Multiblock multigrid grid generation algorithms: Overcoming multigrid anisotropy

摘要

The paper presents multiblock multigrid algorithms for the generation of boundary-fitted grids (BFGs) able to avoid undesired multigrid anisotropy arises. The algorithms combine a domain decomposition technique, the elliptic grid generation and the multigrid computation and provide satisfactory numerical solutions and fast convergence. Block matching, based on two-surface overlapping, saves grid characteristics through the block boundaries. We assume two- and three-dimensional generation systems allowing line-spacing control and we design block by block multigrid computation to solve the transformed finite-difference systems. We identify anisotropy as an undesirable multigrid effect and we overcome anisotropic behaviors by investigation over coefficients of the generating systems. They can provide information to select appropriate relaxation algorithms. A wide series of performance evaluation results is available. One-block and multiblock so as one-grid and multi-grid algorithms have been evaluated and meaningful results are collected in tables. Visualizations of 2D and 3D computed grids and displays of convergence histories, comparing multigrid components, are shown by figures.