Different Results between ke LP and kw SST

Questions and remarks about code_saturne usage
Forum rules
Please read the forum usage recommendations before posting.
Post Reply
sirlb
Posts: 34
Joined: Mon Mar 17, 2014 11:54 am

Different Results between ke LP and kw SST

Post by sirlb »

Hello,

I have a case of external hydrodynamics, with an objective to compute drag force on a body.
I have compared turbulence models k-e LP and kw SST.

For k-e LP, the results show differences with my reference (experimental dataset) up to 20%.
For kw SST, the results show differences up to 4%.

My Re is 5e7, flow is incompressible and modelisation is steady-state with variable step in time and space (SIMPLEC coupling).

I am just trying to understand what would be the reasons for such differences knowing that with other codes we do not observe such differences between k-e and k-w models.

Any feedback is welcome. Thank you.
Design_BIB

Re: Different Results between ke LP and kw SST

Post by Design_BIB »

Hi,

To my experience, There are a lot of parameters taking an effect for accuracy in your results. It's might be best if you post your study domain and essential parameters, if you can, in order to a lots of professional users provided you a value suggestion in meshing algorithm, mesh quality and turbulence model.

In this case, if you want to compare only these two models, I suggest the "k-w SST" result because SST was developed for near wall bounds so flow around objects solving is better than "k-e LP". These two of models was used different equations to solving a simulation of flow.

In addition, Your Re is too high so you should completely trust your mesh quality and change your model for more convergence results


Best regards,
Chayawat
sirlb
Posts: 34
Joined: Mon Mar 17, 2014 11:54 am

Re: Different Results between ke LP and kw SST

Post by sirlb »

Hi,
f you want to compare only these two models, I suggest the "k-w SST" result because SST was developed for near wall bounds so flow around objects solving is better than "k-e LP"
That's what I am trying to figure out, is there any limitation on the use of k-e LP ? Does this implementation of k-e is validated only for specific Re numbers or applications ? I'm suprised to observe such differences between models. Mesh is validated for other codes after multiple convergence tests and gives good results for CS as well with kw SST.

Best Regards,
CAVT
Posts: 9
Joined: Wed Mar 14, 2012 9:10 pm

Re: Different Results between ke LP and kw SST

Post by CAVT »

Luc, for what I remember a mesh that works with K-e will work with K-w SST, but the opposite is not guaranteed. K-e asumes a boundary layer model, so your Y+ should be well above 1; there's a post regarding reccomended values I think for various models (if I recall properly Y+ should be greater than 30 but smaller then 100 for K-e, whereas K-w allows Y+ lower than 1).
From my experience K-w has been more robust than K-e regarding poor quality meshes, so I suggest you to play around with the different schemes Code Saturne offers (upwind, least square, etc.). That has direct impact in convergence and not all the schemes suit all models nor all types of mesh elements.
You may want to check the University of Manchester's website, they have a CFD group using Code Saturne and did a validation of the code comparing several turbulence models, K-e an K-w SST among them.
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: Different Results between ke LP and kw SST

Post by Yvan Fournier »

Hello Luc,

In any case, it will be difficult to understand the difference better if you do not provide additional information as in the forum usage recommendations: with no info even on the code version you are using, nobody can even tell you if you are using a version with a known bug.

As I recall, a bug was fixed after 3.0 for k-omega without scalable wall functions with very low (< 10) y+, which can be worked around using a bit of relaxation for omega. I assume you are not in this range, so the difference is probably "normal", but without details, I can only guess.

Just the same, with no info on your mesh, I don't even know if your near-wall mesh is good or not. Did you visualize your y+ values, or near wall mesh quality ? You may also check the influence of advanced options such as scalable wall functions or

I second CAVT's recommendation t check out the University of Manchester website, as it has more details on turbulence models (basically I recommend this site for everything except detailed turbulence model info, and the Manchester website for more details on turbulence). Check for example http://cfd.mace.manchester.ac.uk/twiki/ ... e/SSTModel. You'll also see examples of cases where k-omege SST is not adapted, such as p8 of http://code-saturne.org/cms/sites/defau ... 8-KNOO.pdf, but which is in a different range than where you are. Combining the Manchester web site and the user meetings section of this site, you'll find quite a bit of info (which is unfortunately hard to summarize).

Regards,

Yvan

Yvan
Post Reply