"Segmentation fault (11)" when writing solution of a 2D cylinder in euler flow SU2
 

Hello to everyone!

i've made a structured mesh in gmsh 3.0.6 of a 2D cylinder and i've exported it from gmsh with the .su2 format (enabling the option "save all"). (I've attached a picture of the mesh and the .su2 file)

When i run: "SU2_GEO cylinder.cfg", the mesh file is correctly recognized.

I've set an EULER flow with mach number of 0.1 with 0 AoA and standard reference pressure and temperature.

When i run: "SU2_CFD cylinder.cfg", the calculation proceeds well until the point when the intemediate solution is written ( i've set for example WRT_SOL_FREQ= 20), at this point the program stops with the following message:

-------------------------- File Output Summary --------------------------
Writing comma-separated values (CSV) surface files.
Loading solution output data locally on each rank.
Sorting output data across all ranks.
[TPE-E540:09518] *** Process received signal ***
[TPE-E540:09518] Signal: Segmentation fault (11)
[TPE-E540:09518] Signal code: Address not mapped (1)
[TPE-E540:09518] Failing at address: 0x3e16fe8



and i have to push "ctrl+c" to exit the process.

I've attached the config file "cylinder.cfg" if you want to test it with my mesh.

I've tried to set WRT_SOL_FREQ with a big number to see if the calculation proceeds well and i've notice that the residual value reach the convergence after 12261 iterations but i receive the same error (i think allways when it tries to write the solution).

I'm not using the multigrid (set to 0).



Have you got any idea about the origin of this problem?

Thanks in advance to everyone!

Andrea

In gmsh make sure to create volume zone for the fluid and enable "Save All" when exporting it to the SU2 format.

Tim

Thanks for your answer Talbring.

With "volume" i think you mean a "phisical Surface" for the 2D fluid domain, is it right?

And if you have defined phisical lines for the boundaries and phisical surface for the fluid i think that there's no need to activate "Save all", right?

For these case i've solved the problem in writing the solution generating a different mesh (i think i had too low quality of certain elements).

Then for the convergence i've experienced good results activating the LAX method.