Parallelisation for finite-discrete element analysis in a distributed-memory environment

A parallel computational strategy based on a distributed-memory environment is presented for simulating combined finite-discrete element systems comprising a large number of separate bodies. An explicit central difference scheme is used for the temporal integration of the governing equations. Some key issues, such as partitioning algorithms, load balance schemes and contact handling methods are discussed. A dual-level domain decomposition strategy is proposed to perform the dynamic domain decomposition. An implementation of this proposed strategy on cluster computing systems is described. MPI is adopted as the message passing library in this implementation. Numerical examples show that this implementation is suitable for simulating large scale problems. A dragline bucket filling model with 3 million degrees of freedom is built to demonstrate the parallel efficiency and scalability on a PC cluster.