[JanH]

Hi,

In order to perform an aero-structural optimization of a wing I need sensitivities of drag w.r.t. wing twist and wing vertical (Z) position at several spanwise positions on the wing. I can deform the wing at these spanwise positions easily using the FFD-boxes inside the SU2 code, however the problem arises when I try to obtain sensitivities w.r.t. these deformations.

I have successfully changed the deformation routine to the design variables I need (For example: local wing dihedral, all nodes at j_index=1 upwards, see the attached figure) and I use SU2_DOT to project the (converged) adjoint solution onto my design variable. However the sensitivity results I get from SU2_DOT are a factor 10³ too high when compared with the sensitivity gained with finite differencing of the same design variable (I use a displacement of the control points of 1E-4, unlike what is shown in the attachment).

Are more changes to the SU2_DOT necessary (apart from adding the new design variable) to be able to project the adjoint solution correctly onto multiple displaced FFD control points at once, controlled by the design variables explained above?

Any help is much appreciated,

JanH

[demanosalvas]

JanH,

Attached is a PDF that will give you a better idea on how to implement the adjoint for customized variables, and might be able to help you figure out if there are any bugs on your previous implementation.

Cheers,

David