open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am studying the dynamics of the Southwestern Atlantic Shelf between 28,5°S and 42,8°S with TELEMAC V6P1 (attached file "STR3.txt"), and I have to use the k-e or K-w models, but when I use the VERTICAL TURBULENCE MODEL= “3” ( K-e) or "7" (K-w) don't works. I have made many tests changing the solvers for diffusion and schemes for advection, but I didn't get any success. I would really be glad if anyone can give me some advise about settings/turbulence-solvers-schemes combinations

.

Thank you,

Ernesto Luiz Casares Pinto

Hello Ernesto

what do you mean by 'k-eps and k-w don't work'? it crashes or it gives unphysiscal results?
options for vertical turbulences will depend strongly on your application and on what you want to focus on. Moreover, it will depend on your mesh quality, vertical mesh transformation, time step, advection schemes, etc...
I saw that you are using 10-6 for vertical diffusion velocities which is the order of molecular viscosity and thus, depending on application nature, could be very low. your time step looks big also.
In a nut shell, try to optimize all numerical options (time step, advection schemes, mesh quality, etc) and then you can focus on turbulence by going with an increasing complexity. you can start with constant velocity diffusion and tune the value in order to obtain physical behaviour.
A final remark, with RANS approaches we are aware that we can not reproduce satisfactorily any kind of turbulence structures. Moreover, the highly anisotropic nature of meshes (element size for horizontal mesh very large compared to those of vertical direction) makes the uses of RANS models questionnable.

with my best regards
Riadh ATA

PS: please upgrade to last version in order to benefit from last developments and bug fixes.

Dear Riadh,

Thank you very much for your response. In reality the simulation using SMAGORINSKY (HORIZONTAL TURBULENCE MODEL=4) and K-eps or K-w (VERTICAL TURBULENCE MODEL=3 or 7) runs aproximately 3 days and it crashes with the message “GRACJG(BIEF): EXCEEDING MAXIMUM ITERATIONS 5000 RELATIVE PRECISION:NAN”.
I have already tested many combinations of solvers and advection schemes according to the operating and reference mannuals. The study area is of the order of 2000km along the coast (parallel) and 800km offshore. The unstructure mesh was built using the software Bluekenue and the discretization was ( first- 14km, second near the middle of the total area- 7km, third near the shel break- 3.5km, forth near the Prata River- 2km, and Patos Lagoon it reaches 60m). The result is a 2D mesh with 80193 nodes and 158567 elements. I think I must to improve the mesh. For example, trying to use 10km as the first discretization, instead of 14km.

I will follow your suggestions and return back soon.


Once more, thank you very much for your help!

Best regards,

Ernesto