[Parth Kumar]

Hello,

I am a researcher at Georgia Tech and am using SU2 for some work here. I wanted to improve my understanding of a few parts of SU2, mostly regarding the fixed CL mode and the discrete adjoint, and so had a couple questions:

1. Regarding the EVAL_DOF_DCX option: When I run a direct solution, it would conduct a finite difference step at the end of the computation, regardless of whether I had this parameter as Yes or No. I also found that the option did not affect the discrete adjoint. What does this parameter actually do?
2. Partially out of curiosity, how is the gradient of drag (or moment) with a fixed lift constraint computed with only 1 adjoint calculation? I would have thought two sensitivities would be required. I could not find the applicable portion of the source code, though I may have overlooked it. I did some finite difference testing and found the gradient to match within 1% to the discrete adjoint, so I assume it is not an approximation?
3. My final question is on the surface_sens.csv file that is generated during the discrete adjoint calculation. Theoretically, if I manually compute the DX, DY, of the surface points on the boundary after a Hicks-Hennes or similar shape function perturbation and compute Sum(DX*Sens_X + DY*SensY) over all the mesh points this should return the correct gradient?
   
   I did this and the results matched for the several cases I ran, but thought to verify here once as well.

I would appreciate a response, thanks!


EDIT: I am using SU2v7



-Parth