open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi there,
I am trying to test the Smagorinksi model on a straight channel and I have two cases to compare in order to identify the effect of using such a model.
I have a high flow and a low flow (in both cases I am using the same mesh (horizontally and vertically in terms of the no of vertical layers).
For a constant viscosity run, I used to have a higher diffusion coefficient for the low flow run to help it to stabilise in a reasonable no. of timesteps without lowering the timestep too much.
However now when testing the Smagorinksi model on such a comparison what do I do?
Do I set the diffusion coefficients (vertical and horizontal to default (molecular diffusion)?

If I kept the horizontal diffusion coefficient at the values of the constant diffusion run,, and set the vertical diffusion to default: the run with the higher flow rate reaches a steady state in a reasonable time (naturally because it is less in need for Smagorinksi) but the one with the low flow struggles for a good while and fails to acheive a steady state.
Any thoughts or ideas with this?
Thanks alot,
Zeinab

Hello Zeinab,

If you choose Smagorinski for horizontal and vertical diffusion, a subroutine smago3d is called.

If you choose Smagorinski only for horizontal a subroutine smago is called (and the vertical diffusion is done with another model).

In both cases there is a hard-coded parameter CS equal to 0.1D in them, this parameter can be case-dependent.

In both cases the diffusion coefficients given in the parameter file must be the molecular diffusion (they are added to the result of smago or smago3d).

This may ex^plain what you get,

With best regards,

Jean-Michel Hervouet

... to complement what Jean-Michel said ...


remember also that the K-Epsilon model is a 3D TURBULENCE model in TELEMAC-3D, so if you have Smagorinski for horizontal and K-Epsilon for vertical, TELEMAC will automatically ignore your Smagorinski for horizontal and replace it with K-Epsilon.

Sébastien.

Thank you Jean-Michel and Sébastien for your responses which shed light on matter and helped me understand better the model.
I noticed the smag and smago3d modules being called in TELEMAC3d.f if ITURBH or ITURBV = 4 (Smag model).
Also noticed if I chose to leave the horizontal turbulence at say constant viscosity but use Smag. for the vertical, the choice of the horizontal turbulence is over-riden, same as when using k-epsilon or k-omega.
I will experiment with the model and will absolutely come back with feedback.
Thanks alot for the help, its very much appreciated.
Zeinab

Hi Jean-Michel and Sébastien,
I promised to come with feedback on the Smagorinksi model.
The struggle is continuing to reach a steady state with the model when setting the diffusion (horizontal and vertical to molecular).
I decrease the timestep and increase the duration of the run, yet no steady state.
I tried setting the horizontal diffusion coefficient to say 0.1 but use molecular diffusion for the vertical. While this has improved results, still it is a struggle to reach steady state.

I noticed the same difficulty with k-epsilon and k-omega which also both require diffusion to be set to molecular value.
With the k-epsilon model, I tried using another turbulence scheme (mixing length) to intiate the run and then use the mixing length run as an intial condition for the k-epsilon model. Also trebled the length of the channel. Still not working well as the timestep requirement is very small which makes the investigation computationally expensive.

Any suggestions, experiences, success stories with these models?
Best Regards
Zeinab