|November 2, 2009, 23:08||
How to calculate the convergence rate of Gauss-Seidel line iteration?
Join Date: Mar 2009
Posts: 40Rep Power: 8
For the two dimensional Poissonís equation,
second-order centered difference with uniform grid is used to discretize. Then, we can obtain
where a<<c (e.g. a=0.01, c=1), hence y-line iteration should be used.
a[v(n,i+1,j)-2v(n+1,i,j)+ v(n+1,i-1,j)]+c[v(n+1,i,j+1)-2v(n+1,i,j)+ v(n+1,i,j-1)]=0
By using the Fourier expansion of the above equation, we can obtain the convergence rate mu(alpha,beta) as
According to some book, the maximal value of the above equation is
mu(alpha,beta)= | a/[ 2(a+c)-c*cos(beta)-exp(I*alpha) ] | .
I canít understand how to calculate max(mu). Anybody know? I have tried to analyze the eigenvalues of my iteration matrix, the one with maximal norm is about 0.005, however, the convergence rate in my actual numerical experiment is closed to 1/sqrt(5). How to calculate the convergence rate 1/sqrt(5) theoretically?
Thank you so much.
|Thread||Thread Starter||Forum||Replies||Last Post|
|BlockMeshmergePatchPairs||hjasak||OpenFOAM Native Meshers: blockMesh||11||August 15, 2008 07:36|
|BlockMeshmergePatchPairs polyTopoChanger||benru||OpenFOAM Native Meshers: blockMesh||3||June 29, 2008 21:24|
|errors||Fahad||Main CFD Forum||0||March 23, 2004 14:20|
|Problems of Duns Codes!||Martin J||Main CFD Forum||8||August 14, 2003 23:19|
|error while compiling the USER Sub routine||CFD user||CFX||3||November 25, 2002 16:16|