CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)

 dictatore_bozorg April 29, 2012 10:23

a symposium about large eddy simulation

i want to simulate an external Fluid Flow (air) on a flat plate with streamwise zero pressure gradient.

http://s15.postimage.org/xjpc5fjxz/Untitled2.jpg

for this purpose i made a cubic computational domain in GAMBIT like below.

http://s18.postimage.org/ti9o9dlj9/Untitled.jpg

and choose below Boundary Condition (BC):

INLET:

for inlet i choose VELOCITY_INLET boundary condition, cause we want simulate ONLY turbulent part of Boundary Layer, if we use
velocity_inlet BC we can use fluctuating velocity algorithm option in Fluent.

Question 1.
in some case i see we can choose PERIODIC BC, what is your Opinion about velocity_inlet BC and PERIODIC BC?

Question 2.
if the inlet velocity profile will be parabolic our simulating is much better. i know by writing UDF its possible but
is it possible to have parabolic with fluctuation in inlet by UDF? if you have any knowledge please share with me?

Question 3.
if we use Velocity_inlet BC on inlet surface and use FLUCTUATING VELOCITY ALGORITHM (for example Vortex method) the the vortex will made on hole inlet surface whereas in real Flow we have vortex only in boundary layer, indeed we can separate our inlet in two domain one of them contain boundary condition and the other contain invicid Flow, what is you suggestion for this Problem?

http://s18.postimage.org/a4bi6s6et/Untitled3.jpg

Bottom:
we choose WALL BC for bottom Surface.

SIDES and UPPER SURFACE:

cause i want to simulate an external flow i choose OUTLET_PRESSURE for
right, left and upper surface.

Question 4.
what do you think about these BC? did i choice right BC for sides?

OUTLET
for outlet i choose OUTLET_PRESSURE, cause i want a ZPG flow.

base on my studies the grid size for LES can be 0.1 of DNS grid size and we can calculate DNS grid size base on kolmogorov's scales.
for choosing a right time steps we must calculate Courant Number (CO=(V*delta(t))/(delta(L)) and CO number must be less than 1 (CO<1) in any point of computational domain.

Please share you knowledge and if possible answer upper question and let keep on this thread step by step

 LuckyTran April 29, 2012 10:47

Is it an infinite flat plate? This problem appears to be 2D.

I must ask, what is the purpose of you running this LES and what is that you are trying to study? Your setup is not suitable to study the turbulence. At best, you can learn to run LES using this setup. Unless you are running a very high Reynolds number simulation, there is little that can be drawn from this work. For LES, the most important task is to give the proper inflow conditions and your inflow conditions will not produce the best results.

1) After a brute superdomain simulation, periodic with rescaling is the next way to do boundary layer simulations, but you do not have the correct setup for that. and you'll need to write your own rescaling code. inlet velocity with fluctuations is not terrible, but it is too much an ad hoc fix, I would not trust it unless I had to.
2) parabolic profile is for laminar flows. I thought you were simulating a turbulent one? You need to get this profile exactly correct. That is why periodic-bc with rescaling or generating the inflow conditions through LES is better than velocity with fluctuations.
3) I imagine this is possible but I do not know how to do it. As an alternative, you can create 2 velocity inlets, 1 with the inlet profile and fluctuating velocity, and another with only freestream velocity and no fluctuations.
4) Side surfaces should have the periodic boundary condition with 0 pressure gradient. Also check to make sure your cube is wide enough so that your sides do not introduce errors. Top surface can be done in many ways. I think pressure outlet should work.

 All times are GMT -4. The time now is 20:22.