open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am working in the last days, trying to find a good combination regarding the 3D settings in my cas file, "playing" with the turbulence models and solvers.
Since time is always short, I would like to ask if any of you already did some tests around turbulence, e.g. which mixing length model/K-eps ..etc.

For the project I am working at, I have built a model of a bay of around 100 km2, with unstructured triangular grid with an avg element length of ca. 150 m. This bay is located in the middle part of a reservoir of ca. 800 km2. In my model I included part of the reservoir main stream to permit inflow and outflow in/from the bay, where water is almost stagnant. The main driving force of velocity in the bay is the wind, which takes the bay to velocities of maximum 1 cm/s. Also, along the boundaries of the bay, I have low water depth (lower than 3 m). This was just as information about my computational domain.

What I noticed so far, that for 3D simulations I got better results using k-epsilon for horizontal turbulence and Prandtl for vertical. Although with the solvers CGSTAB and GMRES computational time is really long. With 2D-simulations, instead, constant viscosity was good enough for the results and, combined with the conjugate gradient method on normal equation was really fast on HPC with parallel processing.

I am simulating at the moment wind-induced flow with mean flow conditions, but in the next future I should implement tracer and heat transport. I have read that Quetin and Tsanis are good for wind drifts. I would really be glad if anyone can give me some advise about some settings/turbulence-solvers combinations that I could try out and that could be better for my specific case.

Thank you in advance.

With my best regards

Hello,

If the Prandtl model works well with your model, I would advise you keep with it, and maybe use a constant value or the Smagorinski for the horizontal. If you wish to improve your results, I would consider incrasing the number of layers in the vertical.

Hope this helps,
Sébastien

Hello,

sorry for the late reply, but I had a long work trip...

About the layers, I chose 14, refined near the bottom (one at 1 % of the water column) and near the surface (one at 2 % and one at 6 % of the water column), while the other levels were equidistant at 10 % of the water column. As information, the mean water depth of the entire computational domain is 9.62 m and the mean water depth of the bay is 3.77 m.

Do you still think I should increase the number...?

Additionally, what do you mean with "if the Prandtl model works well with your model"? Unfortunately I don't have measurements to compare results and calibrate the model. It is not totally clear to me how I can be sure about the turbulemce model to use.
What I noticed, is that, when I take wind into account, iteration number may exceed 200 for CGSTAB and 3400 for GMRES and, therefore, also the computational time gets higher, even on HPC with parallel processing.
Do you have suggestions about all this?

Many thanks in advance.

Best regards

Hello,

If you have shallow waters you should look at the influence of the keyword:

THRESHOLD DEPTH FOR WIND

because wind may trigger large velocities if applied on dry zones.

With best regards,

Jean-Michel Hervouet

Hello,

thank you Jean Michel, but my velocities are not are not unphysically high, therefore, it is probably not necessary in my case.

I tried different combinations in the steering file and I found fast and reasonable if I use Prandtl for the vertical mixing length and I use the constant viscosity for the horizontal turbulence model.

I attach the output file, if one of you may check it...



I created some sections with the help of Janet and Postel3D and I visualize them with the help of Rubens. I attach a couple of ones that I consider correct and a couple of ones that I do not really understand (... the only reason I could think is that they look like that because low velocities in the bay... ). The first 2 are taken near the inflow (in the main stream), the other 2 in the mid of the bay. Both without and with wind, alternatively. The sections taken are small, the width of one section is approximately the length of an element edge.


Thanks for the help.

With best regards

... and 2 taken between the main stream and the mouth of the bay...







Thanks!