open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,
I have a problem with a 2D-simulation of a flood.
After about 1/3 of the simulation time, I receive this message: "EQUNOR <BIEF>: EXCEEDING MAXIMUM ITERATIONS: 100 RELATIVE PRECISION: 2.222222" an infinitive number of times. I tried to reduce the time-step, to change the boundary conditions, to modify the structural mesh, but nothing changes, except the moment in which the simulation start to have this kind of problem.
I already performed a similar simulation without problems, and the only thing I modified is the elevation of some nodes (I increase the elevation of the railways and decrease the elevation of some canals).
Attached you can find the files that are needed to perform the simulation, together with a screenshot of the message that I receive and the folder with the partial results.




I'm using opentelemac, version v6p3-3.0 for Windows
Please help me! :)

Thanks in advance,
Francesca

Hello,

The files are not attached but provisionnally we can say that it is a stability problem, so we need to look at the numerical parameters. I am surprised by the use of the normal equation method, which is meant for non symmetric matrices. The first thing to check is thus whether you have :

TREATMENT OF THE LINEAR SYSTEM : 1 or 2

2 (pseudo-wave equation, symmetric system) is much faster and a bit more stable, and in this case SOLVER = 1 (conjugate gradient) is the best method. Then there are a number of possibilities to stabilise : diffusion, compatibility of free surface gradient, etc., depending on your case. Another thing to look at is whether you have dry zones or not and how you deal with them (option for the treatment of tidal flats, treatment of negative depths,...).

With best regards,

Jean-Michel Hervouet

Thanks for the quick reply.
At the moment I have:
SOLVER = 3
TREATMENT OF THE LINEAR SYSTEM = 1 (coupled)

Should I change them?
I try to attach my files again, sorry.




Best regards,
Francesca

How can I attache my file? Sorry for the stupid question, but I add the zip folder but it can't be submitted...

Hi
Take care about the file size!
There is a limit on the forum for attachment.

And Yes, try the recommendations of Jean-Michel and change the keywords value

Regards

I can't find the limit for the files size on the forum rules, but I have the same problem also if I try to attach a file that is 3KB.
Btw, I will try to change the parameters that Jean-Michel suggested.
Thanks.

Regards

Thank you all, it works!!!!

Hi, i changed my parameters into
SOLVER = 1
TREATMENT OF THE LINEAR SYSTEM = 2
and it worked.

But now I simply added a culvert and it gives me the same problem as before:
GRACJG <BIEF>: EXCEEDING MAXIMUM ITERATIONS: 100 RELATIVE PRECISION: NaN.

Do you have some other suggestion please?

Thanks in advance,
Francesca

Although this warning, the simulation works until the end. Thanks!

Hello,

No, NaN (Not a Number) says that the solver crashes, so there is something wrong in the computation, you cannot consider that it works.

With best regards,

Jean-Michel Hervouet