Effect of sand particle shape on micromechanical modeling in direct shear testing

Soil representation in discrete element method (DEM) simulations is a complex process due to the intrinsic variability in grain shape. Many methods have been proposed to include shape effects in DEM, from rolling friction to aspherical particle representations such as ellipsoids or polyhedra. In this paper the micromechanical differences between two commonly used methods of shape representation, rolling friction applied to spherical particles and ellipsoids, are investigated for direct shear testing. This simulation study is supported by dry direct shear tests in the laboratory using a poorly graded sand and three vertical confining pressures: 50, 100 and 200 kPa. While the bulk behavior observed in the laboratory can be matched using both shape representation approaches, the micromechanics show significant differences. The average cumulative rotation of the ellipsoidal particles is an order of magnitude lower than for the spheres with rolling friction. The addition of rolling friction significantly affects the mechanical redundancy of the system, making the spherical-particle samples more hyperstatic than their ellipsoidal equivalents. Although using ellipsoids enables a better representation of the micromechanical behavior, this increases the computational cost five-fold compared to using rolling friction and spheres. •Direct shear testing of dry sand using rigid wall boundaries in DEM.•Numerical samples composed of ellipsoidal and spherical samples were tested.•Calibration and validation were done using data from laboratory experiments.•Ellipsoidal-particle samples matched laboratory data without rolling friction.•Considerable differences were found in particle rotation and redundancy index.