology is essential for DEM modelling because particle size, form, angularity and roughness jointly affect packing, contacts and deformation. Existing particle generation methods can reproduce realistic particle shapes or selected descriptor distributions, but the intrinsic dependence between size and multi-level morphology is seldom preserved and isolated explicitly. This study proposes a Gaussian-Copula and spherical-harmonic (SH) based particle generation framework for realistic sand. Empirical marginal distributions of particle volume, elongation index (EI), flatness index (FI), angularity index (AI) and roughness (RG) are combined with a Spearman-corrected Gaussian Copula to sample physically consistent descriptor vectors, which are then converted into triangular meshes by inverse SH optimization. The generated library reproduces the target marginal distributions and the main rank-correlation structure. DEM drained triaxial simulations show that the correlation-preserving cloned particles reproduce the macro- and micro-mechanical responses of the original particles under the examined monotonic drained triaxial conditions. An independent-Copula control further shows that removing size-morphology correlations has limited influence on the large-strain strength and volumetric response in these tests, but changes the contact behaviors before the critical state. The proposed framework therefore provides a controlled platform for separating marginal-distribution effects from intrinsic size-morphology-correlation effects.