Numerical triaxial simulations using the discrete element method (DEM) were performed to examine how the adopted contact models and the associated parameters affect the response. The increase in the shear modulus G in the Hertz-Mindlin’s contact model can enhance the small-strain Young’s modulus and reduce the initial volumetric contraction but the influence from the Poisson’s ratio, ν, used in the same model can be neglected. Samples with a higher interparticle friction coefficient, f_s, give rise to higher shear strength and greater volumetric dilation. However, the associated peak and critical-state friction angles, ∅_p and ∅_cs, have a non-proportional relationship with increasing f_s. The addition of rolling resistance can render a similar effect to increasing f_s but enable the overall...[ Read more ]

Numerical triaxial simulations using the discrete element method (DEM) were performed to examine how the adopted contact models and the associated parameters affect the response. The increase in the shear modulus G in the Hertz-Mindlin’s contact model can enhance the small-strain Young’s modulus and reduce the initial volumetric contraction but the influence from the Poisson’s ratio, ν, used in the same model can be neglected. Samples with a higher interparticle friction coefficient, f_s, give rise to higher shear strength and greater volumetric dilation. However, the associated peak and critical-state friction angles, ∅_p and ∅_cs, have a non-proportional relationship with increasing f_s. The addition of rolling resistance can render a similar effect to increasing f_s but enable the overall response, including the resulting ∅_p, ∅_cs, ε_p (the strain at the peak strength), and dilatancy behavior, closer to the experimental observations. As the rolling coefficient, J_n, and the coefficient of rotational sliding, η, increase, the ∅_p and ∅_cs also increase until they both reach a limit and become saturated. Moreover, the resulting ∅_p and ∅_cs from the samples with a fixed rolling resistance but different f_s also show a similar non-proportional response as f_s increases. When f_s reaches a certain value, the failure is mainly by rotational sliding and not by frictional sliding because rotational sliding can occur more easily. Hence, the frictional resistance cannot be fully developed and ∅_p (or ∅_cs) ceases to increase no matter how much f_s increase. Similarly, when η reaches a certain value, the failure switches into a frictional sliding mode so ∅_p (or ∅_cs) also stops increasing even as η continues to increase.

In this study, DEM simulations on triaxial creep tests of dense and loose samples were carried out to examine the involved micromechanics during creep in sand. The simulated creep responses fairly reproduce the published experimental results. During the primary creep, the creep rate continuously decreases. This is due to that the contact forces are gradually transferred from decreasing tangential to increasing normal forces to form columnar particle structures. The columnar structures eventually completed formed and the creep rate reaches a minimum. However, the structures become meta-stable and are susceptible to buckling. This explains why a sand packing does not show an extended period of the secondary creep in the experiment. Buckling of the columnar structures also gives rise to maximum dilatancy and a sharp transition of the major fabric orientation from the horizontal to the vertical directions. The continuous process of buckling of columnar structures increases the creep rate and sliding ratios during the tertiary creep. In addition, the trend that contact tangential forces decrease while contact normal forces increase is reversed. Finally creep rupture occurs as the creep stress-strain line intersects the complete stress-strain curve. All the creep samples follow their original volume-change tendency to continue the dilation or contraction response during creep.