Implicit Large Eddy Simulations
I have recently found out about Implicit LES: The idea of the method is to use LES without subgrid model, i.e. run laminar simulation on a fine grid, which is much coarser than greeds used for DNS.
The justification of the method is based on the fact that energy of the eddies with space size less than l is proportional to l^2/3, so if our problem scale is L and the grid spacing is l, then the unaccounted eddies carry only (l/L)^2/3 part of the turbulent energy (see e.g; http://www.mit.edu/~auranga/presentations/Uranga_presentation-AIAA-2009-SanAntonio.pdf ).
Now, I would like to run flow around an airfoil at Re=50K, the characteristic size of the problem is the boundary layer thikness, which is relatively big at this Re:
Does it mean that I can run Implicit LES for a 3D airfoil, (which, e.g. can be laminar calculation in FLUENT) with grid spacing at foil surface of about 1/50 of the boundary layer size and get reliable results? (y+ for such spacing is order of 1 at this Re)?
People doing ILES claim that numerical error plays role of subgrid model. Will this work with traditional finite-volume schemes implimented in, say Fluent, or this methods needs special numeric schemes.
In my experience, it almost always turns out that what you are actually doing in Fluent is ILES, in the sense that the SGS has always a very limited effect, if any.
This is determined by the numerical schemes available, which almost always overwhelmes the SGS part (hence their use is questionable at least).
Are those schemes suitable for ILES? In my opinion NO (neither they are for explicit LES) but the answer can change according to the application (bluff body flows could be best suited for this).
To convince yourself of this just run two simulations, one with a SGS scheme and one without it. Will the difference in results justify the overhead of a SGS model (even if a low one)?
However, the approach to ILES in FLuent is to use a laminar model (which means no turbulence model at all), a suitable convective scheme (bounded central) and a "proper" grid and time steps. What proper means for the grid is application depending, for the time step you need a courant lower than 1 (2 if unavoidable)
Is the convective scheme you have mentioned (bounded central) is a of Non Oscillatory Finite Volume (NFV) type? As far as I understood one has to use this type of discretization scheme for ILES.
I will be grateful if you could give more info on discretization options to set in Fluent laminar unsteday simulations for run it as ILES.
|All times are GMT -4. The time now is 08:40.|