A novel granular discrete element method (DEM) is introduced to simulate mixtures of particles of any convex shape. To quickly identify pairs of particles in contact, the method first uses a broad-phase and a narrow-phase contact detection strategy. After this, a contact resolution phase finds the contact normal and penetration depth. A new algorithm is introduced to effectively locate the contact point in the geometric center of flat faces in partial contact. This is important for a correct evaluation of the torque on each particle, leading to a much higher stability of stacks of particles than with previous algorithms. The granular DEM is used to generate random packings in a cylindrical vessel. The simulated shapes include non-spherical particles with different aspect ratio cuboids, cylinders and ellipsoids. More complex polyhedral shapes representing sand and woodchip particles are also used. The latter particles all have a unique shape and size, resembling real granular particle packings. All packings are analyzed extensively by investigating positional and orientational ordering.

@en

A hybrid collision integration scheme is introduced, benefiting from the efficient handling of binary collisions in the hard sphere scheme and the robust time scaling of the soft sphere scheme. In typical dynamic dense granular flow, simulated with the soft sphere scheme, the amount of collisions involving more than two particles are limited, and necessarily so because of loss of energy decay otherwise. Because most collisions are binary, these collisions can be handled within one time step without the necessary numerical integration as needed in a soft sphere method. The remainder of the collisions can still be handled with the classical soft sphere scheme. In this work the hybrid collisions integration scheme is shortly described and tested with a bounding box problem. The hybrid scheme is capable of solving the same problem as a classic soft sphere scheme but is roughly one order of magnitude faster.@en