open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello!

I working on a coupled model of Telemac and Tomawac simulating the currents inside a lake.
When running the model on a cluster with 40 cores I get an error, which I'm not really sure how to solve it even though the systems gives some hints. Here the error:

MODULE STREAMLINE, SUBROUTINE SCARACT
SIZE OF BUFFER SIZEBUF TOO SMALL IN 1 PROCESSORS
POSSIBLE SOLUTIONS:
1) CHECK THE C.F.L. NUMBER,
LOOK FOR LARGE, OR INFINITE VELOCITIES?
MAYBE YOUR CASE IS GOING TO CRASH ANYWAY
2) REDUCE THE TIME STEP
3) REDUCE THE NUMBER OF PROCESSOR

What exactly does "C. F. L. number" mean. I thought of
C: wave celerity
F: Froude number
L: Courant number.
Is that correct? I checked them and they seem ok. Is there a combined values which needs to be in a specific range?

I also tried to reduce the time step and the no. of processors, which only let the error appear a bit later during the time I model.

Thanks for help. If needed I am happy to provide any files.

Hello,

C.F.L. stands for Courant Friedrich Levy (Here courant is a name nothing to do with current), this adimensionnal number is calculate as
C dt/dx

where C is your celerity (here the celerity of waves) dt your time step and dx your space step. The closer to 1 this number is the better your scheme works.

Roughly it shows how many number of cells your wave is crossing in one time step.

As it is advised reduce the time step will decrease your CFL. Time step should "follow" the smallest space step, I guess you have big difference of space step in your mesh.

Hope it helps

Thierry

Hi Thierry,

Thanks for the detailed explanation. This helps a lot and clears my confusion.
You are right my time-step is quite large and I tested the variable time-step option and could see that the system decreased the time-step significantly.

Will lower the time step.
I also checked my mesh in detail and realized that it has some overconstrained triangles, which may cause some troubles. Any suggestion how to get rid of them and clean my mesh in a easy way?

Hello

I have never done it but my colleague says that it can be easily done using stbtel.py using the keyword OVERSTRESSED TRIANGLES CUTTING.

If you are using stbtel you can also add NODES RENUMBERING which will give you a better numbering to solve system.

Hope it helps.

T.

Hello

Thank you I will give it a shot. Will let you know if it works.

Btw. I used OceanMesh2D to create my mesh and they have an option to cut such triangles as well. This work very well for my case so far, but maybe stbtel.py does it as well.

Hej Thierry!

I'm working on another project now and finally thought of using stbtel.py with the keywords you mentioned.

However, I do not get it to run as it always ends up with the same error:

ERROR 1003 DURING CALL OF OPEN_BND:NGEO
ERROR TEXT: HERMES_FILE_NOT_OPENED_ERR
ERROR WITH IF 1 : OPEN_BND_SRF:GET_OBJ

I'm not sure how to treat this error, since I actually do not know what it means. Do you may have an idea how to solve the issue?

C.

Hello,

I am not a specialist on this but i would say there is a problem with your boundary file and it can't open it. Did you forget to specify a boundary file, does it exist ?

Hello,

Thanks for your answer.
That was also my thought. I added the boundary file, which basically only has closed boundaries atm. I just add the .cas file here it could be that I miss something necessary to run the stbtel.py.

Hello

i tried to run stbtel on your cas file (just changing UNIVERSAL FILE and BOUNDARY CONDITIONS FILE by the geo.slf and geo.cli of the test case telemac/examples/stbtel/refinement/) and it didn't find any error so there is nothing missing un your cas file. Though i would avoid to put twice

GEOMETRY FILE FOR TELEMAC = 'geo_Bolmen.slf'
and
GEOMETRY FILE FOR TELEMAC = 'geo_Bolmen_1.slf'

But it doesn't seem to be the problem.

I think it is something with your mesh.