Dear Friends,

I am having difficulty in running my k-w model on fine meshes, where the aspect ratios are quite large. I am seeing that my results are showung the same trend as the experimental/existing computations which I am expected to match with. In fact a good match is achieved nearly 1e-3 to 1e-4 residue fall, but the solution leads to an underprediction and a failure, at times if I run the code further to say 1e-6 or 1e-8. I suspect that there is a greater production of dissipation, which causes and underprediction. I have ensured that no quantity amog k,w or eddy viscosity becomes negative. Please note that I have no problems on caorser meshes, and hence my conclusion. Can anybody suggest some remedy, possibly in terms of a lower bound for k or w or their production etc... Anticipating suggestions and comments

Regards and Thanks in advance,

Ganesh

Perhaps it is a problem related to the near-wall region. The omega equation has a very stiff wall boundary conditions which, depending on how it is implented, can cause problems on fine near-wall meshes. Does you simulation on a finer mesh also have a lower y+ value and a finer resolved boundary layer?

Dear Jonas,

Thanks for the reply. The y+ values are around 2 to 3 for the finest meshes I have, although I do see the problems I reported earlier on meshes with y+ around 10-12. I am setting the wall b.c. as below,( with suitable non-dimensionalisation)

K_w = 0 ; W_w= 60 nu_w/(Re*beta*d**2),

I know that there is also a bc for W_w based on the surface roughness, as reported in Wilcox. Do you think this can be of any help ?

Thanks for the comments and suggestions

Regards, Ganesh

Yes, from my experience the surface roughness wall boundary condition as described in Wilcox's book is much more stable. You can still set a roughness which is so smooth that it gives a hydrauliclly smooth surface. Several years ago when I did my PhD i experimented with a few various wall boundary conditions for omega and my conclusion was that you should always use the surface roughness model even when you have a smooth wall (and set the roughness so small that you get a smooth wall).

Try the time scale bound k-w model by Medic and Durbin(Transactions of ASME, Journal of Turbomachinery, Vol. 124, pp. 187-192, April 2002). See page 191 Appendix. There is an wall bodundary condition for w-equation.

I hope this helps.

Halim

Dear Jonas and Halim,

Thanks for the comments and suggestions.

Regards,

Ganesh

Hi,

the Fluent Manual suggests to have the y+ for k-w within the log layer 30 < y+ < 300 - a value close to the lower bound is most desirable.

I do not know, if thats a good idea. I have bad results compared to the same msh with k-e and wall functions.

Hope, it helps

Ralf

Dear Jonas,

I have an intersting observation. I am seeing that my code blows at lower CFL on coarse grids but with higher CFL, it goes on well, and I end up with the right results. This looks perplexing, since stability would be more with a lower CFL. I am not sure how to explain this, or if it is a result of a bug, but I would like to know if you had some such experience.

Awaiting you reply,

Thanks in advance,

Ganesh

This is only valid if you run the k omega model with wall functions. The original k omega model is a low-Re model which requires a resolved boundary layer with y+ for the first cell preferably below 1.

Can't say that I remember having had this problem before, sorry.

I am not sure what scheme you are using to solve the flow and k-omega equations. All schemes are dissipative to some extend, and you will often see that numerical dissipation is higher at higher CFL (assuming an implicit scheme). It may seem counter intuitive (based on linear stability analysis) that the stability of your scheme is larger at higher CFL, but in practice it's entirely possible. You also mentioned that you don't have any problems on coarser meshes. Again, it could simply be a damping effect, caused by higher numerical dissipation on a coarser mesh.

Dear Mani,

Thanks for the comments. I am using a Roe scheme for solving the flow and turbulence equations in a coupled manner. I am also thinking to try out other schemes, but I cannot find a possible explanation to the failure on finer grids.

Any suggestions are most welcome

Regards,

Ganesh

Hi,

In Fluent, there is NO possibility to turn of or on a wall function when using k-omega, not to mention setting the type of wall fuction.

So do you say, if we are using k-omega with high Re, then there is automatically a wall fuction used??

The Fluent manual (6.2) says: "...if [Transitional Flows] option is not active, then the mesh guidelines should be the same as for the wall functions"

So k-omega "sees" the Re number and decides if a wall function is neccessary or not?

Ralf