USING DELAUNAY TRIANGULATIONS TO INVESTIGATE THE EFFECT OF INTERPARTICLE FRICTION ON CRITICAL-STATE DEM SIMULATIONS

KJ Hanley and X Huang and C O'Sullivan, PARTICLE-BASED METHODS IV- FUNDAMENTALS AND APPLICATIONS, 890-899 (2015).

It is now broadly accepted that failure in a granular materialis dominated by buckling failure of strong force chains oriented in the direction of the major principals tress. In this paper, Delaunay triangulation is used to explore the role of the inter particle friction coefficient (mu) in stabilising these strong force chains. An isotropic numerical sample was created in 3D which contains 20,164 polydisperse spherical particles inside a periodic cell. mu was changed to 0.0, 0.1, 0.25, 0.5, 0.75 or 1.0 to create six samples, each of which was sheared triaxially until critical state was attained. The contact network was obtained at the end of each simulation by joining the centroids of the contacting particles. The Voro++ software library was used to compute two types of periodic Voronoi tessellation of the particle centroids: the conventional unweighted tessellation and the radical tessellation which is weighted by particle diameter. These were converted to their Delaunay duals to facilitate direct comparison with the associated contact networks. The numbers of edges (i.e., contacts) that are in the contact networks decrease consistently as mu is increased towards 1. This contrasts with the numbers of edges, faces or tetrahedra in the Delaunay triangulations, all of which increase with increasing mu up to mu = 0.25 and become approximately constant there after. Two quantities were defined: the percentage of faces in the triangulation comprising three contacts (P-fc) and the percentage of tetrahedra comprising six contacts (P-tc). P-fc shows a linearly-decreasing trend with increasing angle of shearing resistance and both P-fc and P-tc decrease continuously as mu is increased. This indicates that triangular motifs become less widespread as friction increases and that the weak supporting network which is present at low mu values is effectively absent at high mu.

Return to Publications page