open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi Telemac-Users,

I am still struggeling with the water levels in Telemac 3D and Telemac2D.
As I posted some weeks ago, I am working on a reservoir case with a weir structure and a tailwater section.

I modelled it 2D and 3D and found out that the 2D water levels correspond to the measured water levels in the physical model (standing in our lab), but the 3D water levels are about 2km upstream more than 40cm too high ( see www.opentelemac.org/index.php/kunena/21-...els-in-3d-than-in-2d).

First I thought that the weir structure could be the reason, but the results stayed more or less the same when I removed the pillars. Then I thought that maybe there are some kind of mesh-effects, because of some very fine meshed parts. I thus meshed it perfectly regular again, but the 3D water level stayed still more or less the same.

I used my standard cas file from my other reservoir calculations. Changing the turbulence model from k-epsilon to const.eddy/prandtl mixing length changed the 3D water level just about 3cm. The effect of the variation of the bed roughness is in the same threshold.

So, how can I dump the 3D water level to match the physical model?



Hoping for some hint,
Gabi

Hello Gabi,

In view of your curves, the problem comes from the weir that causes too much head loss in 3D. Are you in non hydrostatic mode ? Posting your steering file would be helpful. Physical models are always a difficulty because they sometimes fall within the treatments of tidal flats, e.g. there is a keyword:
THRESHOLD FOR VISCOSITY CORRECTION ON TIDAL FLATS
that is set to 0.2D0 by default and that should probably be set to 0. in your case.

With best regards,

Jean-Michel Hervouet

Hello,

Maybe yet another physical model problem, to start with see my post #16738 immediately before this one.

Regards,

JMH

Hello again,

Another thing, an important question in your case are the upstream boundary conditions on k and epsilon. You can modify the relevant subroutines in T2D and T3D (kepscl and kepscl3 I think) to impose already established turbulence corresponding to a logarithmic profile. I'm afraid that given your geometry turbulence has little room to develop after the upstream boundary, so that what you find in your domain is more or less you you feed into the domain.

Regards,

Jean-Michel Hervouet

Hi Jean-Michel,

maybe I havn’t pointed it out clearly, but I removed the complete weir structure in the last simulations. Therefore it is now just the river bend left. The water depth is about 5-7m and there are just some tidal flats on the river bank, but I am going to try THRESHOLD FOR VISCOSITY CORRECTION ON TIDAL FLATS.

I thought I posted my cas-file, but unfortunately it has not been attached in the post.

Why should the turbulence not develop in this case? I have about 2km river upstream of the river bend (with the removed weir structure), the river is 130m wide and 7m deep and it is more or less a channel geometry (with some sediment deposition in it).



Many thanks,
Gabi

The cas_file

Hello,

thanks, so the first thing to try is :

THRESHOLD FOR VISCOSITY CORRECTION ON TIDAL FLATS : 0.

and then I would try to remove the friction on boundaries, then a mixing length model on the vertical, and on the horizontal a constant diffusion with very low value.

With best regards,

Jean-Michel Hervouet

Hello,

I have the same problem, but vice versa! I'm running a 3D non-hydrostatic model with k-epsilon and the water level is always too low compared to the measured data.

The geometry is very complex (90° curve, 10m deep scours). I've already increased the bottom friction by 1,5, but it didn't have a big influence on the water level. Espacially with higher discharges the influence of the fritction becomes more and more negligible.

Is there maybe some experience or knowledge which parameter I also could take into account due to calibration? I only can think of lowering the coefficients of diffusivity (currently 1.D-06), but that would be kind of "unphysical" in my opinion.

My .cas file is attached

Best regards,

Constantin

Hello,

What is the order of magnitude of "too low" ?

I would advise to try first a constant horizontal diffusion on the horizontal and mixing-length on the vertical. Then look at the effect of the horizontal diffusion, trying 0.1 1. and 10. (k-epsilon may give values which are too small due to the mesh size, the distance from the entrance to establish turbulence, etc.).

Another thing, can you explain your data with the effect of friction or the head loss due to the 90° curve ? If not what would be the cause... ?

With best regards,

Jean-Michel Hervouet

Hello,

thanks for the fast response!

Depending on the discharge, the magnitude varies between -8cm (lower discharges ~900m³/s) and -18 cm (higher discharges ~2000m³/s).

I've already tried mixing length on the vertical and Smagorinsky on the horizontal. Using lower discharges, this combination results in acceptable water levels. But with higher discharges the results of k-eps and ML&Sma become more and more equal (~15cm). The distance from the entrance to the 90°curve is about 2km and almost straight.

The 90° curve is at Rhein-km 551 ("Jungferngrund") and it is marked by a rock at the cut bank and an island at the point bar. The friction coefficients vary between 0,1 and 0,4. But in this section the differences in the water level are acceptable (~ -5cm)

The main head losses occur in a section with a 2km long scour with a maximim depth of 8m below the riverbed (Rhein-km 552-554).

I hope the given information are helpful!

With best regards,

Constantin