I just compute a 3D incompressible unsteady fluid use the LU-SGS methold. But it convergence is too slower. Can you tell me? how to use the parameter

which parameter do you mean ? Referring to what paper ?

LU-SGS based on Roe scheme with roe matrix approximated by its spectral radius or Block LUSGS wich exact jacobians for the main diagonal and exact flux differences for the off-diagonal terms ("exact" = flux difference with the flux function that is also used for the spatial operator)

LUSGS as solver or as preconditioner for Krylov subspace method ?

Frank

I just have read the

1: Lower-Upper Symmetric-Gauss-Seidel Method for then Euler and Navier-Stokes Equations author:Yoon S. and Jameson A. AIAA J. Vol 26 No.9 p1025-1026

2: Three-Dimensional Incompressible Navier-Stokes Solver Using Lower-Upper Symmetric-Gauss-Seidel Algorithm author:Yoon S. and Kwak D. AIAA J. Vol 29 p874-875

they both used the parameters of pseudocompressiblility¡¡¦Â and¡¡¦Á¡¡and ¦Ê¡¡¡£

a).this is the use of the ¦Ê A¡À=1/2[A¡&Ñ(A)I] ¦Ñ(A)=¦Ê max[¦Ë(A)]

b).the following is the using of the ¦Á the discreted Equations are

[I+¦Á¦¤t(D¦ÎA+D¦ÇB+D¦ÆC)]¦ÄQ=-¦¤tR

c).¦Â multiple pseudotime derivative of pressure was andded to the continuity equation.

how to use the a suit of the parameter? I think ¦Ê and ¦Â affected each other at least.

Thank you for your help!

¡¡

Check L. Dailey's recent (after 1996) paper. You may get some useful information.

Thank you John. Can you give me more detail about it, so I can search for the paper.

The LU-SGS scheme of approximate spetral radius version can not directly compute unsteady flow due to lack of time accuracy. To be general, alfa=1, kappa=1, and beta=1 is a good baseline choice. Acceleartion may be obtained from combination of kappa and beta, but no general rules exist for different problems. Another approach os Preconditioning. To my knowlege, there are no strong recommendations using Preconditioning from the authors.

Thank you,Versi. I will take a test of alfa=1, kappa=1, and beta=1. I have never compute this suit.

But can you tell me, there are other methold can accelerate the solution recently.

I don't have the exact reference title, but I guess you can send an e-mail at pletcher@iastate.edu to ask about Dailey's papers and thesis. Dr. Pletcher was his Ph.D advisor. I've used those papers and thesis for my research and I got more than enough information about LU-SGS scheme. Good luck.

Thank you. I have sent a mail to him.

My purpose is to find a effective methold to accelerate the current methold we used and my classmate told me the LU-SGS may be faster. I take a look at the process of this methold and find that it calculates less than ours. So I think it may be work.

I aslo studied other papers and got a methold of ILU(0)+GMRES. I can't find the 3D ILU(0) formula. Referenced from: A Comparison of Impicit Schemes for the Icompressible Navier-Stokes Equations with Artificial Compressibility Stuart E. Rogers AIAA Paper 95-0567

What is your major? Can you help me about this? Thanks for you.