An optimization based model that simulates random packings of polydisperse particles to mimic porous media systems is presented in this paper. The model simulates systems with properties that map back to physically measured quantities. The model is composed of three main procedures: detachment of an existing sphere from a target sphere by moving its center to a new position, movement of an existing sphere to touch the target sphere, and addition of a new sphere to touch the target sphere. Each procedure is performed to satisfy a constrained nonlinear optimization formulation. Input parameters include radius and coordination numbers density functions. A monodisperse and polydisperse packings were simulated by incorporating physically based properties of glass bead packings obtained from 3D computed tomography images. In each simulated system, the packing density obtained was very close to the packing density of the physical system. The packings were tested for isotropy, homogeneity and randomness and then compared to experimental microtomography images. Findings indicate that the packings generated were isotropic, homogenous, random, and in a good agreement with the experimental data. (c) 2007 Elsevier B.V. All rights reserved.