Max Residual

az_f · September 9, 2011, 12:26

Hi
I have a question about Max residual and RMS. Is it necessary that Max residual is more accurate comparing to the RMS.When I used RMS in my model I've got a smooth convergence but when I changed to Max I can see the convergence with fluctuation until 10-5 and then one flat line for 10-6. Does that mean the RMS didn't give me the real convergence or MAX is too restrict.There is 0.1 % difference between the energy calculated from both options, but I'm not sure about the convergence?? Could anyone explain it to me?

Thanks

Biskvito · September 9, 2011, 14:42

hi az_f

from what I know we have a matrix of residuals so that's a set of numbers where RMS is a quadratic mean of all of them and MAX is of course the largest of them. This would mean that reqirement on the MAX residual is always more restrict. But 0.1% of difference is in my opinion very small as I've heard that if the model differs from the real object by up to 5% we consider it as very good.
The RMS is more smooth because it's a mean value so it wouldn't differ very much in each timestep.

...but I'm just a student so I might be wrong

regards

ghorrocks · September 12, 2011, 20:31

The residual is a scalar field variable, and RMS and MAX is two different ways of reducing the variable field to a single number. Obviously RMS gives you an average idea of convergence but MAX is dependant on the worst element. Which one is more appropriate depends on what you are doing.

But for most people the convergence you talk about is tighter than is required. Do a mesh sensitivity check to work out how tight you really need to solve it.

September 9, 2011, 12:26	Max Residual	#1
az_f Member Join Date: Mar 2009 Posts: 48 Rep Power: 17	Hi I have a question about Max residual and RMS. Is it necessary that Max residual is more accurate comparing to the RMS.When I used RMS in my model I've got a smooth convergence but when I changed to Max I can see the convergence with fluctuation until 10-5 and then one flat line for 10-6. Does that mean the RMS didn't give me the real convergence or MAX is too restrict.There is 0.1 % difference between the energy calculated from both options, but I'm not sure about the convergence?? Could anyone explain it to me? Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Extrusion with OpenFoam problem No. Iterations 0	Lord Kelvin	OpenFOAM Running, Solving & CFD	8	March 28, 2016 12:08
How to write k and epsilon before the abnormal end	xiuying	OpenFOAM Running, Solving & CFD	8	August 27, 2013 16:33
Orifice Plate with a fully developed flow - Problems with convergence	jonmec	OpenFOAM Running, Solving & CFD	3	July 28, 2011 06:24
Differences between serial and parallel runs	carsten	OpenFOAM Bugs	11	September 12, 2008 12:16
Could anybody help me see this error and give help	liugx212	OpenFOAM Running, Solving & CFD	3	January 4, 2006 19:07

September 9, 2011, 14:42		#2
Biskvito New Member Join Date: Aug 2011 Location: Cracow, Poland Posts: 7 Rep Power: 15	hi az_f from what I know we have a matrix of residuals so that's a set of numbers where RMS is a quadratic mean of all of them and MAX is of course the largest of them. This would mean that reqirement on the MAX residual is always more restrict. But 0.1% of difference is in my opinion very small as I've heard that if the model differs from the real object by up to 5% we consider it as very good. The RMS is more smooth because it's a mean value so it wouldn't differ very much in each timestep. ...but I'm just a student so I might be wrong regards

September 12, 2011, 20:31		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,852 Rep Power: 144	The residual is a scalar field variable, and RMS and MAX is two different ways of reducing the variable field to a single number. Obviously RMS gives you an average idea of convergence but MAX is dependant on the worst element. Which one is more appropriate depends on what you are doing. But for most people the convergence you talk about is tighter than is required. Do a mesh sensitivity check to work out how tight you really need to solve it.