open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear all,

I simulated wind-induced flow in my domain (bay + part of the river stream) with TELEMAC-3D for 10 days (enough to reach steady state conditions), using constant viscosity and Prantl mixing length as horizontal and vertical turbulence model, respectively.
I am analyzing with Postel-3D and Rubens some particular sections, because I would like to see circulation patterns in the XZ or YZ sections.
But there is something that wonders me. Along the different layers, I notice difference in the arrows direction, without or with the wind, which is normal and expected, due to the wind. But when I look along the vertical and consider only the wind case, the arrows directions are always the same along the layers, with an increasing velocity moving towards the surface (most of the cases). I am not observing any circulation over the vertical direction (see attachments




)
FYI: I chose sections where the flow was in direction or against the wind (approx.) and I cut YZ sections to see the vertical in Rubens.
Do you think it is correct even if I don't see circulation patterns? I will continue with tracer and afterwards heat transport, maybe I will see in those cases some circulation over the vertical..?!

Thank you.
Regards

Hello,

When the wind pushes water at the free surface, it can come back on the bottom, or with 2D recirculations and it seems that it is what you get. However I would question the turbulence model, Prandtl mixing length, which is not really meant for wind. You can try k-epsilon instead (in this case you will have k-epsilon in the vertical and the horizontal). On our Berre lake application we do find the 3D recirculations with the k-epsilon model.

With best regards,

Jean-Michel Hervouet

Dear Jean-Michel Hervouet,
thank you.
For the horizontal direction I use already a constant viscosity model, not a k-epsilon. I tested and calculated the option also with k-epsilon some time ago but the computational time was very high and I thought that, since I have very low range of velocities in the bay and not complicated turbulence patterns, I could use the simple constant viscosity for the horizontal and Prandtl for the vertical, since it is also suitable for tracer and heat transport.
Actually, I couldn't see it in the sections that I chose, but if I analyze the whole bay in the different layers with Blue Kenue, I can see the differences over the vertical (see attachment please where I reported for example the bottom and the surface layers)
Thank you.
Best regards

Hello,

FYI
I am trying again to simulate with K-Eps in both directions to check if I notice differences, but the case with wind require really long computational time: with CV&Prandtl on HPC ca. 30', with K-Eps much more than 5 hours... still waiting...
Is it normal? Therefore also I thought for my case k-Eps was not really suitable...

Thanks.

BR

Hello,

Is this due to a large number of iterations in the k-epsilon model. This is perhaps something to tune, by changing the solver or the accuracy asked.

With best regards,

Jean-Michel Hervouet

Hello,

... I am checking now the results (K-eps in both direnctions) and they seem not to differ relevantly, comparing them with the option constant viscosity & prandtl as horizontal and vertical turbulence model, respectively.

With best regards