[hsieh]

HI,

I am solving a tube flow. The inlet veocity varies with time such that flow changes from laminar to turbulent (and then it may go back to laminar again). How do I deal with BCs for k and epslon in this case? Thanks!

[fabianpk]

I think you will have problems in calculating transition from laminar to turbulent flow using the k-epsilon model, since it cannot handle such flows (correctly).

/Fabian

[hsieh]

Hi, Fabian and Pierre,

Thanks for the repies!

What will be the best model to handle transition? Thanks!

Pei

[hsieh]

Sorry for the typo! I meant "replies" not "repies". (I pressed send too quickly)

Pei

[fabianpk]

I know that LES models can handle transition, it has to do with that you model all the scales of the turbulence in RANS models. This means that there can be no real dissipation of the turbulence according to the kolmogorov -5/3 theory.
Personally, I was quite uncomfortable with the variety of turbulence models before I read books and took classes about it, it's not really a matter of picking one, just any one. They all have different traits and work well or not depending on your flow.
Of course, you could always close your eyes and just run the case, it depends on what kind of accuracy you want.

/Fabian

[vatavuk]

I think you guys are meaning epsilon instead of epson (which is a printer).

[hsieh]

Hi, Paulo,

You never heard of k-epson model? I am surprise!
(just kidding, sorry about the typo). I need a spell checker.

Fabian,

I agree with you 100%. Around 14 years ago, when I attended my first APS conference in Cornell, Prof. Lumley said: Everytime I abroad an airplane, I am nervious because I wonder what turbulence model was used to design this plane.

This problem I am interested is a tube flow with water in it. Re # varies from 1000 - 3000 depending on inlet speed (inlet speed is a function of time). The goal is only to study qualitatively the difference between Re 1000 - 3000 (hopefully seeing the transition if it exists). LES will be time consuming to do. Now, my delemma is: Do I do this with my eyes close or open?

Pei

[fabianpk]

Assuming you don't have so much time, maybe k-epsilon is the way to go. However, I suggest you try more than one model, maybe try a RSM model too?
I am not really an expert in the field, but I would treat the different Reynols numbers as separate cases. Thereby avoiding the transition, and just seeing what the difference between the cases are.

/Fabian