dear all

i just devote myself to les,and i'm a freshman in cfd.some questions defused me greatly,so maybe you could help me.

1.if one use no-slip BC in les,how to set the distance from the wall to the first grid.i know that the dimentionless y+ is less than 5.it means the first grid point is set in linear zone.but i want to know that how to set y,not y+,when i generate my mesh.

2.how to compute the average quantity in simulation,for example the velocity U. Turbulence is unsteady.could i compute it when the turbulence fully developed?

3.the fluctuant velocity u'=U-(u),where (u) is the filted U. can i use u',v'and w' to compute the turbulent kinetic energy.if not ,what i need?

4.about initial condition.in practice could you tell me how to set the initial conditions?especially the fluctuant velocity u'!

I'm a freshman ,so i need help.I appreciate your help and best regards for you.

Hi,

I would suggest an intro textbook CFD such as Versteg and Malalasekera (1995) and on turbulence such as Tennekes & Lumley or Lesieur. Then move on to more complex turbulence modelling issues for LES and DNS. LES is a fairly advanced and intensive way of modelling turbulence; so starting with simpler bases might help.

Bob

1.you can compute y with respect to y+.

2.yes,in my opinion,it is when the turbulence has developed to a "real turbulence" state from your arbitrarily imposed initial condition.

3.the fluctuations should be those with respect to mean velocity components.

4.I don't think u' is needed as initial condition in LES.

thanks. I can compute y with y+ and the friction velocity U_tau,but U_tau is unknown before i simulate!i just evaluate the U_tau from the log law,but how to set the bulk velocity? is it the bulk velocity inlet or the one after fully developed. if it's the later,i do not know how to do. thanks a lot for everyone.

This should be a kind of estimation procedure. See Chap 7.1.5 in "Turbulent Flows" by Pope(2000) for reference.

Hi there,

I will give you guidance and information on something a little bit restrictive but still you should get a lot out of it.

Let's just consider turbulent channel flow. In such case, you should design your code such that you use the U_tau as the dimensionless parameter. Then it's easy to compute the y+ as a function of y and the Re. In such case the Re is based on the friction velocity, U_tau and delta (the channel half width). Most of the computations that were done for LES of a channel make use of a box size: (2*pi*delta) in x, (2*delta) in y and (pi*delta) in z.

For more about LES in channel flow, you should go to the stanford website http://ctr.stanford.edu/ or check publication by Ugo Piomelli. Check also the paper s by Moser et al., Phys of Fluids.

I hope this helps,

Sincerely,

Frederic Felten.

dear Linfeng above all thanks for your answer. i found the book:'turbulent flow' wrote by Pope(2000) and read the Chap 7.1.5. actually i must know the Re with the bulk velocity U_b first,then i can compute the friction coefficient Cf=0.5*sqrt(U_tau/U_inf) from the log law(U_inf is the inlet velocity).is it right? others question:I want to know how to select the numerical scheme? Which scheme is better? Anyone could send the lastest development in LES or the information about it to me.

2.how to compute the average quantity in simulation,for example the velocity U. Turbulence is unsteady.could i compute it when the turbulence fully developed? I can compute the average quantity by time average in experiment,but in CFD,what i should do.I'm confused greatly.

3.the fluctuant velocity u'=U-(u),where (u) is the filted U. can i use u',v'and w' to compute the turbulent kinetic energy.if not ,what i need?

could you help me?

Sincerely,

Tom

thanks to Frederic Felten Bob may you happy.

The numerical scheme depends on your problem and your computer. In genernal, for FDS, it should be conservative. But you know,higher order schems requires more elaboration. See CTR's website for reference.