open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I have found an anomaly when simulating a tracer in telemac2d parallel mode.
the sub-domains are not consistent in respect to the computed tracer plume as if they are not converting back to the correct global position. If I run the simulation in non-parallel the plume is consistent and significantly different than the parallel results even though every thing in the cas file is the same. Has anybody else found this with windows version (I assume it is a bug) The hydrodynamics runs correctly in parallel. I attach output screen dump to demonstrate the anomaly.

regards
Tony

Hello,

If you are sure that this cannot be due to a specific Fortran file that you use for your case we shall have to receive your data files to run the case ourselves.

Regards,

Jean-Michel Hervouet

Hi Jean-Michel
In this case I am not using any princi file. I also came across this problem in December last with another application, it ran in 3d parallel but in 2D had same problem and again no princi file used.
The files are big so I will check latter this evening whether the same thing happens for a smaller project.

regards
Tony

Further to my post
I created a simple test case of a bay with outfall and single tide condition and ran for 2 parallel processors and also withoutparallel . I selected the outfall close to the partition between the 2 submeshes and found again incompatability between the two submeshes. the depth, elev and velocity were fine. I attach the test case, I have also found same problem for other projects.

Hello,

Well received. I'll have a look at this case and will report on it (allow a few days...).

JMH

Hello,

I found a bug in the treatment of characteristics when quasi-bubble velocities and a linear tracer are advected, this is probably the reason of what you noticed. The best bypass of the problem is to have the keyword :

TYPE OF ADVECTION : 1;5;14

to have another method for the tracer (and it will be conservative, unlike the method of characteristics). It may be necessary to add :
TREATMENT OF NEGATIVE DEPTHS : 2

It would be faster to use the wave equation option :

TREATMENT OF THE LINEAR SYSTEM : 2 SOLVER : 1
SUPG OPTION : 0;0

in which case the free surface gradient compatibility = 0.9 will fight the inf-sup wiggles and you can have a linear discretisation :

DISCRETISATIONS IN SPACE : 11;11;11

which would also be a solution to the problem.

With best regards,

Jean-Michel Hervouet