项目名称: 解不可压缩Navier-Stokes方程的若干过滤分解预处理子
项目编号: No.11301420
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 牛强
作者单位: 西交利物浦大学
项目金额: 22万元
中文摘要: 不可压缩Navier-Stokes方程数值求解过程中所产生的大规模稀疏线性方程组具有不定、维数高、病态化等特点。这些大规模稀疏线性方程组的求解对传统预处理迭代算法构成挑战。本项目研究Navier-Stokes方程离散化所产生的结构线性方程组预处理问题。研究内容包括开发能同时消除高频和低频误差的动态过滤分解预处理子;基于矩阵重排和分而治之的多水平过滤分解预处理子;以及高效组合预处理子。项目拟从模型问题入手,以Fourier分析为主要工具,以预处理之后矩阵的条件数达到最优为目标来解决预处理子构造过程中的最优参数确定及多个预处理子的最优组合等问题。该项目的研究成果将为进一步开发并行、多核等高性能预处理技术建立基础,对提高复杂粘性流体的数值模拟效率具有重要意义。
中文关键词: Navier-Stokes equation;预处理子;迭代法;傅立叶分析;
英文摘要: Large sparse linear systems arising from the discretization of Navier-Stokes equations are generally indefinite, highly dimensional and ill-conditioned. The solution of such kinds of linear systems poses challenge to the conventional preconditioned iterative solvers. In this project,we consider high efficient preconditioning techniques for structured linear systems arising from the numerical solution of incompressible Navier-Stokes equations. In particular, we will develop an class of adaptive filtering decomposition preconditioners,which are able to eliminate both the high and the low frequency errors.Based on matrix reordering and the idea of divided and conquer, we will develop a class of multilevel frequency filtering decomposition preconditioners. Besides, we will also consider developing high efficient combinative preconditioners. We plan to use Fourier analysis to solve the related optimal parameter determination problem and optimal combination problem. The results of the project will form a basis for developing parallel and multicore preconditioning techniques, and is significant for enhancing the efficiency of some complex fluid simulations.
英文关键词: Navier-Stokes equation;preconditioner;Iterative methods;Fourier analysis;