|
[Sponsors] |
April 20, 2015, 01:53 |
converge problem
|
#1 |
Member
cglr
Join Date: Apr 2015
Posts: 31
Rep Power: 11 |
Hello,
I have been experiencing a heat transfer analysis. (blowing air(4m/s) between two hot parallel plates) I set gravity at 9.81 and made density boussinesq using k-e standart turbulance model, under rel. factor for energy 0.9 and for pressure I used presto and calculated for 1000 iterations. Although I have a good mesh quality, I had a bad converging(there are many fluctuating as you can see in the attached document) after 400th iteration and at the finish. But until the 400 th iteration I have good converging. What may the possible causes of this be? Do you have any advice to do ? Or, is this attached figure normal? |
|
April 20, 2015, 04:48 |
|
#2 |
Senior Member
Cees Haringa
Join Date: May 2013
Location: Delft
Posts: 607
Rep Power: 0 |
Don't judge convergence by your residuals. They're an indicator whether things are going in the right direction, but not a way to determine whether your solution is good enough. In some cases, residuals of 1e-3 may be more than sufficient, in others e-6 is not good enough. Check whether the quantities you are interested in reach a constant value. In your case: mean velocity, mean temperature, mean surface nusselt number maybe...
|
|
April 20, 2015, 05:59 |
|
#3 |
Member
cglr
Join Date: Apr 2015
Posts: 31
Rep Power: 11 |
thanks for your reply. But what exactly do you mean by right direction?
|
|
April 20, 2015, 07:03 |
|
#4 |
Senior Member
Cees Haringa
Join Date: May 2013
Location: Delft
Posts: 607
Rep Power: 0 |
If the residuals are going up, it means the solution is a progressively worse approximation of the true solution. Occasionally residuals may go up without problem, but on the long term they should always go down - otherwise your solution diverges and eventually may crash.
If your residuals stick at a certain value, it means the solver cannot get closer to the true solution given the settings. It could be the solution at that point is already satisfactory accurate - but that you must judge on other (integral) parameters as the ones named above. If the solution is not accurate enough yet, you can reduce the under-relaxation parameters. This may break the 'dancing' around a certain solution that you pbserve. |
|
April 21, 2015, 07:22 |
bad convergence
|
#5 |
New Member
rim
Join Date: Feb 2015
Posts: 2
Rep Power: 0 |
I'm a beginner in fluent and i'am trying to model a 2d version of squeeze flow between two disks. I made a simple rectangle(1x20) for the geometry in the xy axis and for the boundary conditions I specified the top as a outflow moving, the bottom as an axis , my left as symmetry and the right boundary conditions as velocity inlet (0.01 m/s).First of all I want to know if my geometry is convienient with my model ? Second in the simulation when i put a fine meshing i found a bad results and when i put a large mesh i found a good results and these are illogical .If anyone has any suggestions it would be greatly appreciated.
Thanks all. |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
converge problem | mahdidelfan | FLUENT | 0 | July 13, 2014 01:26 |
Problem with Residual plot raising up to hige numbers na do not converge | fardinkhalili | STAR-CCM+ | 4 | March 25, 2014 09:35 |
Velocity profiles problem behind the elbow (3D problem) | kabat73 | FLUENT | 8 | May 9, 2010 04:26 |
converge problem | snowshovel | CFX | 13 | January 18, 2007 21:05 |
Converge problem of 3D free jet flow | Jen | FLUENT | 11 | January 24, 2005 00:21 |