Existing theories and computer models for packed columns are either incapable of handling complex pellet shapes or based on oversimplified packing geometry. A digital packing algorithm, namely DigiPac, has recently been developed to fill the gap. It is capable of packing of particles of any shapes and sizes in a container of arbitrary geometry, and is a first step towards a practical computational tool for reliable predictions of packed column properties based on the actual pellet shapes. DigiPac can operate in two modes: a Monte Carlo mode in which particles undergo directional diffusive motions; and a Discrete Element mode where translations and rotations of particles are governed by physical laws. The former is faster but in certain cases less accurate, whereas the latter is slower but produces significantly more accurate predictions. Both modes have been used in simulating packed columns of real pellet shapes. Results for cylinders – one of the most commonly used shapes for packed columns – are reported. Comparisons are made between DigiPac predictions under different modes and experimental data obtained using nuclear magnetic resonance (NMR) imaging technique. Good agreement between simulation and NMR results has been observed.