|
[Sponsors] |
Fluent's new turbulence models, how to choose ? |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
February 25, 2012, 14:05 |
Fluent's new turbulence models, how to choose ?
|
#1 |
Member
Charles
Join Date: May 2011
Location: France
Posts: 77
Rep Power: 14 |
Hi there,
I noticed that Fluent 13 has got a few new turbulence models compared to Fluent 6. I've hit the Fluent help about them but that didn't gave me clear explanations. From what I've understand : - "Transition SST" is an evolution of k-epsilon with roughly the same CPU expanse but better results especially in separated flows. I'm currently running some tests on this one. - "Transition k-kl-omega". I haven't understand the difference with transition SST, but as far as I'm concerned the tests I've made were disappointing. The residuals omega and kl kept doing some spikes and therefore I'm not confident in the results. - SAS. Once again the theory behind this one is blurry for me. It seems to be a compromise between the RANS model and the expansive DNS/LES. The results looked cool but the calculations are very long. My main interest concerns high Reynolds, high speed flows for external flows around jet aircrafts. For example I would like to get the frequency and Cd/Cl amplitude for a turbulent flow around something like a pitot probe. I therefore need to run unsteady tests in 2D and 3D. The DNS is out of the question and the LES might be too expansive and too precise for the kind of results I'm looking for. These new models seems to be a nice compromise. From your own experience, which one do you prefer ? Which one would you recommand for my needs ? Thanks, Charles |
|
February 25, 2012, 14:33 |
|
#3 |
Member
Charles
Join Date: May 2011
Location: France
Posts: 77
Rep Power: 14 |
Hi Far,
For the Reynolds that would be around 7M (180 m/s, caracteristic lenght 0.6m). My mean Y+ is around 10. |
|
February 25, 2012, 14:46 |
|
#4 |
Super Moderator
|
DNS is impossible. LES is out of question for this Reynolds number coz you need very fine mesh (find the formula on Google about the No. of nodes in terms of Reynolds number).
Transition is also not possible at Reynolds number as flow is fully turbulent. I would suggest SA and SST model. SST model by default (V 13 & 14) uses the automatic wall function similar to CFX SST model. Which is valid for all yplus. I recommend yplus between 6-15. |
|
February 25, 2012, 14:55 |
|
#5 |
Member
Charles
Join Date: May 2011
Location: France
Posts: 77
Rep Power: 14 |
Thanks for your reply. So transition model can only be used when the flow is laminar on a significant distance and turbulent. What would be the maximum Reynolds number for this model ?
SA didn't look not good enough because the flow was oily and slightly returned from a transient state to a totally steady flow which is not possible. So you think I shall use k-omega with the SST model and that might get more accurate results that the SA ? But k-omega is still RANS right ? |
|
February 25, 2012, 14:58 |
|
#6 | |
Super Moderator
|
Quote:
Transition model is useful for Re 1e05 |
||
February 25, 2012, 15:19 |
|
#7 |
Member
Charles
Join Date: May 2011
Location: France
Posts: 77
Rep Power: 14 |
Thanks for your help, I'll try the k-w SST model.
|
|
February 25, 2012, 15:22 |
|
#8 |
Member
Charles
Join Date: May 2011
Location: France
Posts: 77
Rep Power: 14 |
Oh and what do you think of the SAS ?
|
|
February 25, 2012, 15:29 |
|
#9 |
Super Moderator
|
SAS is good and and can be classified in between DES and RANS. It has flow structure like DES and mesh requirements like RANS. But it is unsteady method and requires skills and thorough reading.
|
|
February 25, 2012, 15:48 |
|
#10 | |
Super Moderator
|
Quote:
|
||
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
LES & DES turbulence models | Seva Reehal | CFX | 5 | October 18, 2009 08:44 |
Are there more FLUENT Turbulence Models? | Travis | FLUENT | 3 | December 30, 2008 01:45 |
Additional turbulence models for FLUENT | Travis | FLUENT | 0 | December 19, 2008 11:58 |
Turbulence models for GCV method dosen't work! | Sima Baheri | Phoenics | 0 | January 6, 2007 04:08 |
values for turbulence models | Ben | Siemens | 8 | July 12, 2004 13:42 |