LES solver for DNS vs. dnsFoam
 

Hi all,

in the description of dnsFoam (dnsFoam.c) there is written:
...
Description:
Direct numerical simulation solver for boxes of isotropic turbulence
...

Is this solver only suitable for boxes?? Is it possible to simulate pipeflow with this solver.
- In my opinion, if I switch off the force-field it should be possible??

Is it possible to do DNS or quasi-DNS with the pimpleFoam solver using Smagorinsky-Model and set Smagorinsky-parameter = 0????

Or is icoFoam better for DNS?

Which is the best solver in OpenFoam for DNS or quasi-DNS for different flows (pipeflow, impinging jet etc.)???

kind regards
Florian

Just my thoughts:
to use LES for DNS, you have to decrease the grid size and time step to Kolmogorov scales. Then the subgrid scale stress tensor should become zero automatically. Therefor you dont have to change the Smagorinsky coefficients, and you can "use" any sgs model as it will have no effect at all.
But I am unsure about the pimpleFoam solver, because it is for large time steps. Maybe pisoFoam is better?!

Hi HanSolo,

thx for your reply.
You are right, if my grid is fine enought then nuSgs tending to 0 (it dosn't matter which LESModel I use. But this should be only right for "real" DNS. I will do some quasi-DNS, then my grid is nearly fine enough to solve all turbulence scales.
At the moment I use pimpleFoam with CFL of 0.2 -> Timestep is really small. And the gridsize is in the range of Kolmogorov-scale. I'm not sure if there is a better possibility in OpenFOAM. Pherhaps dnsFOAM without the force or pisoFoam.

I will check my results with pimpleFOAM next week. Pherhaps there is somebody out there, who has done a DNS or q-DNS with openfoam and can share his experience.

By the way I have read the following paper about q-DNS in OpenFOAM:
"Quasi-DNS capabilities of OpenFOAM for different mesh size"
and the thesis of Steven van Haren.
So there should be experience out there ;)

kind regards
Florian

Hi All

Can you please tell what parameter in the LESProperties file should be set to 0 to convert the LES to quasi-DNS?
Thanks.

Regards
Nihar

Hi itchy!


in my opinion you should keep a source term in the pipe's axial direction, ensuring that turbulence is properly sustained (under an energetic point of view). If no energy is provided in your pipe, the flow will eventually stop. Indeed, including a source term in axial momentum equation, is what is usually done in any publication concerning turbulent pipe flow.


regards


Lorenzo

Just a comment: by putting a source term in the momentum predictor you are not "injecting energy", you are setting a bodyforce! Subtle, but important difference.