The distinct element method (DEM) is a numerical model to describe the mechanical behaviour of discontinuous bodies. The method is based on the use of an explicit scheme that monitors the particle interaction through contacts and calculates the particle motion through Newton's second law. Calculations alternate between the application of Newton's law to the particle motions and a force-displacement law at the contact points. The method has been tested against experimental results obtained through the use of photoelastic discs in determining contact forces, displacements and rotation of the individual disks. The flexibility of numerical modelling in load configurations, particle sizes, size distributions and physical properties of the particles adds more power to the method's accuracy to make it a viable research tool into the behaviour of two-dimensional granular media. Neverthless, the potential of the DEM is not fully exploited due to the limited computer power used with its sequential algorithm. A typical simulation of the DEM, as applied to the molding sand in the lost foam casting technology, takes unreasonable amount of time on currently available RISC computers for particles more than a thousand.
An efficient parallel implementation of the DEM, as pointed out by its original authors Cundall and Hart, may make it feasible to run simulations with large number of particles. Hence, in this paper we present a parallel DEM algorithm and its implementation on a distributed-memory multiprocessor, particularly the Intel architectures. Preliminary results are indeed good and they signal a potential for better performances.