open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,

I'm working on a (big) telemac modelling.
I need to model a dam break. In my 2D modelling, I have to inject the wave hydrograph (in which the discharge evolve from 10 to very high discharge in a few seconds ...).

I chose to inject discharge along my upstream open boundary, et use a QH law downstream.

I ran a first short simulation to "wet" meshes along my upstrem boundary (see. CONDIN.JPG).




Then I try to run the simulation with a "false" hydrograph (Just to chack everything's ok before to test with very high discharge).

I tried in sequential (1 processor) and parallel (8 and 18 processor), but the calculation stop.
I attached the log file.



i get this message :

@STREAMLINE: :SCARAT: THE NUMBER OF TRACEBACK INTERFACE CROSSINGS IGEN < 99

I have read several threads on this forum that mention this error.

- I tested the hypothesis of an instability linked to the division of the domain into sub-domains (1 processor, 8 processors and 18 processors) --> For parallel calculations (8 and 18 processors) the log is the same, it is the one that is attached. For the sequential calculation (1 processor), the calculation is still in progress (!! it is so slow ...) but it has not yet crashed whereas I have just passed the point of crashing the parallel model.

- I'm trying to find where coulb be an instability, but, But as this 1 processor simulation is still running, I think that the problem is not there, but more on a "programming" subject.

And then, I am completely lost!

So I'm coming to you today for some help. I still don't use Telemac very much, so I might miss some "basic" points for you. So I ask for your indulgence :)


For your convenience, I am attaching my case.txt.



Thanks in advance to the community.

Hello Laurie,

A few remarks quickly, looking at the log you have uploaded:
- beginning: I am not sure you installation is perfect with your smpd messages,
- you seem to use an old release (v8p0 or older), you should upgrade if you want to be assisted, many bugs have been fix for 4 years (at least) and maybe your issue is already solved,
- you have nodes with up to 12 neighours which is big. You should redo your mesh,
- using stage-discharge curves, I hope you start from a steady state compatible with,
- I can read "PRERES : NOMBRE DE COURANT MAXIMUM : 154.8530" which is a rather big value. First idea is to decrease the time step.

Looking at the zoom of your mesh, the growth ratio between elements may be too big (sometimes more than 2). You may encouter trouble.

I would recommend you to rebuild your mesh and upgrade to the latest release (v8p3r1) first.

Hope this helps,

Chi-Tuan

Hi Chi-Tuan,

Thanks a lot.
I'm re-using an old mesh, and I have not source data. That's why I try to avoid to recreate this mesh. But I know this is a possibility ..
I'm trying with V8P0. It's the more recent version I have. I tried to install the last one. But I didn't succed. I'll try again !


And I'll try to decrease the time step. We will see.

Kind regards,

Hi,

To identify step by step which are the problem of my modelling, I extracted a sbset from my mesh.
I only keep the upper (upstrem) part of my river.
Until I'm using v8p0.

I get this error message :
PRERES: MAXIMUM COURANT NUMBER: 2.531338
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
TELEMAC2D INITIALIZED

Backtrace for this error:

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
unable to read the cmd header on the pmi context, Error = -1

The complete log is here :




The CAS file is here :




Last point, my subset mesh specifications are here :




I don't understand the error message ? Someone knows what I have to check ?

Thanks a lot,

Laurie

Hello Laurie,

I still read a possible too big value for Courant number:
"PRERES: MAXIMUM COURANT NUMBER: 2.531338".
The first idea I have already written is to decrease the time step.

Using the search engine of this forum with "Segmentation fault", you can find some ideas, in particular, this post:
www.opentelemac.org/index.php/kunena/16-...-running-error#39413

Moreover, to check your mesh, please use once the option CHECKING THE MESH = YES (VERIFICATION DU MAILLAGE = OUI in French).

Hope this helps,

Chi-Tuan

Hi again :)

I followed your advice decreasing the time step, but I will decrease again.

Thanks for the topics, I'll read it.

And thanks for the key word to check my mesh.

Hi all,

I fixed this issue.

In my case, the error was linked to an instability caused by a bad parameterization of my downstream condition: a lot of water was entering the model from the downstream with high velocities... The thing was not realistic at all.

I tried to apply as much advice as possible from the experienced people who answered me. My work is not perfect, but at least everything seems to work for the moment. I'll have to be very careful with calculation errors...

Thanks in any case!