Parallel CFD for Flow and Transport Applications Including Unstructured and Adaptive Grids

This chapter describes some recent research on parallel simulation of fluid flow and transport processes using "PC clusters" and "supercomputers." Coupled nonlinear interactions such as those arising in the Rayleigh-Bénard-Marangoni (R-B-M) problem involving buoyancy and "thermocapillary surface tension" are discussed in the chapter. Heat transfer in a viscous incompressible fluid with a free surface is considered. The effects of buoyancy due to heating are introduced as a source term in the momentum equations via the Boussinesq approximation. The associated parallel algorithms are based on implicit schemes with "Krylov space solvers" using domain decomposition with static and dynamic partitioning. Comparative scaled speedup performance is graphed for several PC cluster systems. Part of the chapter is directed to efficient "parallel adaptive mesh refinement" (AMR) strategies. This allows resolving solution features and promoting more reliable, accurate, and stable simulations. The chapter presents "parallel computational fluid dynamics" (CFD) results from representative R-B-M flow calculations and for applications modeled by non-Newtonian fluids. Some remarks concerning large-scale parametric studies in a GRID computing environment are presented in the chapter.