open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I'm currently trying to calibrate the 3D hydrodynamics near a dam on a particular river. The dam is mainly constituted of 32 piers (see image below). All floodgates are considered raised here. The headloss just upstream to downstream of the dam is rather low (few centimeters). The water surface profile is accurately represented with model.




Upstream of the dam at an approximate distance of 100 m, I get acceptable results of the velocity profile compared to the ADCP measurements using the ke turbulence model in both vertical and horizontal directions.




However, just downstream of the dam, at an approximate distance of 30 m, I can't reproduce the acceleration near the right bank part of the river as show on the measurements:





I also tried with a constant velocity diffusivity instead. I get a slight acceleration on the right bank but still, the velocity profile at this specific location is not reproduced accurately enough. Additionally, with constant velocity diffusivity, the velocity profiles upstream are not as clean as with the ke turbulence model.




I join my cas file for review.

I'm looking here for calibration tips from experienced user in similar hydrodynamic influenced by multiple piers cases.


Any help would be gladly appreciated.

Have a nice weekend


Mathieu

Hello,

A nice mesh!

My problem is that I do not understand why there should be an acceleration on the right bank, as everything seems symmetrical in your case, unless the extra structure on the right is a place where more water is released ? The Telemac-3D result downstream shows a symmetry that seemingly answers to the symmetry of the problem, so I probably miss something. Anyway the only parameter left is the friction on the bottom... in case it is not symmetrical.

Trying a constant viscosity model on the horizontal and a mixing-length on the vertical could also be useful, in case the k-epsilon underestimates the diffusion, what I always suspect with the mesh size in free surface flows.

Did you try also the non hydrostatic version?

With best regards,

Jean-Michel Hervouet

Hello Jean-Michel,

Thank you for your quick reply.

The dam is almost symmetrical. However, there is an apron that was refined in the mesh located on the right side just behind the piers. The acceleration is located within this apron.

I tried the non hydrostatic version, but the results did not change significantly.

Concerning the mesh size required for the k-e model, what is the recommended size so that the diffusion is not underestimated ??? Also, what is the spacing between z layers that would be required ???

Regards

Mathieu

Hello,

Hum, very complicated question. With internal flows the mesh size for k-epsilon model is of an order of centimeters. Certainly too small to do it in our applications. On the vertical we got good results with 32 planes on the vertical, or 16 if you refine near the bottom. An extra unsolved problem is the aspect ratio opposed to the fact that k-epsilon has an assumption of isotropic turbulence. Only convergence tests with different meshes can tell you if the mesh size is enough.

In your case don't you have different friction coefficients between the apron and the rest of the domain?

With best regards,

Jean-Michel Hervouet

Hello,

I tried lowering the friction coefficient to an acceptable limit only where the apron is located (where the acceleration was measured). There was only a minor acceleration not enough to represent the measurements.

If I have enough time, I'll do a mesh sensitivity test with a much finer mesh at the dam location and add more z layers. I'll post the results if they are relevant.


Thank you very much for your time and tips.

Mathieu