Hello,

Im trying to simulate an incompressible, turbulent flow (Re bulk=56000) in a channel and validate it with experimental data.

Im using simpleFoam for this with the turbulentmodell LaunderGibson. The problem is that my simulation behaves more laminar then the experiment does. I guess this is because of my seting of the kinematic viscosity parameter is to high. I have calculated it from experimental data for bulk values of velocity and Reynolds number and the hydraulic diameter.

-Now Im wondering if it might be wrong to base the kinematic viscosity on the bulk velocity?

-An other thing thats bothering me is that i get very different results from different gradings towards the channel wall. How do I know how big the grading needs to be? Now it is set to be 0.01.

[olwi]

Hej Erik,

Since the turbulence model uses wall functions, you should usually AVOID resolving the boundary layers too much... The criteria for an appropriate reslution near the wall is to check the value of yplus (y+), which is the non-dimensional wall distance of the first computational node. (Look in a good textbook on turbulent flows, e.g. that of Stephen Pope.) I think there is a postprocessing utility with openfoam that will compute yplus for you. Then plot it on wall surfaces. Yplus in the range 30-80 is probably a good target for you.

Is see nothing wrong with the way you set your nu. As for the definition of Re, you should of course define it in the same way as the guys who made the experiment! That's the only thing that counts...

If I were you I would start off with the simple K-epsilon model. For a simple channel flow an RST model is probably overkill. K-epsilon is tuned well to channel flows in particular. Often it's more important to have well-converged solutions and good grids, than to fiddle around with different turbulence models. If you insist on using the RST model, and you have problems with convergence, it could be a good idea to restart from a converged K-espilon solution. How you do that in OpenFOAM is beyond med at this point...

Good luck!

/Ola

Thank you very much.

Its a big help using checkYPlus as a guideline. unfortunatly I got one of my best results from a LaunderGibson turb. model with grading 0.01. But im trying to find an alternative now using a better mesh.

/Erik