September 7, 2016, 06:43
|
Discussion: LES SGSmodels and solvers
|
#1
|
New Member
Giulio
Join Date: Feb 2016
Location: Birmingham
Posts: 13
Rep Power: 10
|
Dear Foamers ![Big Grin](https://www.cfd-online.com/Forums/images/smilies/biggrin.gif) ,
I am an OF beginner, playing with OF LES skills on a complex geometry with highly separated flow. I have a series of concerns regarding LES, its implementation and its relationship with OF solvers.
- pisoFoam would be the most natural solution for solving highly transient flow, but it's very sensitive to time steps and y+. I usually put 2 to 4 loops and then 1 to 2 nonOrthogonalCorrectors (usually 1). Am I overusing it?
- pimpleFoam allows a more stable simulation, but the settings of the Pimple loop are not clear to me (it appears I need much more iterations than PISO and still convergence is very slow). Is it necessary to relax p and U like for SIMPLE? I put 20 to 50 outerCorrectors, 2 Correctors, 1 nOrtho + relaxation (0.3-0.7). Is it sensible (still it's really slow in convergence);
- simpleFoam (which is the SIMPLEC of Ansys as far as I understand): can it be used for transient flows? According to the literature yes, but still I don't think it's a good idea to use it with LES (am I right?);
- Adaptive time Step is it a good idea to use it with LES? Any suggestion on how to implement it? Any trick (because it usually ends up with a very low time step: 1e-6...)
- Any other alternative for solvers?
For LES I mean SGS Smagorinsky model. My idea was to then switch to more advanced SGS models:
- dynamicSmagorinsky is not implemented (just the kSGS and the Lagrangian, which is very sensitive to the mesh and the maxCo) from v.3 on, right? (I wanted to try it out)
- WALE can be a solution to tackle y+ requirements, is it alright? If so what would you suggest for a y+ value? Furthermore, the simulation is much slower than with Smagorinsky (but I've only tried Pimple).
- vanDriest damping is an alternative to WALE for tackling behaviour of Smagorinsky near wall. I get good results even with y+=5. Is it the case, that vanDriest helps to get better solutions even for not perfect y+? (or it's beginner's luck?);
I don't mention the Schemes because we'll end up with an infinite post ![EEK!](https://www.cfd-online.com/Forums/images/smilies/eek.gif) . I just wanted to hear from more expert users if you have suggestions/experience and if you find something is wrong with what I wrote!
This is a cylinder upstream to a thick blunt plate, without any wake modelled (infinite plate). The mesh is huge, therefore I need to cheat (I cannot make it uniform, I have to stretch cells outside of the separation bubble zone I'm interested in)...
Thank you very much,
Sorry if I've said stupidities!!!
Giulio
|
|
|