open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

Yes, if you have strong changes in the depth and several layers of elements drying at once, velocity may take some time to adapt. The treatment of friction on dry land is a bit tricky, it is based on the depth at previous time step, it must be high enough to cancel velocities when depth is 0, but not too much to allow flooding at a correct speed.

On the number of cells: depending on the machine, and number of time steps, it is variable, but we are currently doing studies with 100000 elements, we use desktop Linux machines with 8 cores, or Linux clusters. Look at the Malpasset test case (26000 elements, 4000 time steps) it takes here about 10 s with 8 processors. Another user reported 4 s on a 12-core machine.

With best regards,

Jean-Michel Hervouet

Hi,

So i guess the best configuration would be to have at least several cells in the riverbed (3...), or to use a water surface profile as IC.
Regarding the cells number, my current geometry has 117000 elements, and the computation takes 6-7 minutes for 3000 time steps in non-paralel mode, which is very good ! My question was more to know if telemac has a maximal number of cells that we should respect. Maybe the limit is given by the size of the result file, which shouldn't be more than 2Go on Windows machines if i understood right...

Regards,
Pierre-Louis

Hello,

Well the limitation is rather several millions of elements on a desktop computer, and computer time probably becomes then the first problem. Daresbury lab tested 200 M elements, but with thousands of processors.

JMH

And the 2 Go limitation is only for 32 bits machines, JMH.

Hello !

I have another question regarding another type of geometry.
I am modeling a river with overbanks. The terrain profile at the downstream BC location is illustrated on the graph below.

I am wondering which type of BC should i set for such a geometry. So far, i have tested BC = liquid with velocity profile = 1.
This configuration does not work properly : i have messages in the log stating that "Probleme mal pose, vitesse libre entrante par exemple au point... / Probleme mal pose, hauteur libre sur frontiere avec vitesse entrante et ecoulement torrentiel par exemple au point...".

As it is not possible to define a rating curve on such a profile, i then tried to set artificially a critical flow at my DS BC by creating a waterfall on the 2 last rows of nodes. The problem remained the same.

Does someone have any advice ? It is really not possible to define as BC "critical flow" for example ? Should i subdivide my BC and set for each of them a rating curve ?

Thank you in advance,
PL

Hello,

I am now modeling another geometry, with a rating curve as DS BC.
The model runs normaly, but when checking the rating curve as computed by Telemac and when comparing it with the law defined as BC, there is a slight difference (it seems to behave like a hysteresis).
Is this normal (I would have expected the 2 laws to be identical...) ?

Furthermore, does anyone have an indication to give me regarding my previous post ?

Thank you in advance,
PL

Hello,

I think this is due to the relaxation. The prescribed discharge given by the rating curve is not imposed at once but with a delay, to avoid brutal oscillations. This can be removed or tuned if it is a problem.

With best regards,

Jean-Michel Hervouet

Thank you Jean-Michel.

I will try to set such a velocity profile and see if the model behaves better at the downstream BC.

Regarding the rating curve, how can I remove this relaxation feature ? I don't think i will face oscillations since i am modeling the filling and emptying of a large reservoir, with small water surface evolutions.

Thank you!
PL

Hello!

You can do what you want with relaxation if you take in your Fortran file the subroutine:

sta_dis_cur.f (library telemac2d)

You will see that at the end there is a relaxation coefficient of 0.02. This could be a keyword if users want (to have yet another keyword...).

Regards,

Jean-Michel Hervouet

Hello,

You can perhaps try a velocity profile equal to 4 or 5 (the latter only in version 6.1), so that the velocity decreases when depth is small, it will avoid supercritical flow at the entrance.

JMH