Residual plot for LES Case

AS_Aero · April 11, 2017, 12:25

Hi All

I am running an LES case and I have my residuals 1e-7 as the RMS value and that was reached while I did a RANS itself and now I used this RANS result as initialization and then did the LES and now I have my Residuals like shown in the image, could someone tell me if its wrong or is it normal ?
Thanks in advance.

ghorrocks · April 11, 2017, 19:24

You have reached the limit of numerical precision on that simulation. It will not converge further until you improve the numerical precision.

But note:
1) That level of residual is converged for most cases. You probably don't need to go tighter.
2) This appears to be a steady state simulation, but you mentioned you are doing LES. You cannot do steady state LES, this is meaningless.

AS_Aero · April 12, 2017, 05:55

There are two things !! My residual value I gave was 1e-6 as RMS value. I started from a RANS solution where the residual value was reached and the simulation stopped. Then I used that .res file to run the LES which is transient ofcourse and my monitor point values are not stable or converged so am still running them. I am not sure what would have gone wrong. Or what should I do ?

ghorrocks · April 12, 2017, 08:11

LES does not converge by definition. It models the large turbulence structures so will always have random fluctuations. You need to run it long enough such that an average will be representative of the flow.

If that is the residuals for a transient simulation then you probably have not set it up correctly. If you use adaptive time stepping homing in on 3-5 coeff loops per iteration the solver will find a time step for you. Also note that residual tolerances for transient runs can often be a little looser than steady state runs and still obtain good convergence.

AS_Aero · April 12, 2017, 09:38

Hi Glenn

So I gave adaptive time stepping !!
Kindly have a look on my settings I have given on CFX and tell me if I did something wrong !!

My simulation is running as of now and the contours plots and the profiles on the flow seems to be ok and comparable to the RANS except the level and the trend seems to be normal.

ghorrocks · April 12, 2017, 18:45

You are using courant number. I recommend number of coefficient loops, homing in on 3-5 coeff loops.

Your convergence tolerance is tight for a transient run. Does it need to be that tight? Do a sensitivity study to find out.

I would make the max number of coeff loops 10.

AS_Aero · April 13, 2017, 04:50

Yes Courant number have to be 1 or less than one right (Theoretically) ? And Coefficient loops I just gave it randomly. And convergence tolerance I just gave according to the best practices guide. And what do you mean by tight ? As of now my case is running without any issue except that my RMS residual looks like the one on the picture.

ghorrocks · April 13, 2017, 19:05

What theory says the Courant number has to be less than 1? There is none. CFX is an implicit solver and is not restrained by Courant number. Explicit solvers are limited by Courant number - but CFX is not an explicit solver. This is a very common misunderstanding.

AS_Aero · April 13, 2017, 19:17

In general CFL has to be 1 or less than one right atleast according to some books. Regarding CFX so what can I give ? Adaptive time stepping of some timestep with 1e-6 or -5 ? And do I have to consider the courent number limit ? If CFX doesnt matter then why is the option for courent number minimum and maximum in CFX

ghorrocks · April 13, 2017, 19:26

Read my previous post and read the "some books" again, carefully. It does not apply to CFX as CFX is an implicit solver, not explicit.

I recommend adaptive time stepping, homing in on 3-5 coefficient loops per iteration. That works very well for CFX in most cases.

CFX has many options in many menus which are not recommended in many cases. There are a small number of cases where they are used and that is why they are present.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
chtMultiRegionSimpleFoam: maximum number of iterations excedeed.	Nkl	OpenFOAM Running, Solving & CFD	19	October 10, 2019 02:42
High Courant Number @ icoFoam	Artex85	OpenFOAM Running, Solving & CFD	11	February 16, 2017 13:40
HeatSource BC to the whole region in chtMultiRegionHeater	xsa	OpenFOAM Running, Solving & CFD	3	November 7, 2016 05:07
How to write k and epsilon before the abnormal end	xiuying	OpenFOAM Running, Solving & CFD	8	August 27, 2013 15:33
Unknown error	sivakumar	OpenFOAM Pre-Processing	9	September 9, 2008 12:53

April 11, 2017, 19:24		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,694 Rep Power: 143	You have reached the limit of numerical precision on that simulation. It will not converge further until you improve the numerical precision. But note: 1) That level of residual is converged for most cases. You probably don't need to go tighter. 2) This appears to be a steady state simulation, but you mentioned you are doing LES. You cannot do steady state LES, this is meaningless.

April 12, 2017, 08:11		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,694 Rep Power: 143	LES does not converge by definition. It models the large turbulence structures so will always have random fluctuations. You need to run it long enough such that an average will be representative of the flow. If that is the residuals for a transient simulation then you probably have not set it up correctly. If you use adaptive time stepping homing in on 3-5 coeff loops per iteration the solver will find a time step for you. Also note that residual tolerances for transient runs can often be a little looser than steady state runs and still obtain good convergence.

April 12, 2017, 18:45		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,694 Rep Power: 143	You are using courant number. I recommend number of coefficient loops, homing in on 3-5 coeff loops. Your convergence tolerance is tight for a transient run. Does it need to be that tight? Do a sensitivity study to find out. I would make the max number of coeff loops 10.

April 13, 2017, 04:50		#7
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	Yes Courant number have to be 1 or less than one right (Theoretically) ? And Coefficient loops I just gave it randomly. And convergence tolerance I just gave according to the best practices guide. And what do you mean by tight ? As of now my case is running without any issue except that my RMS residual looks like the one on the picture.

April 13, 2017, 19:05		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,694 Rep Power: 143	What theory says the Courant number has to be less than 1? There is none. CFX is an implicit solver and is not restrained by Courant number. Explicit solvers are limited by Courant number - but CFX is not an explicit solver. This is a very common misunderstanding.

April 13, 2017, 19:17		#9
AS_Aero Senior Member Join Date: Mar 2013 Location: Germany Posts: 357 Rep Power: 14	In general CFL has to be 1 or less than one right atleast according to some books. Regarding CFX so what can I give ? Adaptive time stepping of some timestep with 1e-6 or -5 ? And do I have to consider the courent number limit ? If CFX doesnt matter then why is the option for courent number minimum and maximum in CFX

April 13, 2017, 19:26		#10
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,694 Rep Power: 143	Read my previous post and read the "some books" again, carefully. It does not apply to CFX as CFX is an implicit solver, not explicit. I recommend adaptive time stepping, homing in on 3-5 coefficient loops per iteration. That works very well for CFX in most cases. CFX has many options in many menus which are not recommended in many cases. There are a small number of cases where they are used and that is why they are present.