Convergence Acceleration of Fluid-Flow Computation with Special Emphasis on Multigrid Technique

Abstract

This thesis is concerned with the study and analysis of various convergence-acceleration strategies that can help CFD practitioners to obtain their simulation results quickly. Convergence-acceleration can be achieved either by choosing a fast algorithm or by parallelizing the solution strategy. The multigrid method lies in the former category and is considered as one of the most powerful convergence-acceleration strategies. CUDA, which is a relatively new method for parallelizing CFD codes falls under the second category and as a part of the convergence-acceleration efforts embodied in this thesis, CUDA is used to parallelize LBM-based computations