Do I HAVE TO get "Solution Converged" message?

aggie · April 18, 2010, 23:28

Hi all,
Do I have to get "Solution Converged" message in order to decide that solution is converged and correct?
Usually I don't get "solution converged" message even after 60000 iterations but my variables like maximum velocity, pressure, mach number and density are constant. Does that mean the solution is converged although "solution converged" message doesn't appear?
Thanks.

ic7 · April 19, 2010, 00:48

Howdy aggie,

It is my understanding that you do not necessarily have to obtain a significant residual convergence[computational accuracy] (<1e-6) so long as your nummerical model accurately depicts to the real world manifestation of your problem, and that the results obtained are accurate, to the degree of your liking, when compared or validated with experimental data.

It may also be a good idea to run your case on a refined grid (finer mesh) to see if your results are the same.

Best,
Ian

heidar · April 19, 2010, 01:56

hi sir
could ypu tell me what is your choice for the amount of your residuals?
your answer might be correct or viceversa it depends on your mesh resuloution and the determined residuals.

1- clear me the selected residuals.

aggie · April 22, 2010, 16:08

Thank you all.
Now I have grid independence and the results seem to be good. I think the solution is correct and converged but I still did not get "solution converged" message.
I increased the mesh number with Yplus and Static Pressure adaptions. I have around 150,000 nodes now and solution is still same.
Convergencce ccriteria for residuals is 0.0001.

Chris D · April 22, 2010, 17:41

Why do you think your solution has converged? You said earlier that velocity, pressure, density, etc. are all constant, but where are they constant? Over the entire domain or at one point? Also, what are you calculating? I would hesitate to say that your solution is grid-independent after only performing a few localized grid adaptions.

aggie · April 22, 2010, 22:38

I checked the values at several points, and maximum values are also constant.
I do not know whether or not they are constant at each point over the entire domain- is it possible to check it ?
I am also not %100 sure that it is converged but it seems like that.
Adaptions are not a few, I adapted the domain roughly 3k times for Yplus which should be less than 5. For static pressure, I adapted the domain about 4k times.
Do u have any idea about what should be done more ?

Chris D · April 22, 2010, 23:31

I wouldn't check each variable over the entire domain. That would be overkill. Instead, I would monitor some things that you're interested in calculating, like maybe lift, drag, flow rate, etc. What is the problem that you are trying to solve?

With respect to adapting by y+, all you have done is set the distance of the first cell off the wall. That doesn't do anything for the rest of the boundary layer. So, while your y+ might be in the recommended range for your turbulence model, it's possible that the boundary layer itself is not sufficiently resolved, meaning that your solution is not grid independent.

Same with adapting by pressure. You may resolve the flow sufficiently in one region, but what about the rest of the flow?

April 18, 2010, 23:28	Do I HAVE TO get "Solution Converged" message?	#1
aggie Member Jack Join Date: Mar 2010 Posts: 32 Rep Power: 16	Hi all, Do I have to get "Solution Converged" message in order to decide that solution is converged and correct? Usually I don't get "solution converged" message even after 60000 iterations but my variables like maximum velocity, pressure, mach number and density are constant. Does that mean the solution is converged although "solution converged" message doesn't appear? Thanks.

April 19, 2010, 01:56	heida	#3
heidar Member heidar jafarizadeh Join Date: Mar 2010 Location: Iran- Tehran Posts: 43 Rep Power: 16	hi sir could ypu tell me what is your choice for the amount of your residuals? your answer might be correct or viceversa it depends on your mesh resuloution and the determined residuals. 1- clear me the selected residuals.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Guide: Getting Started with the CFD Online Discussion Forums	pete	Site Help, Feedback & Discussions	8	July 29, 2016 06:00
single directional message transmit in parallel model	su_junwei	OpenFOAM Programming & Development	1	December 17, 2009 08:00
Error message in CFX-10	CFDworker	CFX	2	October 12, 2007 08:23
error message	susan	Siemens	0	August 17, 2007 01:27
Error Message on es-ice [BUG]	Wendy Tjia	Siemens	0	February 10, 2005 09:40

April 19, 2010, 00:48		#2
ic7 New Member Ian Join Date: Apr 2010 Posts: 4 Rep Power: 16	Howdy aggie, It is my understanding that you do not necessarily have to obtain a significant residual convergence[computational accuracy] (<1e-6) so long as your nummerical model accurately depicts to the real world manifestation of your problem, and that the results obtained are accurate, to the degree of your liking, when compared or validated with experimental data. It may also be a good idea to run your case on a refined grid (finer mesh) to see if your results are the same. Best, Ian

April 22, 2010, 16:08		#4
aggie Member Jack Join Date: Mar 2010 Posts: 32 Rep Power: 16	Thank you all. Now I have grid independence and the results seem to be good. I think the solution is correct and converged but I still did not get "solution converged" message. I increased the mesh number with Yplus and Static Pressure adaptions. I have around 150,000 nodes now and solution is still same. Convergencce ccriteria for residuals is 0.0001.

April 22, 2010, 17:41		#5
Chris D Senior Member Chris Join Date: Jul 2009 Location: Ohio, USA Posts: 169 Rep Power: 17	Why do you think your solution has converged? You said earlier that velocity, pressure, density, etc. are all constant, but where are they constant? Over the entire domain or at one point? Also, what are you calculating? I would hesitate to say that your solution is grid-independent after only performing a few localized grid adaptions.

April 22, 2010, 22:38		#6
aggie Member Jack Join Date: Mar 2010 Posts: 32 Rep Power: 16	I checked the values at several points, and maximum values are also constant. I do not know whether or not they are constant at each point over the entire domain- is it possible to check it ? I am also not %100 sure that it is converged but it seems like that. Adaptions are not a few, I adapted the domain roughly 3k times for Yplus which should be less than 5. For static pressure, I adapted the domain about 4k times. Do u have any idea about what should be done more ?

April 22, 2010, 23:31		#7
Chris D Senior Member Chris Join Date: Jul 2009 Location: Ohio, USA Posts: 169 Rep Power: 17	I wouldn't check each variable over the entire domain. That would be overkill. Instead, I would monitor some things that you're interested in calculating, like maybe lift, drag, flow rate, etc. What is the problem that you are trying to solve? With respect to adapting by y+, all you have done is set the distance of the first cell off the wall. That doesn't do anything for the rest of the boundary layer. So, while your y+ might be in the recommended range for your turbulence model, it's possible that the boundary layer itself is not sufficiently resolved, meaning that your solution is not grid independent. Same with adapting by pressure. You may resolve the flow sufficiently in one region, but what about the rest of the flow?