CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > FLUENT

Divergence Problem

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   May 25, 2013, 14:22
Default Divergence Problem
  #1
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
I am experiencing some divergence problems with my residuals using Fluent. I am using the k-omega turbulence model running under steady state conditions.
I refined my mesh as best as I can. The qualities (orthogonal qulity, aspect ratio, skewness, etc...) are around suggested values, but for some reason my residuals still diverge. I do not think this is a mesh quality problem.

Also my solution actually reaches the "residuls have converged" pop-up when i run the case, but if I turn convergence criterion off, it would converge then diverge.

I read that I could change the relaxation factors lower, but isn't that kind of cheating the system? My experience with lowering the factors is that it only "delays" the divergence, but not "eliminate" the divergence.

Any help would be greatly appreciated.


P.S I am pretty positive my setup is correct. In solution methods, I am mainly using second order solvers, except for the dissipation rate solver (using 1st order for this one).
pyroknife is offline   Reply With Quote

Old   May 25, 2013, 14:39
Default
  #2
Member
 
misagh
Join Date: Apr 2012
Posts: 63
Rep Power: 6
misagh is on a distinguished road
the convergence pattern must be steady.It may reach the criteria you want to ,but it shouldn't oscillate.
By the way lowering the factors is not cheating.It is factors to ease the equation solution process.And in some special cases(that actually i am not familiar with those ones) you need to increase the factors.
bests
misagh is offline   Reply With Quote

Old   May 25, 2013, 15:36
Default
  #3
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
Quote:
Originally Posted by misagh View Post
the convergence pattern must be steady.It may reach the criteria you want to ,but it shouldn't oscillate.
By the way lowering the factors is not cheating.It is factors to ease the equation solution process.And in some special cases(that actually i am not familiar with those ones) you need to increase the factors.
bests
I have tried lowering the factors, but all it seems to do is delay the divergence, not actually eliminate it from my simulation.
pyroknife is offline   Reply With Quote

Old   May 26, 2013, 00:26
Default
  #4
Member
 
刘峰
Join Date: Mar 2013
Posts: 31
Rep Power: 5
Liufeng_ustb is on a distinguished road
I also wonder if lowering the re-factors can help eliminate divergence...

Besides, pyroknife, I want to know whether your simulation involves gas mixture? If you do, what's your option about settong the properties of gas mixture.

Regards
Liufeng_ustb is offline   Reply With Quote

Old   May 26, 2013, 15:30
Default
  #5
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
Quote:
Originally Posted by Liufeng_ustb View Post
I also wonder if lowering the re-factors can help eliminate divergence...

Besides, pyroknife, I want to know whether your simulation involves gas mixture? If you do, what's your option about settong the properties of gas mixture.

Regards
My simulation is currently involving methane (the one FLUENT has in it's material library). Does that count?
I remember using incompressible-ideal gas, polynomial, and 2 kinetic theory options.
pyroknife is offline   Reply With Quote

Old   May 26, 2013, 17:04
Default
  #6
Senior Member
 
Lefteris
Join Date: Oct 2011
Location: UK
Posts: 199
Rep Power: 6
Aeronautics El. K. is on a distinguished road
k-omega turbulence model is sensitive and it depends greatly on the initial values of k and omega. If the values of these variables are not ok, the solution will most probably diverge. Try the SST k-omega.
As for the underrelaxation factors of course it's not cheating! Don't forget that turbulence model's equations are stiff and make convergence even more difficult...

EDIT
-----
The free stream values of the turbulence variables I meant to say, not the initial values of k and omega. My bad, sorry. (http://www.cfd-online.com/Wiki/SST_k-omega_model)
__________________
Lefteris


Last edited by Aeronautics El. K.; May 28, 2013 at 13:31.
Aeronautics El. K. is offline   Reply With Quote

Old   May 26, 2013, 18:38
Default
  #7
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
Quote:
Originally Posted by Aeronautics El. K. View Post
k-omega turbulence models is sensitive and it depends greatly on the initial values of k and omega. If the initials values of these variables are wronge, the solution will most probably diverge. Try the SST k-omega.
As for the underrelaxation factors of course it's not cheating! Don't forget that turbulence model's equations are stiff and make convergence even more difficult...
Oh I see. Can you set the initial values of k and omega? If so, how?
I am currently using SST k-omega.
pyroknife is offline   Reply With Quote

Old   May 27, 2013, 00:12
Default
  #8
Member
 
刘峰
Join Date: Mar 2013
Posts: 31
Rep Power: 5
Liufeng_ustb is on a distinguished road
Quote:
Originally Posted by pyroknife View Post
My simulation is currently involving methane (the one FLUENT has in it's material library). Does that count?
I remember using incompressible-ideal gas, polynomial, and 2 kinetic theory options.
Hi, friend.

I once experienced a divergence problem caused by the selection of the properties of mixture gas.

My mixture gas is composed of two species which are hydrogen and water vapor. Initially, I selected ideal-gas-mixing-law, then when I ran, divergence problem occured. But when I revised it using mass-averaged, this problem disapperaed.

Your simulation is a single gas species. But you can try to modify ideal gas option.

Wish you luck!
Liufeng_ustb is offline   Reply With Quote

Old   May 27, 2013, 03:01
Default
  #9
New Member
 
Muhammad Aqib Chishty
Join Date: May 2013
Location: Islamabad, Pakistan
Posts: 23
Rep Power: 0
MAC88 is on a distinguished road
Send a message via Skype™ to MAC88
As you are using k-omega turbulence model with under steady state conditions, it is sensitive to free stream conditions.
What time step size you are using?
Try to reduce the time step size.

Wish you best of luck!
MAC88 is offline   Reply With Quote

Old   May 27, 2013, 16:57
Default
  #10
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
Quote:
Originally Posted by MAC88 View Post
As you are using k-omega turbulence model with under steady state conditions, it is sensitive to free stream conditions.
What time step size you are using?
Try to reduce the time step size.

Wish you best of luck!
I am running under steady state. Isn't time step only for unsteady?




Also, it appears that only my energy equation is experiencing divergence.
pyroknife is offline   Reply With Quote

Old   May 28, 2013, 00:09
Default
  #11
New Member
 
Muhammad Aqib Chishty
Join Date: May 2013
Location: Islamabad, Pakistan
Posts: 23
Rep Power: 0
MAC88 is on a distinguished road
Send a message via Skype™ to MAC88
Quote:
Originally Posted by pyroknife View Post
I am running under steady state. Isn't time step only for unsteady?




Also, it appears that only my energy equation is experiencing divergence.
I think that was my typing mistake.
You have to run your simulation in an unsteady mode.
Have you tried that?
MAC88 is offline   Reply With Quote

Old   May 28, 2013, 03:31
Default
  #12
Member
 
Join Date: Dec 2009
Location: China
Posts: 79
Rep Power: 8
ringtail is on a distinguished road
Quote:
Originally Posted by pyroknife View Post
I am experiencing some divergence problems with my residuals using Fluent. I am using the k-omega turbulence model running under steady state conditions.
I refined my mesh as best as I can. The qualities (orthogonal qulity, aspect ratio, skewness, etc...) are around suggested values, but for some reason my residuals still diverge. I do not think this is a mesh quality problem.

Also my solution actually reaches the "residuls have converged" pop-up when i run the case, but if I turn convergence criterion off, it would converge then diverge.

I read that I could change the relaxation factors lower, but isn't that kind of cheating the system? My experience with lowering the factors is that it only "delays" the divergence, but not "eliminate" the divergence.

Any help would be greatly appreciated.


P.S I am pretty positive my setup is correct. In solution methods, I am mainly using second order solvers, except for the dissipation rate solver (using 1st order for this one).
Many factors could cause instability problems. I guess you should show the history of residual as well as Cd and so on to provide more detail information.

generally, since you use the k-omega model, so the y+ should be about 1.
And, be careful, a too fine mesh for RANS simulation may also cause divergence problems for fine mesh may capture more instable flow phenomena.

What's more, for high order discretization scheme, lower relaxation factors is necessary, especially for pressure (set to 0.1 maybe, it depends on your p-v coupling scheme). You can also enable High order Term Relaxation, and have a try, it may works.

Please correct me if I am wrong.
ringtail is offline   Reply With Quote

Old   May 28, 2013, 10:26
Default
  #13
New Member
 
Yasir hayat
Join Date: May 2013
Posts: 4
Rep Power: 5
yasir_hayyat is on a distinguished road
Hi all . I am facing same problem. I want to simulate cryogenic boiling in fluid film bearings. I am using k omega sst model , with multi-phase mixture model but some where in the iteration my solution diverges and i get floating point error. i have noted that when i use energy equation it diverges and i have noticed the residua graph of temperature shoots up and i get floating point error. Can any one guide me why this is happening and any possible approach to tackle this problem.

Regards

Yasir hayat
yasir_hayyat is offline   Reply With Quote

Old   May 28, 2013, 12:24
Default
  #14
Member
 
Join Date: Jan 2013
Posts: 39
Rep Power: 5
pyroknife is on a distinguished road
Quote:
Originally Posted by MAC88 View Post
I think that was my typing mistake.
You have to run your simulation in an unsteady mode.
Have you tried that?
I have yet to try unsteady. My simulation is a steady case, however.

Quote:
Originally Posted by ringtail View Post
Many factors could cause instability problems. I guess you should show the history of residual as well as Cd and so on to provide more detail information.

generally, since you use the k-omega model, so the y+ should be about 1.
And, be careful, a too fine mesh for RANS simulation may also cause divergence problems for fine mesh may capture more instable flow phenomena.

What's more, for high order discretization scheme, lower relaxation factors is necessary, especially for pressure (set to 0.1 maybe, it depends on your p-v coupling scheme). You can also enable High order Term Relaxation, and have a try, it may works.

Please correct me if I am wrong.
Thanks. That is (too fine of a mesh) what I think may be causing the problem. I did not save the data, but if i experience the divergence again, i'll post it on here. I am basically simulating pipe flow, but a bit more complicated than the normal 1 inlet/outlet, pipe flow problem. I ran a successful simulation with .4million elements using cutcell mesh. However, when I increased the length of my pipe, I used a 1.2million cutcell mesh and for some reason this failed. FYI, the .4 million element mesh had WORSE quality than the 1.2 million one, so I don't think the divergence is caused by poor element quality.
pyroknife is offline   Reply With Quote

Old   May 28, 2013, 13:31
Default
  #15
New Member
 
Muhammad Aqib Chishty
Join Date: May 2013
Location: Islamabad, Pakistan
Posts: 23
Rep Power: 0
MAC88 is on a distinguished road
Send a message via Skype™ to MAC88
Quote:
Originally Posted by yasir_hayyat View Post
Hi all . I am facing same problem. I want to simulate cryogenic boiling in fluid film bearings. I am using k omega sst model , with multi-phase mixture model but some where in the iteration my solution diverges and i get floating point error. i have noted that when i use energy equation it diverges and i have noticed the residua graph of temperature shoots up and i get floating point error. Can any one guide me why this is happening and any possible approach to tackle this problem.

Regards

Yasir hayat
Reduce the under relaxation parameters.

Best of luck!!!
MAC88 is offline   Reply With Quote

Old   May 28, 2013, 22:26
Default
  #16
Member
 
Join Date: Dec 2009
Location: China
Posts: 79
Rep Power: 8
ringtail is on a distinguished road
Hi, pyroknife
More details is needed to solve your problem.
Such as Y+ for the all computation domain, the discretization scheme, your PC setting and so on.
If it is possible, post the geometry even the case file will be greatly helpful.

Regards
ringtail is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Divergence problem Smaras FLUENT 13 February 21, 2013 06:03
Meshing and divergence problem ghost FLUENT 9 February 5, 2010 13:24
divergence problem vincent FLUENT 3 August 3, 2006 15:44
strange divergence when solving multiphase problem tanghao FLUENT 2 July 27, 2006 19:47
divergence problem Ayyappan.T FLUENT 2 May 16, 2005 12:10


All times are GMT -4. The time now is 08:09.