|
[Sponsors] |
pimpleDyMFoam for real industrial case taking too long without converging |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
March 23, 2016, 00:25 |
pimpleDyMFoam for real industrial case taking too long without converging
|
#1 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Hello All,
I'm running pimpleDyMFoam on an industrial case fan system. But the time step is very low (10e-7) at the beginning but it increases to 10e-6 later. The problem is that for almost 3 weeks now the case doesn't converge after many revolutions. I tried to put higher the maxco to increase the time step but no real effect observed. The mesh is about 6 millions cells. Knowing that start the pimpleDyMFoam running using the steady state (simpleFoam-MRF) to have a jump-start. My question am I doing something wrong ? And How can I get a good options for reasonable time computation? Attached the controDict, fvSolution and fvScheme. Thanks in advance. Kossivi controlDict.txt fvSchemes.txt fvSolution.txt |
|
March 28, 2016, 09:50 |
|
#3 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Hello kenyoh,
Thanks for reply. p iterations become 3 or 2 iterations. See attached a part of the log file (too heavy to upload). Thanks, Kossivi log1.txt
__________________
Thanks in advance, Kossivi. |
|
March 28, 2016, 20:53 |
|
#4 |
New Member
|
Hi Kossivi
It seems to be very stable caluclation. I guess there is no way except finding mesh where Co is high and re-meshing there. Or, increasing maxCo, increasing parallel numbers. Sorry I couldn't help you. kkk |
|
March 30, 2016, 14:46 |
|
#5 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Thanks very much for the help.
I increase the maxCo and it speeds up the timeStep. At which extent the maxCo can be increased generally for real case? Thanks, Kossivi
__________________
Thanks in advance, Kossivi. |
|
March 30, 2016, 23:11 |
|
#6 |
New Member
|
Hi Kossivi
Actually, I have not set maxCo more than 10. Simulation usually become in fail when I set maxCo more than 10. But I've heard around 100 was OK from someone. Generally simulation is not constrained by Courant number in implicit method. You would be able to set maxCo as much as higher possible. But it means you extremely decrease the accuracy of the time-discrete comparing with one of space-discrete. I guess it's better to decrease the accuracy of the space-discrete and set smaller maxCo. Regards, kkk |
|
March 30, 2016, 23:48 |
|
#7 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Hi kenyoh,
Thank you for your answer. It really helps. I have another question. It is regarding the convergence of the solution. I know for transient the convergence is not straight as is steady state case. But what I do for this case is to monitor the moment and the pressure rise (p_outlet - p_inlet) and compare them to the steady state results with the (simpleFoam-MRF). If the results after let's say 2 revolutions are similar to the one of the steady state so I conclude for convergence. of course I look at the pressure convergence too. Is that way of doing correct ? regards, Kossivi
__________________
Thanks in advance, Kossivi. |
|
March 31, 2016, 10:07 |
|
#8 |
New Member
|
Hi Kossivi
Yeah, I think so. I guess you might have to monitor more long time for making sure it really become steady state. If transient simulation become steady state, the monitoring value will become constant after that. If the monitoring value does not change after 2 revolutions, Then you can stop the simulation. Regards, kkk |
|
March 31, 2016, 13:35 |
|
#9 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Hi kenyoh,
Thanks very much for your reply. It really helps. regards, Kossivi.
__________________
Thanks in advance, Kossivi. |
|
April 4, 2016, 22:03 |
|
#10 |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Hi kenyoh,
I have a question regarding the comparison of a steady turbomachinery (MRFSimpleFoam for example) and the unsteady turbomachinery calculation. How can I compare the results a the same angle ? Is simple for know the angle of the unsteady solution. But how to know it for the steady one ? Any idea on it will be appreciated. Thanks, Kossivi
__________________
Thanks in advance, Kossivi. |
|
April 4, 2016, 23:05 |
|
#11 |
New Member
|
Hi Kossivi
Actually I have never used MRFSimpleFoam, so this is my guess. I guess the rotation force might be added to N-S equation as a inertial force in MRFSimpleFoam. Just imagine that you were seeing the result of pimpleDyMFoam on the rotating inertial system. The results of both might be the same. Then you could evaluate it at any angle you like. Regards, kkk |
|
April 20, 2016, 18:39 |
|
#12 | |
New Member
Kossivi GOKPI
Join Date: Jun 2010
Location: France
Posts: 28
Rep Power: 15 |
Quote:
Thanks.
__________________
Thanks in advance, Kossivi. |
||
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Is Playstation 3 cluster suitable for CFD work | hsieh | OpenFOAM | 9 | August 16, 2015 14:53 |
pimpleDyMFoam case setup | sivakumar | OpenFOAM Pre-Processing | 0 | December 11, 2013 06:24 |
bubbleFoam kEpsilon not converging for k on tutorial case | vbchris | OpenFOAM | 1 | March 31, 2013 23:45 |
Missing math.h header | Travis | FLUENT | 4 | January 15, 2009 11:48 |
Free surface boudary conditions with SOLA-VOF | Fan | Main CFD Forum | 10 | September 9, 2006 12:24 |