Question on residual convergence

rks171 · January 4, 2012, 18:03

I ran a solution and monitored some physical parameters (mass flow rate at inlet and outlet and linear pressure drop) as well as the residuals. My continuity and z-momentum residuals only drop about an order of magnitude (the z-direction is the dominant flow direction), but they are normalized to the initial values, which were set pretty close to the converged values. So it makes sense to me if they don't drop that far from the beginning, however, they start to oscillate as the iterations continue on and I'm not sure if this should be expected or not. See the attached picture. Is this oscillation acceptable as long as my inlet/outlet mass balance is good and the pressure distribution makes sense?

rks171 · January 16, 2012, 17:31

Okay, so I inspected some additional physical parameters, namely average flow area velocity and it turns out that the solution wasn't converged. The area averaged velocity never settled on a number, but just kept changing through the iteration. I played with many variables for a while before I found something that allowed my solution to converge on realistic physical parameters AND for the residuals to look reasonable. My near wall prism layer was very thin in comparison to the 2nd prism layer and surrounding polyhedral cells, so I doubled its thickness. I also increased the gap fill percentage to 48% to try and get rid of some of the very small cells in the tight crevices of the models. Running the model without a prism layer exhibited ugly residuals, but using either the polyhedral mesher or trim mesher with the thicker near wall prism layer gave me better results. My residuals now look as follows:

January 4, 2012, 18:03	Question on residual convergence	#1
rks171 Member Join Date: Dec 2011 Location: State College, PA Posts: 87 Rep Power: 15	I ran a solution and monitored some physical parameters (mass flow rate at inlet and outlet and linear pressure drop) as well as the residuals. My continuity and z-momentum residuals only drop about an order of magnitude (the z-direction is the dominant flow direction), but they are normalized to the initial values, which were set pretty close to the converged values. So it makes sense to me if they don't drop that far from the beginning, however, they start to oscillate as the iterations continue on and I'm not sure if this should be expected or not. See the attached picture. Is this oscillation acceptable as long as my inlet/outlet mass balance is good and the pressure distribution makes sense?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
alphaEqn.H in twoPhaseEulerFoam	cheng1988sjtu	OpenFOAM Bugs	15	May 1, 2016 17:12
MRFSimpleFOAM goes divergenced!	renyun0511	OpenFOAM Running, Solving & CFD	0	November 19, 2009 03:11
SimpleFoam k and epsilon bounded	nedved	OpenFOAM Running, Solving & CFD	1	November 25, 2008 21:21
MRFSimpleFoam amp cyclic patches	david	OpenFOAM Running, Solving & CFD	36	October 21, 2008 22:55
IcoFoam parallel woes	msrinath80	OpenFOAM Running, Solving & CFD	9	July 22, 2007 03:58

January 16, 2012, 17:31		#2
rks171 Member Join Date: Dec 2011 Location: State College, PA Posts: 87 Rep Power: 15	Okay, so I inspected some additional physical parameters, namely average flow area velocity and it turns out that the solution wasn't converged. The area averaged velocity never settled on a number, but just kept changing through the iteration. I played with many variables for a while before I found something that allowed my solution to converge on realistic physical parameters AND for the residuals to look reasonable. My near wall prism layer was very thin in comparison to the 2nd prism layer and surrounding polyhedral cells, so I doubled its thickness. I also increased the gap fill percentage to 48% to try and get rid of some of the very small cells in the tight crevices of the models. Running the model without a prism layer exhibited ugly residuals, but using either the polyhedral mesher or trim mesher with the thicker near wall prism layer gave me better results. My residuals now look as follows: