open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all!

I am modelling the outlet of a lake into the sea which consists of 3 parallel concrete canals. The canals' bottom is located higher than the natural bottom in the lake and in the sea, with locally steep slopes - up to the canals and down to the sea. The natural sea level is located higher than the canals' bottom. The flow in the canals is subcritical or near critical.

I modelled it in 2D and 3D and I get rather different results. In 2D, I have a lower water surface (0,2 - 0,3 m) in the canals than in 3D. As a result, I have a stronger tailwater effect in 3D which leads to a higher lake level, for the same discharge.

I use the same mesh and the same numerical parameters (time step, friction coefficients and law, turbulence model and diffusivity coefficients [ie Smagorinsky, on both horizontal and vertical in 3D ; 1.E-6 m2/s], implicitation coefficients, advection schemes, etc). The 3D model is non hydrostatic and has 14 layers.

The same case was studied on a physical model and the comparison tends to show that results from the 2D model are closer from the reference state. It is surprising since I expected the 3D model to show better results (mainly due to the strong bathymetry changes on both ends of the canals).

What are the expectable differences in 2D/3D when modelling an outflow on a very steep bottom (becoming deep)? It does not seem to be a energy dissipation issue since I have rather similar values in 2D and 3D (although computed from the depth averaged velocity...), but what can explain the water surface differences?

I attach my T3D steering file.

Any help would be higly appreciated!

Thanks in advance,
Best regards

PL

Hi PL and Telemac Users,

I have noticed the same phenomenon in one of my cases. I am modelling the flow over a weir structure and the water level in 3D is upstream of the weir 0.2-0.3 m higher than in the 2D calculation. I have also the same mesh and the same friction coefficients. The results of Telemac 2D compare also better with the physical model.
I varied the turbulence model and several parameters, but the water level is still too high in 3D.

I am using v7.0. My last idea ist to run the simulations with v6.3 and check the results.

Happy for any hint,
Gabi

Hi Gabi,

Thanks for your feedback.

We managed to lower somewhat (only 4 cm, still 8 cm to high!) the upstream lake level by changing the vertical turbulence model to Prandtl (mixing length).
We don't see any other trick to do apart from tuning the Smagorinsky coefficient, but this is not the right way to solve the problem since we have a good accordance between 2D-3D for other geometry configurations without tuning this coefficient.

Another run in hydrostatic mode and with a more classic distribution for vertical levels (MESH TRANSFORMATION = 1, we had 2 with a refined mesh near the bottom in non-hydrostatic) also gives lower lake level (- 5 cm, but still 7 cm higher than 2D and physical model).

The explanation that seems the most likely is that the head losses are somehwhat higher in 3D than in 2D. But we cannot guaranty that.

Best regards,
PL

Hi

In a certain way, it could be logical to have different results in 2D and 3D if you have the same friction coefficient. In 2D there is a vertical average as in 3D you have some planes...
I remember there was a short notice on this particular subject in the past. Not sure it's always available on the site.
Hope this helps

Hi PL,
hi Cath,

yes, it is logical, that the water level in 3D is higher than in 2D.
But unfortunately fits the water level in 2D the water level in the physical model better.
I did some sensitivity analysis concerning turbulence models in 2D and 3D two years ago with v6.1 and in this cases the k-e model gave lower water levels than the mixing length model (@PL – I tried the mixing length model yesterday, it lowers the water level about 0.1m, but I am still 0.2m to high right upstream of the weir).
In my case, the water level in the 3D model is ok about 1.4 km upstream of the weir, but then it seems like the water is turning into honey and not fitting through the weir, causing the higher water level straight in front of the weir (with backwater effect of 1.3km).
So, how to lower the water level without turning down some coefficients in a physically incorrect way?
Thanx,
Gabi

Hi all,

Thanks for your answers.

I remerber this topic:
www.opentelemac.org/index.php/kunena/21-...l2d?lang=fr&start=10

It states that with the same friction coefficients (if not Nikuradse law), the 3D hydrostatic version should give the same results than 2D. But this is not excately my case, even though I should make some further tests on turbbulence models to be 100% sure.

But in my case I do strongly believe that the geometry is too complex to be 100% compatible with 2D theory (steep slopes at the canals inlet and outlet) - just as your weir case I guess, Gabi. Therefore I expected the 3D model (non-hydrostatic) to fit better to the physical model results, which is not the case.

In another geometry configuration (with smaller canals and thereby smaller velocities), I don't observe such differences between 2D-3D, which led me to think that head losses are somewhat higher in 3D (which is highlighted in the larger canal alternative, with higher velocities). Can it be the case??

Best regards
PL

As you didn't do the approximation of null vertical velocity in the 3D model you can find a greatest water depth.
In you steering file, you have VERTICAL TURBULENCE MODEL = 4. For a mixing lenght model you have to write 2. Perhaps you did it later when you have tested the Prandtl model.
You can also test different numbers of plans.
You have the same result in the 3D-2D models for the same friction because in 3D (with Strickler law) Telemac uses the verticaly average velocity. Switch to the Nikuradse?

Regards,

Hi,

to point it out again: the problem is not the greater water depth, the problem is that for some reason the 2D calculations of Pl and me are closer to the physical models than with 3D.

I started with some configurations, which worked well in previous projects, but in this case, 2D with constant viscosity 0.5 gave much better results than 3D with k-e (I use 2D with cv 0.5, because 2D with k-e tends to give to low water depth).

I attached a file, with show the water levels and the weir structure. The water level downstream of the weir and at the inflow is ok, but upstream of the weir, 3D gives higher water levels (HQ100 means 100 year flood).

Maybe I miss the point, but I think, the 3D water level should be closer to the physical model.

Gabi

Hi,

OBD, when you say that:

"As you didn't do the approximation of null vertical velocity in the 3D model you can find a greatest water depth."

I can interprete that with non hydrostatic version we would get a higher enery level upstream of a ogee weir compared with 2D. But in 2D the flow is defined by the critical depth on the sill, which corresponds to a discharge coeff of 0,385 (neglecting friction losses). In 3D with vertical acceleration we should get a greater coeff and thus a lower energy level. Or?

As mentionned by Gabi, the main concern is that 2D shows better agreement than 3D with physical model results...

Best regards
PL

I have well understood that you have a better result with T2D. I wanted to say that generally T3D gives greater amplitude of the free surface variation than T2D.
I understand that you use weir as a boundary condition in both T2D and T3D. Perhaps you can test to mesh the weir with T3D (with an approximation concerning the slope and perhaps the width of the sill)?
I hope this reflexion can help.
Regards