Capturing the random mechanical behaviour of granular materials: a comprehensive stochastic discrete-element method study

This research pioneers a stochastic discrete-element method (DEM) by integrating the probability density evolution method (PDEM), offering a novel approach to connect particle-scale property uncertainties, specifically inter-particle friction coefficient (μ) and particle shear modulus (G p ), with macro-scale soil behaviour. Through 1100 DEM simulations, this study reveals that, for uniform particle size distribution, the uncertainty in μ substantially affects large-strain soil behaviour, with its effect being associated with packing density and soil state. The uncertainty effect of μ remains pronounced at the critical state, while the packing density effect diminishes. Stress distribution appears insensitive to uncertainty of μ, rather suggesting a predominant influence of particle size distributions. In contrast, the uncertainty effect of μ becomes negligible on small-strain behaviour, demonstrating a limited effect on small-strain stiffness. Uncertainty in G p presents limited effects on large-strain behaviour, including stress ratios and dilatancy. At small strains, G p shows a significant impact on stiffness, diverging from minimal influence identified for μ. This study presents a framework that integrates experimental techniques to study particle-scale uncertainty propagation, enhancing predictions of macro-scale soil behaviour. This approach could be beneficial for precise multi-scale simulations, incorporating particle-level uncertainties in engineering-scale models, thereby improving geotechnical practice predictability.