open TELEMAC-MASCARET The mathematically superior suite of solvers

Another try:




It was the pixel dimensions...

Regarding picture uploading:

• The types allowed are jpg, gif and png. The bmp type is not allowed as it is very inefficient.
• Maximum size possible is 500kb.
• Maximum dimension is 1200x1080 pixels (Height and width). Larger files will still be uploaded but downsized to that limit

Finally, I strongly advise users to give a descriptive name to the picture (e.g. "WaterSurfaceElevation_Test#1_Username.jpg"). If you give a generic name like "picture1.jpg" there are risk that the picture will be overwritten in the future by another user .

I hope this helps.

Thank you very much for these advices !

My numerical configuration is identical to what you suggest. I even just tried to change IMPLICATION coef from 0.55 to 0.51, but without any effects on the results.

I will try with the princi.f file, but I face problem with this procedure on my machine (W7 64b).

Maybe i have more chances with the finite volume method. I will try.

When is the version 6p2 to be released ?

PL

... as for your wave profile, the mesh could have an impact, but I would first try to make your numerical coefficients the most appropriate for this case. In particular:

TYPE OF ADVECTION : 1;5
TREATMENT OF THE LINEAR SYSTEM : 2
SOLVER : 1
IMPLICITATION FOR DEPTH = 0.55 # or, if stable 0.51
IMPLICITATION FOR VELOCITY = 0.55 # or,if stable 0.51

The later two should be accompanied by the following change to the LECDON_TELEMAC2D.f (make a copy into your PRINCI file, and modify as follows ...)

The last line with TETAC should also be commented out.

Note that we have made that change in the upcoming v6p2. It used to be there in previous versions to make some of our numerical schemes more stable / faster.

Hope this helps.

Sébastien.

... or, if you have the time, you could also try the FINITE VOLUME formulation, second order to ensure conservation of wave energy.

Sébastien.

Hi PL and Seb,

Nice results, PL. We also have some measurements that compare well to T2D both in amplitude and wave length in quite a complex channel (lots of wave reflexion). If I can get a hold of the comparative figures, I will post some.

Sebastien, I would also be interested to cross-check some of my previous simulations with one of the FV scheme. I think there have been quite a few additions to the FV options since v5p9. Is there a particular numerical configuration that you suggest? I believe recommendations in the user manual are more focused on the FE options.

regards,

jeremie

Hi again,

I am trying to run my case with FV.
I read the documentation and have the following configuration in my steering file:
- EQUATIONS : SAINT-VENANT VF
- SCHEMA EN VOLUMES FINIS : 2
- PAS DE TEMPS VARIABLE : VRAI
- NOMBRE DE COURANT SOUHAITE : 1

The calculation runs fine during a few seconds but then an instability occurs just when the computation ends (with the message "Fin normale du programme").
Furthermore, in the log, i have the following message at each iteration :
Attention CFL demande > 1 !..!
Pas de temps (avec CFL=0,5) = 0,2......

Is there any specific numerical configuration that suits better to a FV analysis and allowing to avoid these instabilities ?

Thanks a lot in advance,
PL

That is what I meant about "if you have the time" ... for finite volume you need to reduce your time step drastically .

The rest of your settings seems OK. Put your COURANT NUMBER to 0.5 to be on the same side, and print to screen (not to file) every time step so you can check time step adaptation.

... and also you may be right -- I did not look at your case in detail -- you may need 3D (?)

Good luck !

Sébastien.

Hi !

Thanks for these advices. I will set the courant number to 0,5 and try again. I also modified the type of discretisation with a quasi-bubble speed, as described in the manual:
DISCRETISATIONS EN ESPACE =12;11

But the first results (before instability) give a smaller wave amplitude than the FE method, which is not matching the calibration data. I think i will rather try to optimize the FE model by playing on the element size and other physical parameters, such as friction and diffusion coefficients.
But i might come to the conclusion that 3D would be even more appropriate... !

I will post here interesting results / feedbacks when i have something to share.

PL

Hello Jermie,

Pilou1253's setting is what you need.
You know we would be happy to help on this forum or off line. Try and let us know.

Good luck !

Sébastien.