Additional PIMPLE convergence criteria

amanbearpig · August 8, 2019, 10:32

I'm working on a modified pimpleFoam type solver, and I'd like the outer pimple loop to check an additional convergence criteria (forces on body). Is there an easy way to do this within the standard pimple solution control? I can check my forces convergence manually and then do an if … break;, but this breaks the pimple loop in a non-satisfactory manner.

Anyone have any thoughts? I have to imagine there is a (hopefully simple) solution I have overlooked, that doesn't involve messing with the actual pimpleControl files.

jcappola · August 8, 2019, 12:43

pimpleControl only checks residual convergence in order to determine when to exit the loop (check out the criteriaSatisfied() function in the pimpleControl.C). I don't think there is a clean or "correct" way to add what you are asking about.

It seems the problem is chasing itself, no? If you make the assumption that your case is well-behaved and solving correctly, the proper convergence of PIMPLE should (IMO) be related to proper convergence of your forcing. What is the motivation behind this additional criterion?

amanbearpig · August 8, 2019, 14:38

Well, it's for a coupled 6 DoF dynamic mesh solver. I want the body to move within the pimple loop. I keep track of the residual based on the change in body forces, then I'd like to cleanly exit the pimple loop once that's converged. If I can do this within the pimple loop, I can easily keep the pimple advantages such as under-relaxation for non-finalIter().

jcappola · August 8, 2019, 16:19

Coupled in what way? FSI? Physical time doesn't advance while PIMPLE is looping over a time-step. Intentionally trying to move your mesh per PIMPLE Loop makes no physical sense.

If you could provide some more context of your case we can move towards finding a solution to the problem.

amanbearpig · August 13, 2019, 07:53

Within the outer time loop I have an inner pimple loop. Each pimple iteration I calculate forces on my body. I move the body based on these forces (taking previous movement within the current time step into account). The change in forces calculated on the body is my residual that I'm wanting to account for in the pimple loop.

I've made a modified pimpleControl which I am using at the moment, but for simplicities sake I'd prefer to not have to use a modified file.

jcappola · August 14, 2019, 15:35

This sounds very similar to the sub-iteration method taken in the FSI solver present in the solidMechanics folder of the foam-extend branch. May be worth looking into to see if you can't adapt any code from that.

August 8, 2019, 10:32	Additional PIMPLE convergence criteria	#1
amanbearpig Member Join Date: Apr 2011 Posts: 57 Rep Power: 15	I'm working on a modified pimpleFoam type solver, and I'd like the outer pimple loop to check an additional convergence criteria (forces on body). Is there an easy way to do this within the standard pimple solution control? I can check my forces convergence manually and then do an if … break;, but this breaks the pimple loop in a non-satisfactory manner. Anyone have any thoughts? I have to imagine there is a (hopefully simple) solution I have overlooked, that doesn't involve messing with the actual pimpleControl files.

August 8, 2019, 16:19		#4
jcappola New Member Jonthan Cappola Join Date: Apr 2018 Posts: 8 Rep Power: 8	Coupled in what way? FSI? Physical time doesn't advance while PIMPLE is looping over a time-step. Intentionally trying to move your mesh per PIMPLE Loop makes no physical sense. If you could provide some more context of your case we can move towards finding a solution to the problem. Last edited by jcappola; August 8, 2019 at 16:20. Reason: Emphasis

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence	Centurion2011	FLUENT	48	June 14, 2022 23:29
Residual transient case, convergence criteria	Ursmooth	Main CFD Forum	1	September 30, 2017 03:18
Convergence Criteria in FLUENT	mahi007	FLUENT	9	March 1, 2012 06:46
CAVITATION (convergence criteria & discretization)	ROOZBEH	FLUENT	1	October 6, 2003 09:32
Replace periodic by inlet-outlet pair	lego	CFX	3	November 5, 2002 20:09

August 8, 2019, 12:43		#2
jcappola New Member Jonthan Cappola Join Date: Apr 2018 Posts: 8 Rep Power: 8	pimpleControl only checks residual convergence in order to determine when to exit the loop (check out the criteriaSatisfied() function in the pimpleControl.C). I don't think there is a clean or "correct" way to add what you are asking about. It seems the problem is chasing itself, no? If you make the assumption that your case is well-behaved and solving correctly, the proper convergence of PIMPLE should (IMO) be related to proper convergence of your forcing. What is the motivation behind this additional criterion?

August 8, 2019, 14:38		#3
amanbearpig Member Join Date: Apr 2011 Posts: 57 Rep Power: 15	Well, it's for a coupled 6 DoF dynamic mesh solver. I want the body to move within the pimple loop. I keep track of the residual based on the change in body forces, then I'd like to cleanly exit the pimple loop once that's converged. If I can do this within the pimple loop, I can easily keep the pimple advantages such as under-relaxation for non-finalIter().

August 13, 2019, 07:53		#5
amanbearpig Member Join Date: Apr 2011 Posts: 57 Rep Power: 15	Within the outer time loop I have an inner pimple loop. Each pimple iteration I calculate forces on my body. I move the body based on these forces (taking previous movement within the current time step into account). The change in forces calculated on the body is my residual that I'm wanting to account for in the pimple loop. I've made a modified pimpleControl which I am using at the moment, but for simplicities sake I'd prefer to not have to use a modified file.

August 14, 2019, 15:35		#6
jcappola New Member Jonthan Cappola Join Date: Apr 2018 Posts: 8 Rep Power: 8	This sounds very similar to the sub-iteration method taken in the FSI solver present in the solidMechanics folder of the foam-extend branch. May be worth looking into to see if you can't adapt any code from that.