open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I am using telemac3d to simulate a flume. I tried both k-e and smago as horizontal turbulence model. The vertical turbulence model is mixing length.

I got the vertical profile for the velocity along the central line of flume. See image attached. I don't understand why the velocity doesn't decrease at the bottom when I use k-e as horizontal turbulence model. I refined layers and the y+ at bottom are smaller than 30. I think the first layer is close enough to the bottom. Coefficient of horizontal diffusion of velocity I set 1.D-6

The vertical profile for the velocity looks much better when I chose smago. Actually I used the same setting as k-e case but different horizontal turbulence models.I also let model run enough time to make the flow fully developed.





Another thing is that in the code smago.f. I don't understand how (mesh size)**2 be calculated. Are there any reference I can follow?

Best Regards
Yue

Hello,

In fact when you ask for k-epsilon for the horizontal turbulence model the vertical turbulence model is also set to k-epsilon (this is debatable but well, this is the current situation, and it does not explain your result with k-epsilon). Another thing to look at is your law of bottom friction, Nikuradse using the vertical velocity profile while other laws like Strickler are designed to behave like in 2D (to have the same friction parameters in 2D and 3D). Another important thing is the turbulence at the entrance of your domain. In cstkep.f you have also a number of options hardcoded for k-epsilon, e.g. the type of boundary conditions for k and epsilon at the bottom and free surface.

The way mesh_sise**2 is computed in smago.f is explained before the series of calls to vector.f, in fact the finite element formulation implied in vector adds a multiplication by the integral of test functions, which is an area around every point, this area is condidered to be the factor mesh_size**2.

With best regards,

Jean-Michel Hervouet

Hi Michel,

Many thanks. I will run more tests and see what would happened.

Best Regards,
Yue