ABSTRACT
A DEM simulation is a transient analysis where the response of the system at discrete points in time is predicted based upon the system state at slightly earlier times. Therefore, specication of the initial conditions is as important as specifying the boundary conditions. From an applied, geomechanics perspective, the response of a granular material is known to be highly dependent on its initial state (packing density and stress level), the anisotropy of the fabric (determined from the particle and the contact orientations), the anisotropy of the stress state, and the orientation of the principal stresses relative to the fabric. Just as experimentalists expend signicant eort in preparing their samples for physical tests, DEM analysts need to give careful consideration as to how they construct their specimen. While Poschel and Schwager (2005) (who consider particulate simulations from a general perspective) argue that as the initial conditions are used only once in each simulation, users should not need to expend signicant eort in dening the initial conditions, this statement is not really valid from a geomechanics perspective. In fact Bagi (2005), looking at granular materials from a general perspective, identies the production of the random arrangement of densely packed particles
\dembookwcrop" | 2011/2/22 i
needed for many applications, to be a \challenging task." It is worth noting that signicant eort has been expanded by mathematicians to develop algorithms to generate dense random assemblies of spheres (e.g. Sloane (1998) and Jodrey and Tory (1985)) and there is potential to adopt these algorithms in DEM specimen generation. This Chapter discusses some of the approaches used to generate initial particle congurations for DEM simulations. While consideration is restricted here to generation of specimens to ll simple geometries, e.g. cylinders for triaxial compression tests, the methods can also be applied to more complex boundary geometries that might be encountered in simulation of eld-and industrial-scale problems.
