Parallelized domain decomposition techniques for multiphase flows

Direct simulation of multiphase flows is a challenging task due to the moving interface and property variations between phases. In this study, a parallel domain decomposition method is implemented for such flows to lower the computing cost. Specifically, the approach consists of the additive Schwarz method for domain decomposition, the projection method for the Navier-Stokes equations, the immersed boundary method for treating the interfacial dynamics, and the multigrid method to expedite the solution of the pressure Poisson equation. The issues related to load balancing, communication and computation, scalability in regard to grid size and the number of processors, and interface shape deformation, are studied using both SGI Altix and Linux-based Beowulf systems. As the number of processors increases, as expected, the domain decomposition technique results in modest decrease in convergence rate, while the multigrid technique is effective in reducing the computational cost. The additional computational cost incurred by the immersed boundary method for tracking the interface is not significant.