open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello Robinson,

I'm using k-epsilon too in my 3D model. I noticed this before. In some examples I saw HORIZONTAL TURBULENCE MODEL is set to Constant Viscosity while VERTICAL TURBULENCE MODEL is k-epsilon. In fact it's logical to me that both HORIZONTAL and VERTICAL should be consistent with each other. The Smagorinski model you mentioned is the method used in Large Eddy Simulation. I don't know much about that. But I don't see why the k-epsilon cannot deal with Reynolds stress in the horizontal. In theory it's valid for all directions.

The epsilon in the standard k-epsilon model can represent the rate of energy transfer from large eddies to the rate of dissipation in the small eddies since the closure approximations normally used parametrize the various terms in the epsilon equation as functions of large-eddy scales.

You can find this in the book 'Turbulence Modeling for CFD by David C. Wilcox'

Best regards,
Qilong

Hi,

Thanks again for Mr Hervouet and Qilong's very kind help.
From some literatures and my own model tests, it is true that even huge variation of the horizontal eddy viscosity/diffusivity really does not make any change to the model results. On the other hand, the horizontal eddy viscosity/diffusivity is more related to the model stability.
Now I have tried to modified 'lecdon_telemac3d.f' to allow for a combination of Smagorinski and k-epsilon, and it seems to be working well. But as you've mentioned that k-epsilon is valid for homogeneous turbulence, I think the default is more sound.
Thanks again.

Regards.
Robinson