Convergence Issue

zephyrus17 · August 21, 2011, 10:36

I'm doing a Steady State age of air validation case where air @ 25C enters a simple room through a square inlet and exits from an exit across the room. The air enters at 1.68m/s and the outlet is just a 0 Pa outlet. I'm monitoring the age of air and also monitoring the turbulence kinetic energy (is this the most suitable?) to judge convergence. I'm following the journal and using the standard k-e turbulence model with 14% turbulence intensity.

Here is the journal article for it: http://www.inive.org/members_area/me...%5CUFSC492.pdf

The coarse mesh would converge with no problems, however the fine mesh would not. The advection time is around 17 seconds. I tried setting the local timestep to 5, at first; tried playing with the physical timestep, varying from 200 down to 0.17 seconds. No matter what I did, the converge would drop swifty to 1e-3, then just bounce around there. The lowest I got it down to was around 8e-4.

There was nothing wrong with the mesh itself as it showed no issues in the solver .out file.

Any suggestions on what to do next or what I'm doing wrong?

lindner · August 22, 2011, 09:56

Hi,
You can check the values of y+ for the new mesh, it should be around 30 and 300 for the k-epsilon turbulent model.
Look at the streamlines and search for recirculation in the room, maybe you need to change to SST model to capture that right (and refine the mesh to have y+ around 1).

regards,
lindner

zephyrus17 · August 22, 2011, 10:07

The y+ is at 111.7 at it's maximum. And sadly I can't change the turbulence model because I need to follow the journal article to validate it. And I can't change the mesh as well due to the same reasons.

So in other words, given the current situation, I can't change anything else other than the timestep, solver controls, etc to get it to converge down.

lindner · August 22, 2011, 10:19

Try using the Production Limiter Kato Launder on the Advanced Turbulence Control inside Fluid Models (right under Wall Function).

zephyrus17 · August 22, 2011, 10:35

Doesn't work. It stopped the 'bouncing' of the converge around the 1e-3 mark and spread it out, which looks good, but it's risen to around 1.6e-3. And the bounces are now merely more spread out.

lindner · August 22, 2011, 12:37

Did you compare the results even with the non-converged solution? I did this Bartak case last year using OpenFOAM and got good results, but don't remember having issues with convergence. I will check my setup later and try to find something useful.

zephyrus17 · August 22, 2011, 18:30

I did. The results are pretty much the same. The Age of Air monitoring point can be assumed to be stabilised. However, this is a pretty good case in study of how to get something to stabilise for me

swiss_zhang · August 24, 2011, 04:28

Following Ideal:

1: How many iteration steps have you set? Increase the iteration step and try it.

2: use Adaptive Time Step

ghorrocks · August 24, 2011, 06:24

There is a FAQ which will help here, and it goes into much more detail:

http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Note that divergence when going to finer meshes is very common. It usually means the reduced dissipation of the finer mesh is causing the flow to become transient. Try the tricks on the FAQ, and if that does not work you will have to run it transient.

August 21, 2011, 10:36	Convergence Issue	#1
zephyrus17 New Member Join Date: Sep 2010 Posts: 28 Rep Power: 0	I'm doing a Steady State age of air validation case where air @ 25C enters a simple room through a square inlet and exits from an exit across the room. The air enters at 1.68m/s and the outlet is just a 0 Pa outlet. I'm monitoring the age of air and also monitoring the turbulence kinetic energy (is this the most suitable?) to judge convergence. I'm following the journal and using the standard k-e turbulence model with 14% turbulence intensity. Here is the journal article for it: http://www.inive.org/members_area/me...%5CUFSC492.pdf The coarse mesh would converge with no problems, however the fine mesh would not. The advection time is around 17 seconds. I tried setting the local timestep to 5, at first; tried playing with the physical timestep, varying from 200 down to 0.17 seconds. No matter what I did, the converge would drop swifty to 1e-3, then just bounce around there. The lowest I got it down to was around 8e-4. There was nothing wrong with the mesh itself as it showed no issues in the solver .out file. Any suggestions on what to do next or what I'm doing wrong?

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
Convergence	Centurion2011	FLUENT	48	June 14, 2022 23:29
Convergence issue with continuity equation	Jake	FLUENT	8	June 6, 2018 03:41
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 01:17
Convergence issue in SST for Porous model	Raj	CFX	0	May 2, 2008 02:43
CFX-Solver, issue with convergence behavior	Andy	CFX	7	September 5, 2006 03:24

August 22, 2011, 09:56		#2
lindner Member Join Date: Mar 2010 Posts: 42 Rep Power: 16	Hi, You can check the values of y+ for the new mesh, it should be around 30 and 300 for the k-epsilon turbulent model. Look at the streamlines and search for recirculation in the room, maybe you need to change to SST model to capture that right (and refine the mesh to have y+ around 1). regards, lindner

August 22, 2011, 10:07		#3
zephyrus17 New Member Join Date: Sep 2010 Posts: 28 Rep Power: 0	The y+ is at 111.7 at it's maximum. And sadly I can't change the turbulence model because I need to follow the journal article to validate it. And I can't change the mesh as well due to the same reasons. So in other words, given the current situation, I can't change anything else other than the timestep, solver controls, etc to get it to converge down.

August 22, 2011, 10:19		#4
lindner Member Join Date: Mar 2010 Posts: 42 Rep Power: 16	Try using the Production Limiter Kato Launder on the Advanced Turbulence Control inside Fluid Models (right under Wall Function).

August 22, 2011, 10:35		#5
zephyrus17 New Member Join Date: Sep 2010 Posts: 28 Rep Power: 0	Doesn't work. It stopped the 'bouncing' of the converge around the 1e-3 mark and spread it out, which looks good, but it's risen to around 1.6e-3. And the bounces are now merely more spread out.

August 22, 2011, 12:37		#6
lindner Member Join Date: Mar 2010 Posts: 42 Rep Power: 16	Did you compare the results even with the non-converged solution? I did this Bartak case last year using OpenFOAM and got good results, but don't remember having issues with convergence. I will check my setup later and try to find something useful.

August 22, 2011, 18:30		#7
zephyrus17 New Member Join Date: Sep 2010 Posts: 28 Rep Power: 0	I did. The results are pretty much the same. The Age of Air monitoring point can be assumed to be stabilised. However, this is a pretty good case in study of how to get something to stabilise for me

August 24, 2011, 04:28		#8
swiss_zhang New Member Zhang Yang Join Date: Jun 2011 Location: Zürich Posts: 28 Rep Power: 15	Following Ideal: 1: How many iteration steps have you set? Increase the iteration step and try it. 2: use Adaptive Time Step

August 24, 2011, 06:24		#9
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,690 Rep Power: 143	There is a FAQ which will help here, and it goes into much more detail: http://www.cfd-online.com/Wiki/Ansys...gence_criteria Note that divergence when going to finer meshes is very common. It usually means the reduced dissipation of the finer mesh is causing the flow to become transient. Try the tricks on the FAQ, and if that does not work you will have to run it transient.