[AlbertoPi]

Hi,
I'm using SU2 for 3D simulations with 8M elements.

I'm trying to use mesh_adaptation.py. But SU2 diverges when I run it again with mesh_out.su2

The size of mesh_out.su2 file is 604MB and of restart_flow.dat 476MB only. There is a very high decrease from the initial mesh and restart files before the grid adaption. They were around 1.1-1.5Gb.

I tried to change:
ADAPT_BOUNDARY= to NO
and
SMOOTH_GEOMETRY= to NO

But it keeps diverging.
Could anyone say me what have i doing wrong?

Thx. Alberto.

[hlk]

Originally Posted by AlbertoPi

Hi,
I'm using SU2 for 3D simulations with 8M elements.

I'm trying to use mesh_adaptation.py. But SU2 diverges when I run it again with mesh_out.su2

The size of mesh_out.su2 file is 604MB and of restart_flow.dat 476MB only. There is a very high decrease from the initial mesh and restart files before the grid adaption. They were around 1.1-1.5Gb.

I tried to change:
ADAPT_BOUNDARY= to NO
and
SMOOTH_GEOMETRY= to NO

But it keeps diverging.
Could anyone say me what have i doing wrong?

Thx. Alberto.

Thanks for the question.
The output mesh file size should be very close to the initial mesh file.
I'm not sure what the problem is, but the first thing I think of is to check to make sure that WRT_VOL_SOL = YES (YES is the default value used if this option is not included in your config file).
When you say it diverges, does it run a few iterations and then diverge? Are there any other error messages output?
If it is simply that it starts running ok for a few iterations and then diverges eventually, you may need to decrease the CFL number and/or check that the output mesh is not doing something weird like producing negative-volumed cells.

[AlbertoPi]

Originally Posted by hlk

Thanks for the question.
The output mesh file size should be very close to the initial mesh file.
I'm not sure what the problem is, but the first thing I think of is to check to make sure that WRT_VOL_SOL = YES (YES is the default value used if this option is not included in your config file).
When you say it diverges, does it run a few iterations and then diverge? Are there any other error messages output?
If it is simply that it starts running ok for a few iterations and then diverges eventually, you may need to decrease the CFL number and/or check that the output mesh is not doing something weird like producing negative-volumed cells.

Hi, thanks for your answer Heather.
My config file didn't include WRT_VOL_SOL value. I tried to use it with YES value and it was still diverging. The simulation begins ok, and after a "few" iterations the CLift gets up to -1000 and CDrag up to 1000 slowly and then appears the message "SU2 has diverged".

I have no messages about no-physical points.
I use CFL=1.
The original mesh file is in CGNS format and has 255Mb. The mesh_out.su2 file has 605Mb. And in my expiriences the SU2 format is much heavier than CGNS.
The original solution_flow.dat file is 1.5Gb and the restart file after grid adaptation is 476.7Mb. If i use, after the adaptation, the SU2_SOL to ckeck flow.vtk file, it hasn't values about the flow.

Now I tried to use the mesh_out.su2 without the restart options to check only the mesh, and it works right. So, I can't use the restart file after the grid adaptation.

Thanks again. Cheers.
Alberto

[hlk]

Originally Posted by AlbertoPi

Hi, thanks for your answer Heather.
My config file didn't include WRT_VOL_SOL value. I tried to use it with YES value and it was still diverging. The simulation begins ok, and after a "few" iterations the CLift gets up to -1000 and CDrag up to 1000 slowly and then appears the message "SU2 has diverged".

I have no messages about no-physical points.
I use CFL=1.
The original mesh file is in CGNS format and has 255Mb. The mesh_out.su2 file has 605Mb. And in my expiriences the SU2 format is much heavier than CGNS.
The original solution_flow.dat file is 1.5Gb and the restart file after grid adaptation is 476.7Mb. If i use, after the adaptation, the SU2_SOL to ckeck flow.vtk file, it hasn't values about the flow.

Now I tried to use the mesh_out.su2 without the restart options to check only the mesh, and it works right. So, I can't use the restart file after the grid adaptation.

Thanks again. Cheers.
Alberto

Sorry, I thought from the initial comment that you were saying that the output mesh.su2 file was smaller than the initial .su2 file prior to adaptation. That is why I was recommending checking WRT_SOL to make sure that the volume portion of the mesh was written; that was the case by default and it sounds like that wasn't the problem.

Based on the additional details, it sounds like there is enough of a change between the initial and adapted mesh to make the restart solution far enough off that starting from scratch was a more stable computation. You can try using the CFL adaptation parameters with a minimum CFL less than 1 if you prefer to use the restart files. (See http://www.cfd-online.com/Forums/su2...parameter.html)

[AlbertoPi]

Originally Posted by hlk

Sorry, I thought from the initial comment that you were saying that the output mesh.su2 file was smaller than the initial .su2 file prior to adaptation. That is why I was recommending checking WRT_SOL to make sure that the volume portion of the mesh was written; that was the case by default and it sounds like that wasn't the problem.

Based on the additional details, it sounds like there is enough of a change between the initial and adapted mesh to make the restart solution far enough off that starting from scratch was a more stable computation. You can try using the CFL adaptation parameters with a minimum CFL less than 1 if you prefer to use the restart files. (See http://www.cfd-online.com/Forums/su2...parameter.html)

Thanks for you help.
I wasn't sure if my mesh.out was smaller that the initial mesh because the initial mesh is in CGNS format. Now i know that it's ok. But the new solution_flow.dat doesn't contain results.

I have attached screenshot of the strange values that i see in paraview.

I appreciate a lot your help