ABSTRACT
Numerous laboratory studies have shown that sample preparation methods result in different initial fabrics, thus affecting the observed macro-response even when specimens are prepared at the same densities (e.g., Oda 1972, Mulilis et al. 1977, Vaid et al. 1999). Sample preparation methods traditionally used in the laboratory were developed to recreate a given depositional pattern or in situ soil fabric. Preferential contact orientations can be induced even for samples comprised of spherical particles (Oda 1981). The resulting inherent anisotropy has been shown to greatly influence stress-strain response, especially for cases involving principal stress rotation (Sayao 1989). Particulate discrete element methods (DEM) codes require the specimen preparation process to be simulated and do not allow for specification of an initial dense packing or a pre-defined stress state. If the DEM simulations are to provide valuable insight into the behavior of simple shear element tests, it is important for the initial fabric and stress state of the computer generated sample to match the physical reality.
