[lazzini]

Hello,

I am trying to run a shape optimization of a two-phase nozzle. I am starting with a simple inviscid case. I am currently working on the branch feature_turbo2phase.

The code seems to work fine, but the problem is the mesh created through SU2_DEF.
As a check, we re-did the mesh with another branch, feature_turbomachinery, and everything works fine. When comparing the two meshes, we saw that the connectivity of the first one is completely off on the wall.
Due to this, even when running finite_differences.py the mesh is not deformed properly, and the gradients in output are always 0.

I attach here my config file Moore.cfg, the initial mesh inv_mesh.su2, and the two meshes created, nozzle_ffd_feature_turbo2phase and nozzle_ffd_feature_turbomachinery.

I did not change anything outside SU2_CFD and the config_structure.cpp, but the first merge with feature_turbomachinery was made almost a year ago, maybe something went wrong.

Did someone encounter this problem? If I do the mesh with another branch, the code itself is working fine, and does what it has to do. The problem seems limited to the command SU2_DEF.

[talbring]

Hi Lucia,

it is not the connectivity that is different, it is the global index of the points associated to the FFD box which differs in both cases. That is weird, I just had a quick look at some diffs, but I cannot find the reason for that. With the feature_turbomachinery branch finite differences are working ? And how does it behave with the current develop ?

Thanks,
Tim

[fpalacios]

Tim is right, if I were you I would take a look at
void CSurfaceMovement::WriteFFDInfo(CSurfaceMovement** surface_movement, CGeometry **geometry, CConfig **config) { }
void CSurfaceMovement::MergeFFDInfo(CGeometry *geometry, CConfig *config) { }

The problem may be related with Get_PointIndex or GlobalPoint[iFFDBox][iSurfacePoints]

Best,
Francisco

[lazzini]

Thanks for the replies.

First, yes, with the branch feature_turbomachinery everything is working, finite differences and adjoint. I checked the CSurfaceMovement, everything seems ok.

However, I managed to solve somehow the issue. It's more a trick than a solution, I went on the feature_turbomachinery branch, configure everything and then switch to the 2phase branch and compile.
I don't know why, but in this way it works. So the code itself seems to be correct.

I'll try to have a look and make some tests to find the source of the error, but in the meantime I can make my simulations

[fpalacios]

Great! if you are developing SU2 do not forget to take a look at www.su2devsociety.org.

Best,
Francisco