How we can determine convergence by the imbalances of the equations?

AliMadayen · October 1, 2014, 16:52

Hi Everyone

I remember reading some posts regarding the problem that CFX has some problems with very fine meshes. right now i'm working on a problem and I'm checking grid independency of the results.
In the coarse meshes, solution easily converges. But as I make the mesh finer, residuals do not fall below a certain value and begin to alter around that specific value, without decreasing. As I make the mesh finer, this certain value Increases.
However, the domain Imbalances in all cases are very small for all equations. Like 0.3%, at most.

now the question is, can I consider my solution converged, when the residuals are about 1e-4 but imbalances smaller than 1?

P.S. I searched the from as much as I could. I hope I'm not posting a repeated question.
P.S. 2: The problem is cavitation in a minichannel. i've used physical timescales as small as 1e-6, yet the residuals do not decrease more than a certain amount. I use Rayleigh-Plesset model. and these results have good consistency with experimental images of the cavitation region. I just wanna make sure i'm not faking the validation.

ghorrocks · October 5, 2014, 07:33

As you refine the mesh the result is probably becoming transient. So you will probably need a transient simulation to fully converge it.

AliMadayen · October 12, 2014, 18:51

Tnx for your reply.
So you mean that the imbalances alone are not reliable? I just want to grab the concept of that.
I heard someone saying that the sum of cut-off error in fine mesh prevents the solution from being fully converged. Is that true?

ghorrocks · October 12, 2014, 19:01

It is not as simple as that. It is discussed in this FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

The imbalances alone are generally not a reliable measure of convergence. That is why the equation residuals is the default convergence criteria - it is more generally applicable, but even it is not universally applicable.

Yes, round off error can prevent tight convergence. But so can many other things too.

AliMadayen · October 13, 2014, 06:36

I got what I wanted.
Thanks a lot.

October 1, 2014, 16:52	How we can determine convergence by the imbalances of the equations?	#1
AliMadayen New Member Ali Madayen Join Date: Apr 2014 Posts: 22 Rep Power: 12	Hi Everyone I remember reading some posts regarding the problem that CFX has some problems with very fine meshes. right now i'm working on a problem and I'm checking grid independency of the results. In the coarse meshes, solution easily converges. But as I make the mesh finer, residuals do not fall below a certain value and begin to alter around that specific value, without decreasing. As I make the mesh finer, this certain value Increases. However, the domain Imbalances in all cases are very small for all equations. Like 0.3%, at most. now the question is, can I consider my solution converged, when the residuals are about 1e-4 but imbalances smaller than 1? P.S. I searched the from as much as I could. I hope I'm not posting a repeated question. P.S. 2: The problem is cavitation in a minichannel. i've used physical timescales as small as 1e-6, yet the residuals do not decrease more than a certain amount. I use Rayleigh-Plesset model. and these results have good consistency with experimental images of the cavitation region. I just wanna make sure i'm not faking the validation.

October 5, 2014, 07:33		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	As you refine the mesh the result is probably becoming transient. So you will probably need a transient simulation to fully converge it. AliMadayen likes this.

October 12, 2014, 19:01		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	It is not as simple as that. It is discussed in this FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria The imbalances alone are generally not a reliable measure of convergence. That is why the equation residuals is the default convergence criteria - it is more generally applicable, but even it is not universally applicable. Yes, round off error can prevent tight convergence. But so can many other things too. AliMadayen likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Guide: Writing Equations in LaTeX on the CFD Online Forums	pete	Site Help, Feedback & Discussions	27	May 19, 2022 03:19
problem with Min/max rho	tH3f0rC3	OpenFOAM	8	July 31, 2019 09:48
accelerate the convergence	entropie	OpenFOAM Running, Solving & CFD	2	October 10, 2014 18:58
Time step dependence of convergence behavior of steady state simulations in CFX	Chander	Main CFD Forum	5	December 23, 2013 05:31
convergence parameters	amv	Phoenics	4	September 21, 2005 06:49

October 12, 2014, 18:51		#3
AliMadayen New Member Ali Madayen Join Date: Apr 2014 Posts: 22 Rep Power: 12	Tnx for your reply. So you mean that the imbalances alone are not reliable? I just want to grab the concept of that. I heard someone saying that the sum of cut-off error in fine mesh prevents the solution from being fully converged. Is that true?

October 13, 2014, 06:36		#5
AliMadayen New Member Ali Madayen Join Date: Apr 2014 Posts: 22 Rep Power: 12	I got what I wanted. Thanks a lot.