Comparison of contact treatment methods for rigid polyhedral discrete element models

Abstract

Blocky rock and stone structures (such as jointed rocks mass, masonry structures and railway ballast) are often submitted to dynamic effects. Their mechanical behaviour can be simulated with the discrete element method (DEM) using polyhedral elements. DEM models differ in how they compute the contact forces. Several studies have compared different contact constitutive laws, but little information is available in the literature on which contact treatment method (i.e. number of contacts between two elements, contact detection, contact orientation, area and overlap computation) to choose. The inclined slope, the perpendicular impact and the rocking block problems were simulated and applied as benchmark tests in the present study to compare three most popular polyhedral DEM contact treatment models for rigid elements, namely the Eliáš (available in Yade), the Gilbert-Johnson-Keerthi (GJK, widely implemented, here tested with PFC3D) and the Sub-Contact model (available in 3DEC), to see which ones could be precisely calibrated to analytic and experimental solutions. Sensitivity studies were also carried out on the effect of subdivision density of element surfaces and on the admissible timestep length. Experiences showed that while there are serious doubts about the applicability of the Eliáš volumetric model as it is sensitive to both variations in geometry and the results vary randomly with the timestep length, the GJK model is suitable to simulate granular materials because of its computational efficiency and robustness, and the Sub-Contact method is the proper choice to model structures where the contact surfaces are relatively large, as it operates multiple contacts between two elements and gives information about the force distribution on the surface of the elements.