文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > On convergence of iterative methods for a fourth-order discretization scheme
侵权投诉

On convergence of iterative methods for a fourth-order discretization scheme

We prove that, under certain conditions, some classical iterative methods converge for the linear system resulting from a fourth-order compact discretization of the convection-diffusion equation. Key words: Linear systems, iterative methods, fourth-order c

However, we are not aware of any analytical result to prove that any of the classical iterative methods converge with these fourth-order compact schemes. A rigorous justi cation is always desirable in spite of the fact that numerical experiments have been successfully conducted, In this paper, we give some conditional convergence results for some classical iterative methods using the fourth-order compact scheme developed by Gupta et al 4]. Although our results are limited, they are a rst step towards the convergence analysis of such iterative methods for the fourth-order approximation schemes. This paper is organized as follows: In Section 2, we give the stencil of the fourth-order compact scheme. In Secti

on 3 we prove the convergence of the point Jacobi and point Gauss-Seidel methods with the fourth-order scheme for the di usion-dominated case (de ned below). In Section 4 we symmetrize the coe cient matrix and give a bound for the line Jacobi iteration matrix when the coe cient matrix is symmetrizable. Conclusions are given in Section 5.

2 The Fourth-Order Compact Scheme Let h= 1=(n+1) be the uniform meshsize. The nite di erence formula for a grid point (x; y) which is denoted by`u0 ' involves the eight neighboring mesh points, which are denoted by u; i= 1; 2:::; 8. The discrete values f; i= 0; 1;:::; 4; are labeled similarly. The discretization results in a linear system (for details, see 3]): Au= f; (2) where A= (a ) 2 2 is a square matrix, which is usually nonsymmetric and non-positive de nite. Each equation of (2) is of the form: i i i;j n n

2 (3) u= h (f4+ f3+ f2+ f1+ 8f0)+ (f1? f3)+ (f2? f4)]; 2=0 where the coe cients; i= 0;:::; 8, are described by the computational stencil: 0 0 1 1?(1? )(1+ )?2(1+ )2? 2?(1+ )(1+ ) C 6 2 5 B 3 0 1 C= B?2(1? )2? 2 20+ 4 2+ 4 2?2(1+ )2? 2 A: (4)@@ A?(1? )(1? )?2(1? )2? 2?(1+ )(1? ) 7 4 8 Here= ph=2 and= qh=2 are referred to as the cell Reynolds numbers 2]. When maxfj j; j jg i i i i

8 X

1, we say that the linear system (2) (and the discretized boundary value problem (1)) is di usiondominated, otherwise it is convection-dominated. The numerical experiments conducted in 3] showed that classical iterative methods with this scheme converge for any values of p and q. We also showed numerically in 8] that the multigrid method with this scheme converges for all values of p and q even when they are functions of x and y.

3 Convergence for the Di usion-Dominated Case Lemma 3.1 The coe cients of the nine-point stencil (4) satisfy?2; i= 1; 2; 3; 4; i

(5)

for all values of and .

Proof. Direct veri cation. 2

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email