How to ascertain convergence

hedley · June 7, 2014, 02:24

In laymans terms how would one check if solution is converging or has solved . Any practical example eg look at xxxx column in the history file and if it started at 11.2 in value it should reduce to yy in 5000 iterations .

Stating residuals should be 6 orders of magnitude less just confuses me .

hlk · June 10, 2014, 13:45

Thank you for your question.
The residuals are printed as the log of their value. This means that for a 6 order of magnitude decrease the number printed should decrease by 6. In your example, a six order of magnitude convergence would be indicated by 11.2 - 6 = 5.2. The number of iteration it takes to converge will depend heavily on the problem, the mesh, and the CFD settings chosen. Tools like multigrid and the CFL number can accelerate convergence.

hedley · June 10, 2014, 13:51

Thank you for this - I have been watching the output and 5000 iterations on my PC take approximately 3 hours and the residuals hardly move - This is on the sample wing SU2_CFD turb_RAM6.cfg

has anyone run this on a PC so I know what to expect

June 7, 2014, 02:24	How to ascertain convergence	#1
hedley Member Hedley Join Date: May 2014 Posts: 52 Rep Power: 12	In laymans terms how would one check if solution is converging or has solved . Any practical example eg look at xxxx column in the history file and if it started at 11.2 in value it should reduce to yy in 5000 iterations . Stating residuals should be 6 orders of magnitude less just confuses me .

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence	Centurion2011	FLUENT	48	June 15, 2022 00:29
Force can not converge	colopolo	CFX	13	October 4, 2011 23:03
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 02:17
Problems with convergence with an easy system	franzdrs	Main CFD Forum	0	June 15, 2009 19:17
Defect correction and convergence	ganesh	Main CFD Forum	4	June 30, 2006 15:20

June 10, 2014, 13:45		#2
hlk Senior Member Heather Kline Join Date: Jun 2013 Posts: 309 Rep Power: 14	Thank you for your question. The residuals are printed as the log of their value. This means that for a 6 order of magnitude decrease the number printed should decrease by 6. In your example, a six order of magnitude convergence would be indicated by 11.2 - 6 = 5.2. The number of iteration it takes to converge will depend heavily on the problem, the mesh, and the CFD settings chosen. Tools like multigrid and the CFL number can accelerate convergence.

June 10, 2014, 13:51		#3
hedley Member Hedley Join Date: May 2014 Posts: 52 Rep Power: 12	Thank you for this - I have been watching the output and 5000 iterations on my PC take approximately 3 hours and the residuals hardly move - This is on the sample wing SU2_CFD turb_RAM6.cfg has anyone run this on a PC so I know what to expect