[Zen]

Dear Developers,

I was testing the SU2_DEF capabilities in terms of deformation of a mesh at the wing-body corner. My wing is a Rood wing and my body is a simple flat plate attached to it. The mesh is completely structured and more info about the case can be found here:

http://cfd.mace.manchester.ac.uk/cgi...html&1&0&0&0&0

I want to perform a deformation of the flat plate in the spanwise direction only (y-axis). Attached is the configuration file I used.
To test SU2_DEF I moved one design variable in the spanwise direction and checked the surface mesh deformation and the volumetric mesh deformation (see figures). The surface mesh deformation is fine, however, the volumetric mesh deformation produces negative volume cells from the first linear iteration.

I then reduced DV_VALUE of a orderd of magnitude every time until I got no negative volume cells. However, this happened for a DV_VALUE=-0.00001 meaning the deformation of the flat plate is negligible. The aim of my work is to optimize the junction area and I am expecting large deformations of the body in that area. Hence, I can't afford having such a small DV_VALUE (which, if I understood correctly, it will take the value of the scaling factor of the OPT_OBJECTIVE term) because it would take an awfully large number of cycles to get to the final shape and I expect negative volume cells to pop up anyways at a certain point.

Do you have any suggestions on how to solve this problem?

Thank you very much,

Zeno

[fpalacios]

If I were you I would try the latest version of SU2 (v3.2.7) and

DEFORM_NONLINEAR_ITER= 10
and
DIVIDE_ELEMENTS= YES

Best Regards,
Francisco Palacios
SU2 Lead developer

[Zen]

Dear Francisco,

Thank you for your help. I have applied the changes you suggested and now SU2_DEF is less prone to produce negative volume cells.

I have then started a full optimization using shape_optimization.py and two converged solutions for the direct and drag adjoint simulation. However, negative volume cells appear again after 5 or 6 optimization cycles and the optimization stops with the divergence of the FGMRES solver. I have tried to reduce the scaling factor in OPT_OBJECTIVE= DRAG*scale.
For scale=0.1,0.05,0.01, negative volume cells appear between DSN_005 and DSN_007 and the FGMRES solver diverges.
For scale=0.005,0.001 the optimization stops at the end of DSN_001 with the "Current function value" equal to the drag value at the end of the direct simulation multiplied by the value of the scaling factor e.g. 0.392647*0.001=0.000392647.

The geometry is the same as above. I enclose a copy of my configuration file. I am now trying to modify some of the "GRID DEFORMATION PARAMETERS" but every time is a shot in the dark as I am not sure how they will affect the deformation. Hence, any help regarding this problem would be appreciated.

Thanks!

Z