This paper investigates the response of Ottawa sand to cyclic loading using virtual oedometer tests and the level-set discrete element method. We study both the macroscopic and the micromechanical behavior, shedding light on the grain-scale processes behind the cyclic response observed in crushable sand, namely stress relaxation under strain control and ratcheting under stress control. Tests without particle breakage first show that asymmetrical frictional sliding during loading-unloading induces these cyclic-loading effects. Then, tests considering particle breakage reveal more pronounced stress relaxation and ratcheting, which decrease in rate over cycles, accompanied by increased frictional sliding and reduced particle contact forces. It is found that the broken fragments unload the most and promote an enhanced cushioning effect. These micromechanical processes contribute to a decrease in breakage potential as the cycles progress, implying that cyclically loaded materials may become more resistant to breakage when compared to the same material loaded monotonically at the same strain level. These new insights highlight the main contributions of the present work, factoring in real particle shapes from 3D X-ray tomography and notably contributing to the existing literature on the topic, where most studies rely on idealized particle shapes and rarely consider crushable grains.

In this work, we numerically investigate the quasi -static shear behavior of ellipsoids under triaxial compression using the level set discrete element method (LS-DEM). Assemblies composed of ellipsoids with various aspect ratios are prepared at the densest states and then sheared to the critical state. Macroscopically, the stress and dilation behaviors are strongly affected by the particle shape, with the spheres having the least shear strength and dilatancy. At the particle scale, more ellipsoidal particles are more resistant to particle rotations and can effectively increase friction mobilizations.We identify the clusters in assemblies via the three-dimensional cluster labeling algorithm and then analyze the structural and mechanical properties of clusters. Based on our analysis, we find that the clusters exhibit the power -law decay in the cluster size distribution and have fractal structures. Upon shearing, the clusters tend to self -organize to gain mechanical stability, indicated by the increasing cluster stress ratio, and mainly support the deviatoric stresses in the assemblies. The mean cluster stress ratio is found to be linearly related to the macroscopic shear strength at the critical state, where more ellipsoidal shapes can gain higher cluster stress ratios, contributing to higher shear strength for the granular assembly. Microscopically, the cluster contributes more significantly to geometrical anisotropy terms while comparably to mechanical ones compared to the non -cluster.