open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I try to validate TELEMAC-3D model with physical model at 1/35 scale (Froude similitude) before to study scale effects with scale 1/1.
My assumptions are described in following points:
- geometry reduction by ratio 35 (on XY and Z);
- Froude similitude applied on time step, simulation time, discharge and water level at boundary conditions, friction coefficient for the bottom.
- others keywords: non-hydrostatic version, 10 horizontal levels (with % fixed on water depth), K-epsilon turbulence model, 10^-6 for diffusion of velocities, smooth turbulence model for the bottom)

Does it exist other parameter to take in account?

There are some strange results in some parts of the domain with high vertical speed (W component) with low water depth... at these points, horizontal mesh seems correct but we have a mesh vertical distorsion superior to 10. Is it only a meshing issue?

Thanks for your help!
Sincerely

Hello,

I think you have all the necessary scales.
I would use Nikuradse roughness for bottom friction, and most probably you have a rough turbulent regime at the bottom?
If you have wetting/drying in your model you should descrease the value for THRESHOLD FOR VISCOSITY CORRECTION ON TIDAL FLATS.

Aspect ratio can be an issue but it is really problem dependent. Hard to say..
You can try to decrease the number of horizontal levels.

Best regards,
Clemens

Thank you for your advice.

Next run with these keywords:
-OPTION FOR THE TREATMENT OF TIDAL FLATS = 2
TREATMENT OF NEGATIVE DEPTHS = 2
BYPASS VOID VOLUMES = TRUE
THRESHOLD FOR VISCOSITY CORRECTION ON TIDAL FLATS = 0.005 (ratio 1/35 basically)
TREATMENT ON TIDAL FLATS FOR VELOCITIES = 1
TREATMENT ON TIDAL FLATS FOR K-EPSILON = 1

Results are better!
But I don't know the keywords effect one by one...

Hello,
I would start with most of the default keywords:
e.g.
OPTION FOR THE TREATMENT OF TIDAL FLATS = 1 and TREATMENT ON TIDAL FLATS... = 0.
For the THRESHOLD FOR VISCOSITY maybe you should consider also the Froude similitude since eddy viscosity has the unit m2/s. However this is now rather an academic view and shouldn't be of much importance. I don't know how this keyword interacts with the use of the k-e model since the k-e model for standard applications calculates eddy viscosities in the range of 1.E-5 to 1.E-3, at least in my experience. Maybe the developer can say more about this keyword THRESHOLD FOR VISCOSITY..
A critical issue can be the chosen time step. Our experience is that in order to get a converged solution the k-e model requires often a (much) smaller time step.
You can sort out some things in the starting steps by giving e.g. the Mixing Length model in the vertical and a constant eddy viscosity in the horizontal. We have quite good experiences with this approach in applications where the horizontal turbulent exchange is not so important and it is much cheaper in terms of computation time.

Best regards,
Clemens