A hybrid parallel approach for fully resolved simulations of particle-laden flows in sediment transport is proposed. To overcome the challenges of load imbalance in the traditional domain decomposition method when encountering highly uneven distributions of particles in space, we develop a hybrid parallel approach adopting the domain decomposition method for the carrier phase and a mirror domain technique for the disperse phase. We modify the mirror domain technique originally developed for point particles to fully resolved particle simulations, which are more challenging since a finite-sized particle may be split into different subdomains; thus, more complex treatments of particle-fluid interactions are needed. By utilizing the mirror domain technique, in which each processor handles nearly the same number of particles regardless of the particle spatial distribution, excellent load balance is achieved. The present hybrid parallel approach also shows strong scalability and high parallel efficiency in a test of a fully resolved simulation case of sediment transport. Furthermore, a novel memory optimization method is proposed for spherical particles of equal size, which can substantially reduce the memory cost and enable the simulation of millions of fully resolved particles on a common highly parallel computing platform. Our code is validated by several benchmark cases, and the results show good agreement with experimental and computational data in the literature.
Download