arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 0805.3041math.NAcs.NAmath-phmath.MP
收藏

实现更好的多重网格求解器收敛性的方法

Methods toward better Multigrid Solver Convergence

John T. Wallis

AI总结:

研究通过调整各向异性、网格层数、预处理器平滑步数、坐标系和初始向量等条件,采用F-cycle多重网格技术求解简单泊松问题,并分析求解器性能。

摘要:

我提出了多重网格技术可以应用的几个领域的动机。我介绍了多重网格求解器可能应用的典型领域。我介绍了平滑器以及如何选择预处理器,并介绍了所使用的多重网格技术。然后,我研究了多重网格技术,同时调整求解器和问题的环境条件,如各向异性、使用的网格层数、预处理器平滑步数、坐标系和初始向量。这里解决的问题是一个简单的泊松问题。本文中使用的多重网格程序采用了F-cycle。我包含了性能研究部分,展示了求解器在不同条件下的行为结果。

英文摘要:

I present a motivation of several areas where the Multigrid techniques can be employed. I present typical areas where the multigrid solver might be employed. I give an introduction to smoothers and how one might choose a preconditionor as well as an introduction of the Multigrid technique used. Then I do a study of the Multigrid technique while adjusting the environment conditions of the solver and the problem such as anisotropies, the grid-levels used, the preconditionor smoothing steps, the coordinate system and the start vector. The problem solved here was a simple Poisson problem. The Multigrid program used an F-cycle in this paper. I include performance study sections displaying results of the solver behavior under the different conditions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑