CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

multigrid methods

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   April 24, 2016, 00:33
Default multigrid methods
  #1
Senior Member
 
david
Join Date: Oct 2012
Posts: 138
Rep Power: 6
davidwilcox is on a distinguished road
did anybody ever have success by using the F cycle for pressure? did you have to use small values for parameters like restriction and prolongation? i noticed on this forum that for the problem of the solution not converging, many suggested reducing the relaxation terms. While this indeed helps, it does not always solve the problem. Please share your experiences?
davidwilcox is offline   Reply With Quote

Old   April 24, 2016, 12:13
Default
  #2
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 2,936
Rep Power: 34
FMDenaro will become famous soon enoughFMDenaro will become famous soon enough
Quote:
Originally Posted by davidwilcox View Post
did anybody ever have success by using the F cycle for pressure? did you have to use small values for parameters like restriction and prolongation? i noticed on this forum that for the problem of the solution not converging, many suggested reducing the relaxation terms. While this indeed helps, it does not always solve the problem. Please share your experiences?

I am not familiar with multi-grid, I just used it many years ago for the 2D vorticity-stream function but I am interested in this question.
What I remember is that you can suppose for example a Jacobi interation on nested grids with restriction and prolongation operations in such a way to remove large wavenumbers of the error.
I which sense multigrid can alter the convergence of the pressure solver?
I immagine that convergence depends only on the spectral radius of the iteration matrix, even for multigrid, isn't it?
FMDenaro is offline   Reply With Quote

Old   April 24, 2016, 20:30
Default
  #3
Senior Member
 
Michael Prinkey
Join Date: Mar 2009
Location: Pittsburgh PA
Posts: 300
Rep Power: 15
mprinkey will become famous soon enough
I'm sorry that my experience is not more current, but I do remember using F-cycle as a compromise for an electric field solver that I built for part of the Solid-Oxide Fuel Cell Model in in FLUENT in the mid-00s. That linear system contained extreme differences in electrical conductivity (aka diffusivity) between metal interconnects and electrodes and also had huge jump conditions due to electrolytic interfaces. It was the ugliest linear system that I have ever tried to solve...so much so that I spent weeks verifying that I had not, in fact, messed up the coefficient formulation. It was just very hard to solve efficiently. The only method that converged reasonably well was W-cycle, but for our large simulations, we needed to run in parallel and (at least at the time), W-cycle scaled poorly. V-cycle did not work well at all and often error-ed out due to some floating point over/underflow. We finally used F-cycle with the ILU(0) smoother and, maybe bicgstab for the outer iterative scheme with the AMG as a preconditioner and got reasonably good results.

So, if you have access to ILU for the MG smoother, maybe try that with F-cycle. That...if I recall correctly...made the most significant improvement. Adding bicgstab just shaved a bit off of the solution time when used with ILU, but did nothing to improve robustness. Again, these are 10-year memories, but I think this is mostly right.

Good luck.
mprinkey is offline   Reply With Quote

Old   April 25, 2016, 01:35
Default
  #4
Senior Member
 
Arjun
Join Date: Mar 2009
Location: Nurenberg, Germany
Posts: 583
Rep Power: 13
arjun is on a distinguished road
Quote:
Originally Posted by davidwilcox View Post
did anybody ever have success by using the F cycle for pressure?
Yes. I think the default scheme for pressure solver for starccm is F and seeing that many users dont change defaults, it is a lot successful.


Quote:
Originally Posted by davidwilcox View Post
did you have to use small values for parameters like restriction and prolongation?
You dont need to do all that, in fact in many cases over relaxing type parameter helps a lot with cenvergence. That is prolongation is more than 1 time.

Restriction usually is not touched.

Quote:
Originally Posted by davidwilcox View Post
While this indeed helps, it does not always solve the problem. Please share your experiences?
Flow solvers convergence is most part have to do with over all algorithm out side of matrix solver.

PS

for pressure equation though, nature of linear solver can affect the convergence but for majority of cases it does not create divergence (As the outer algorithm takes care of things).
arjun is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
About multigrid methods Lionel S. Main CFD Forum 10 April 20, 2007 09:22
multigrid sureshkumar Main CFD Forum 0 June 9, 2006 01:20
comments on FDM, FEM, FVM, SM, SEM, DSEM, BEM kenn Main CFD Forum 2 July 18, 2004 18:28
Multigrid methods Amith Main CFD Forum 3 April 5, 2002 23:50
Multigrid applied to k-e models Paulo Zandonade Main CFD Forum 9 May 24, 1999 08:10


All times are GMT -4. The time now is 03:23.