Question about Multigrid in CFX

Anna Tian · September 16, 2012, 07:13

Hi,

May I ask if there's an automatic way to implement the multigrid method in CFX?

ghorrocks · September 16, 2012, 07:39

Um, yes - it is the default solver option. So if you have done a CFX simulation you have used a multigrid method.

Anna Tian · September 16, 2012, 13:25

Quote:

Originally Posted by ghorrocks

Um, yes - it is the default solver option. So if you have done a CFX simulation you have used a multigrid method.

Thanks. But it seems not work very well. Because from my experience, if we do the multigrid manually, (Run the simulation for the coarse mesh firstly, then come to fine mesh with the result of coarse mesh simulation as the initial value file.), it converges much faster.

Does default multigrid method also work for multi-domain case in CFX?

ghorrocks · September 16, 2012, 18:48

It is unlikely that any problem you are having with convergence is due to the linear solver, and even more unlikely that the manual control you describe will be any better.

If you are having convergence difficulties it is almost certainly due to some setup issue with your simulation. You should fix that problem rather than trying to tweak the linear solver.

The multigrid solver is the linear solver used by CFX for all simulations. So yes, it can handle multi-domain cases as well.

cdegroot · September 16, 2012, 19:45

To clarify, the linear solver in CFX uses algebraic multigrid which works on the coefficient matrix directly rather than a hierarchy of grids. As Glenn said it is the default solver and should work quite well in practically all cases. Manually transferring between grids is unlikely to be helpful.

September 16, 2012, 07:13	Question about Multigrid in CFX	#1
Anna Tian Senior Member Meimei Wang Join Date: Jul 2012 Posts: 494 Rep Power: 15	Hi, May I ask if there's an automatic way to implement the multigrid method in CFX? __________________ Best regards, Meimei

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
question about governing equation in CFX using rotating/non rotating reference frame	rystokes	CFX	0	January 12, 2010 06:14
question about how to define diffusion coefficient in CFX	rystokes	CFX	0	December 12, 2009 05:32
Question about the shock wave in CFX software	nucharin	Main CFD Forum	1	January 25, 2005 08:26
simple question about ICEM CFX	sleepinglily	CFX	0	October 27, 2004 13:15
Question about meshing / solution scheme of CFX	Coriolius	CFX	8	August 1, 2004 18:39

September 16, 2012, 07:39		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	Um, yes - it is the default solver option. So if you have done a CFX simulation you have used a multigrid method.

September 16, 2012, 18:48		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	It is unlikely that any problem you are having with convergence is due to the linear solver, and even more unlikely that the manual control you describe will be any better. If you are having convergence difficulties it is almost certainly due to some setup issue with your simulation. You should fix that problem rather than trying to tweak the linear solver. The multigrid solver is the linear solver used by CFX for all simulations. So yes, it can handle multi-domain cases as well.

September 16, 2012, 19:45		#5
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 17	To clarify, the linear solver in CFX uses algebraic multigrid which works on the coefficient matrix directly rather than a hierarchy of grids. As Glenn said it is the default solver and should work quite well in practically all cases. Manually transferring between grids is unlikely to be helpful.