in using wall functions for les can some one please tell me what the appropriate y+ values are.

how refined should i make my wall region for using wall functions in les.

thank you.

Hi there,

The standard range for wall functions is above y+=30 and beyond. This however apply for equilibrium and stationary flows, what to do for LES (instationary and non-equilibrium) I don't believe anyone know, as it is impossible to derive wall functions for LES-flows. I suggest to do both a resolved and a wall function LES for a small domain, and check the difference. Modify the wall functions to suit your needs and then hold your thumbs that the large simulation (which I presume can not afford to resolve) will be valid. Another choice is to use DES (detached eddy simulation) where both the large scales are resolved using LES, and the near-wall is resolved using some turbulence model with or without wall functions. The interface between the two zones are then of course another point of worry. I guess this didn't help much,

Cheers Jonas

Can someone tell me if the Van Driest wall damping function can be used in conjunction with Wall functions?

See this:

http://www.glue.umd.edu/~ugo/research/pbpss_ijhff.pdf

and especially THIS one:

http://www.glue.umd.edu/~ugo/research/annrev.pdf

thankyou jonas,

you say y+ = 30 and beyond, however surely there is an upper limit to the range too.

surely the wall region can only approximate to a log-profile for a certain distance from the wall.... the wall 'region' must end somewhere?!?!?

please let me know what you think.

Hi,

The upper limit for the wall function to be valid is depent on the Reynolds number. For channels flow, the well known DNS computations (by Kim et al.) have a very narrow span, due to there limit Reynolds number. The Re_tau=180 for instance does not seem to have a proper log region at all. (y+ max = 180). A rule of thumb is not to use the log-law above 10% of the channel height. Using a resolved computation it is "easy" to check for yourself, plot the velocity in a log-log plot and try to estimate where the constant slope disappear. This is of course more difficult to do if you place you first node at the limit of the log-law, however with todays computer resources, I believe it will never be a problem. Run your case with what you believe to be a reasonable resolved wall-function mesh, and you will most probably have a node within the log-law. Check the y+-value for the first node. If it is more than 30, you are safe.

You may check a report a wrote on wall functions http://www.tfd.chalmers.se/~lada/pos..._report_WF.pdf which give some info, and some references for further reading.

Regards Jonas

So the question now arises: what if you can partially (but not completely) resolve the wall layer? What if 30 > y+ > 5 (say)? Do the results of a wall function calculation still hold true?

Hi there,

I have a small idea of that you have been into this question before. As you probably know, the two distinct regions are the log-layer above y+=30, and the viscous sub-layer below y+=5. In-between is the buffer region, where on the contrary to the other two region it is very difficult to specify the velocity profile. (Viscous sub-layer: U+=y+, and log-law: U+=ln(y+)/kappa+B). A hap-hazardous way to treat this region is to implement a switch, use U+=y+ if y+<11.63 (the cross-over point) otherwise the log-law. If the mesh in generally is a wall function mesh, and only in re-circulation, re-attachment and separation points the y+ is low, this will be ok. I tried to modify a k-omega model to be useful on any mesh, however I can't say that the attempt was successful, check if you like: http://www.tfd.chalmers.se/~lada/pos...ME_orlando.pdf

Regards Jonas

Hi,

I am interested in your discussion. I am wondering how to calculate u* (friction velocity)in cfd modelling, both in LES and RANS. Your response would be much appreciated. Please forgive my ignorance of cfd modelling.

Best wishes, Thomas

Friction velocity is defined from the wall-shear as: tau_w=density*u_tau^2 Wall shear stress is computed as: mu*dU/dy for a refined mesh (mu=laminar viscosity)

This fails on a coarse mesh where dU/dy is not correctly predicted. Some codes uses an effective viscosity, to correct the velocity-gradient at the wall in order to produce the correct wall shear stress, i.e.: tau_w=mu_eff*dU/dy Another way is to compute the friction velocity from the log-law, i.e.: U/u_tau=ln(y*u_tau/mu)/kappa+B However you need to solve this equation iteratively.

You should try to use the method used by your code, to retrieve the correct u_tau for you particular case.

Jonas