open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,
I am currently trying to model a simple flow in a current circulation tank with Telemac3D (v6p3). I wish to use a k-epsilon model (vertical and horizontal).
The final goal of this research is to validate an actuator disk formulation intending to model a tidal turbine. Currently, the turbine model is switched off. Therefore, I just want to model the development of turbulence near the bottom.
The current tank is 10m long, the water depth is 0.3m. I use 50 horizontal planes. A velocity profile is given at the inlet (enclosed file v.dat).
As soon as the k-epsilon model is switched on, the results become very strange with the values of k and epsilon rapidly increasing and becoming unrealistic (see the enclosed file k_epsilon.jpg). I have tried several formulations changing the conditions at the bottom (for the velocities and for the k-eps), the values of k and epsilon at the inlet, the scheme for the convection/diffusion of k and eps, the number of horizontal planes... None of them give realistic results.
I would be grateful if someone can send me a similar case (simple flow with vertical and horizontal k-eps model) or suggest any idea.
The case is enclosed to this message (tel.cas, princi.f,struct.slf, struct.cli, v.dat).
Thank you
All the best,
Jérôme

Hello Jérôme,

Could you put all the mass-lumping keywords to 1, the two implicitation coefficients also (IMPLICITATION FOR DEPTH, IMPLICITATION FOR VELOCITIES) and add:
FREE SURFACE GRADIENT COMPATIBILITY = 0.9 to stabilise and get smoother velocities and let's see what happens.
In cstkep KMIN and EMIN are 1.D-10, and you can also consider choosing higher values like 1.D-7.
Another remark: 50 planes is a lot for 30 cm of depth, you must have a large aspect ratio, which is not very good to get good vertical velocities. It has the effect of linking points on a vertical (especially by diffusion terms), with the effect that solution on one vertical is rather disconnected from other verticals, it looks like we have something like this on k.

With best regards,

Jean-Michel

Hello Jérôme,

Could you put all the mass-lumping keywords to 1, the two implicitation coefficients also (IMPLICITATION FOR DEPTH, IMPLICITATION FOR VELOCITIES) and add:
FREE SURFACE GRADIENT COMPATIBILITY = 0.9 to stabilise and get smoother velocities and let's see what happens.
In cstkep KMIN and EMIN are 1.D-10, and you can also consider choosing higher values like 1.D-7.
Another remark: 50 planes is a lot for 30 cm of depth, you must have a large aspect ratio, which is not very good to get good vertical velocities. It has the effect of linking points on a vertical (especially by diffusion terms), with the effect that solution on one vertical is rather disconnected from other verticals, it looks like we have something like this on k.

With best regards,

Jean-Michel

Hello Jean-Michel,
As you suggested, I put all the mass-lumping keywords to 1 as well as the two implicitation coefficients. The free surface gradient compatility is now set to 0.9.
It is now working much better See the enclosed file.
Concerning the other remark, I use 50 planes because I want to discretise my actuator disk accurately. The aspect ratio is too large. I will therefore refine my mesh horizontally to have a good aspect ratio.
Thank you very much for your help and for the promptitude of your response.
All the best,
Jérôme

Hello Jérôme,

Much better indeed, but you see that the turbulence takes time to develop from the bottom and increases gradually in your model, showing that you should perhaps try to change the turbulence at the entrance, in view of what you get at the exit.

Regards,

JMH

Hello Jean-Michel,
I will now impose the turbulence at the inlet according to the turbluence measured in the current circulation tank.
I will keep you in touch as soon as the data-model comparison (with the tidal turbine) is done.
Thank you again for your time.
Best regards,
Jérôme