search

firemodels / fds

Fire Dynamics Simulator
https://pages.nist.gov/fds-smv/
Other
625 stars 610 forks source link

Issue FDS 6.9.1 - Program Exception - access violation #13111

Closed Perry-FSC closed 2 days ago

Perry-FSC commented 6 days ago

Good morning, I'm new to this forum, so I apologize in advance if I haven't mastered all the features. I'm having issues with a simulation I'm currently working on. A fire model with 701k cell with a mesh divided into 64 parts, corresponding to the 64 physical cores of my computer. cell size, initially, 0.5x0.5x0.5m.

I want to carry out a 900s modeling, but systematically at 500s or 600s, FDS crashes with the following error message:

".... Time Step: 6700, Simulation Time: 500.92 s Step Time: 6800, Simulation Time: 506.08 s forrtl: severe (157): Program Exception - access violation PC Image Routine Line Source fds_openmp.exe 00007FF688D65B11 Unknown Unknown Unknown fds_openmp.exe 00007FF688A67BC9 Unknown Unknown Unknown fds_openmp.exe 00007FF688A25DCB Unknown Unknown Unknown fds_openmp.exe 00007FF6893DEBB0 Unknown Unknown Unknown KERNEL32.DLL 00007FFA17FD257D Unknown Unknown Unknown ntdll.dll 00007FFA1914AF28 Unknown Unknown Unknown"

I attach the fds file as an appendix. Hope is is accessible. Can anyone tell me about the error message? Many thanks in advance.

Desktop :

SC0_7+polycarbonates.txt

mcgratta commented 6 days ago

I will run the case.

mcgratta commented 6 days ago

The case failed for me too. I'll try to see what is wrong.

Perry-FSC commented 6 days ago

Many thanks. I retry with less mesh (32 meshes of 22k cells associated with 4 logical cores). With this configuration the fire is fully enclosed in 1 mesh. I now reach 625s, another 1 hour of simulation and we will see

Perry-FSC commented 6 days ago

It's done. I have now reached 900s of simulation in 1h43. It works much better now with 32 meshes instead of 64. It seems that the 126m² fire vent spread over two meshes is causing it to malfunction.

Thank you Kevin.

mcgratta commented 6 days ago

Actually, there are several possible reasons for the crash. BURN_AWAY is always a possibility, but in your case, when I removed it, the code still crashed. Then I removed THICKEN=T and that prevented the crash and the case ran to completion. I am not a fan of THICKEN. It's an old feature and really not necessary anymore except in very special cases. In your case, I suspect that various obstructions were thickening into each other, causing confusion when the obstructions were removed or burned away.

I'll use this as a test case to see if I can pinpoint the exact cause.

Perry-FSC commented 6 days ago

Ok thank you Kevin. Let me know your results