ADI-SOR convergance

November 9, 2005, 19:13

Hi,

I have written an ADI-SOR code and I guess it works fine for 1 X 1 meter domain. But when I decrease the dimensions to for example 10 microns by 10 microns, the convergance time will be very very long. I am solving a 2d rectangular problem, with heat flux at one side, and newtonian BC (h is not zero) at all other sides (All Newmann; no drichlei BCs).

My question is, should lengths of the domain have any effect on convergance time?

I appreciate any help.

Thanks, Mohsen

November 12, 2005, 20:05

Convergence shouldn't be affected by lengths if the same dimensionless parameters are used. You might be getting into round off errors of computer because of too small numbers involved in calculation. If you normalize your equatinons, and solve the normalized equation, you may resolve the problem

wen

November 14, 2005, 20:43

Thanks Wen. My equations are non-dimensional and I am using double-precision vaiables. So you think I am doing ADI in a wrong way?

November 9, 2005, 19:13	ADI-SOR convergance	#1
Mohsen Guest Posts: n/a	Hi, I have written an ADI-SOR code and I guess it works fine for 1 X 1 meter domain. But when I decrease the dimensions to for example 10 microns by 10 microns, the convergance time will be very very long. I am solving a 2d rectangular problem, with heat flux at one side, and newtonian BC (h is not zero) at all other sides (All Newmann; no drichlei BCs). My question is, should lengths of the domain have any effect on convergance time? I appreciate any help. Thanks, Mohsen

November 12, 2005, 20:05	Re: ADI-SOR convergance	#2
Wen Guest Posts: n/a	Convergence shouldn't be affected by lengths if the same dimensionless parameters are used. You might be getting into round off errors of computer because of too small numbers involved in calculation. If you normalize your equatinons, and solve the normalized equation, you may resolve the problem wen

November 14, 2005, 20:43	Re: ADI-SOR convergance	#3
Mohsen Guest Posts: n/a	Thanks Wen. My equations are non-dimensional and I am using double-precision vaiables. So you think I am doing ADI in a wrong way?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Use SOR or ADI for solving the time dependent reynolds equation	Hermano	Main CFD Forum	1	July 5, 2010 09:47
ADI implicit doesn't converge	George Papadakis	Main CFD Forum	2	July 8, 2009 13:27
algo for ADI or fractional step scheme handling crossderivative terms	HaKu	Main CFD Forum	8	April 4, 2009 02:10
Adi in 2D???sos	xiegang	Main CFD Forum	0	October 14, 2003 08:04
ADI of N-S equations	Zhou Yongcheng	Main CFD Forum	5	January 18, 2001 07:58