Discrete element modeling (DEM) is a useful tool for linking global responses of granular materials to underlying particle-level interactions. A DEM model capable of capturing realistic soil behavior must be calibrated to a reference dataset, typically consisting of laboratory experiments. Calibration of a DEM model often requires numerous simulations as contact parameters need to be iterated upon until the simulation results satisfactorily replicate the experimentally observed behaviors. This paper presents a sensitivity investigation that examines the effects of the contact parameters on the drained triaxial compression response of a poorly-graded sand. It then introduces a calibration procedure capable of providing contact parameters that satisfactorily reproduce the results of laboratory triaxial results in a few simulations. Results show that friction and rolling resistance coefficients jointly influence the mobilized peak and critical state friction angles, secant shear modulus, maximum dilation rate, total volumetric strain, and strain softening magnitude. These parameters also influence the mode of failure at contacts and the evolution of fabric anisotropy. The influence of mu r or mu on the triaxial response and particle-level interactions is coupled, becoming more profound as the other parameter is increased. Contact stiffness is shown to influence the shear modulus and volumetric change behavior independently of mu and mu r. An algorithm that estimates values for mu and mu r needed to reproduce experimental results is developed using triaxial response parameters from experimental datasets. The performance of the proposed calibration method is demonstrated for three natural sands showing that it provides appropriate calibrated parameters for poorly graded sands with different relative densities and confined with varying effective stress magnitudes.
