Questions regarding Discrete Adjoint Solver
 

Hi,

I tried simulating the Test case of Onera M6 using the SU2's discrete adjoint solver. I am new to the implementation of the sensitivity solver and trying to analyze the results generated from the implementation, and have the following questions-

1. What is the difference between the Surface_Sensitivity contours generated by surface_adjoint.szplt and surface_sens.szplt? Which ones are usually used for representing the sensitivity of the objective function wrt to design variables?

2. What does FINDIFF_STEP in of_grad_cd.csv represent?

I'm using v 7.1.1. I have attached the config file and gradient results as well.

I appreciate any inputs! Thanks!

Hi,


The first is the adjoint solution (from which sensitivities are derived, it's the objective function sensitivity w.r.t. flow solution), the latter is the objective function sensitivity w.r.t. surface mesh coordinates including mesh deformation, i.e. surface mesh coordinates as design variables.



FINDIFF is used for finite differences, adjoints don't use it.

Best, Ole