Turbulent transition with k-epsilon
 

Hi All,

sorry for the rather fundamental question, I am not an expert in turbulence modelling.

It is possible to simulate turbulence development around a bluff object, with non-turbulent inflow, using RAS with turbulence models such as k-epsilon ?
(I am interested in the transient evolution of nu and k, not the velocity field btw.)

1. What inflow condition should I set (I can't set k, epsilon to 0 as the model would get instable)
2. I hear k-epsilon is not well-suited for simulating non-fully developed turbulence. Low-Re k-epsilon models however only help at the walls, right ? So this won't help the fact that I am trying to simulate a non-turbulent (low-Re) inflow.

Unfortunately, I can't use LES or the like, but need to check how far it is possible to get with RAS (it's for a science project).

Thanks for any hints !

Hi Martin

Interesting problem, but as far as I know, there are no transition models implemented in OpenFOAM (please do correct me, if I am wrong, as I could use them/it as well:)). With respect to the use of k-epsilon, then you could read some on transition in Wilcox "Turbulence Modeling for CFD" (2006), where he describes why k-epsilon in fact are unsuited for use in transitionmodel (division by k is present in the model), whereas k-omega is more suited for that.
I have found some articles by Menter describing the modeling of transition using a set of advection-diffusion equations, however some of the important empirical relations are stated as "properitary" in the article, hence it is not possible to reproduced his results neither trying to extend the use of the model:(

Good luck,

Niels

ABout transient turbulence
 

Hi all!

From general point of view
If there is no suitable model than the direct solution of motion equations can help in you case? (DNS)

There exist equations directly for turbulent eddy diffusivity also together with thermal eddy diffusivity. (not k- e models branch). A-E model.

Sorry but both proposal is not perfect

Hi,

many thanks for the answers !
I had a look at the Wilcox book and the Menter 2006 paper, and decided to see how far I'll get using SST, and maybe try to add a simple transition model if it's not accurate enough. I'd like to implement the Menter model, but like ngj mentioned, Menter didn't specify his calibration data. Is anyone aware of another group that reproduced his results using public domain data ?
(I'll probably not have the time and the 'expert eye' to reproduce these myself )

@kazantsev: No, unfortunately I can't use DNS or LES, we're investigating methods that are explicitely cheaper than that. It not actually a engineering problem we're trying to solve, we're looking for methods that could be adapted in for our simlutations for Computer Graphics.