An efficient parallel multi-scale direct simulation Monte Carlo algorithm to simulate three-dimensional rarefied gas flows over complex geometries is presented. The proposed algorithm employs a novel spatio-temporal adaptivity scheme. Based on the gradients of flow macro-properties, the spatio-temporal adaptivity scheme computes the cell size distribution and assigns the appropriate number of time sub-steps for each cell. The temporal adaptivity scheme provides local time step adaptation through different temporal levels employed in different cells. Spatial representation is based on a hierarchical octree Cartesian grid with low memory storage requirement. The hierarchical octree grid endows the method with straightforward and efficient data management suitable for particle ray tracing and dynamic grid refinement and coarsening. Solid objects, represented by triangulated surfaces, are incorporated using a cut-cell algorithm. A new parallelization scheme suitable for simulating strongly unsteady, non-equilibrium flows is proposed. The parallelization scheme, implemented for multi-core Central Processing Units (CPUs), significantly reduces the computational cost of modeling these flows. Performance of the method is assessed by comparing with benchmarked test cases for various rarefied gas flows.
Read full abstract