I am running a transonic case of a arbitrary body in a transonic flow to try and reduce the drag. The direct N-S (w/ SA) equations are solved fairly well, with residuals dropping to at least 10^-6. I have tried multiple numerical schemes to compare convergence rates and accuracy for the N-S. (All schemes compare well with one another)
The grid was built in pointwise with a structured surface mesh and TREX was used to build the grid in the BL. First point off the wall has y+ under 1.0. Stretching ratio of 1.2. ~10 million cells (5 million are hex). This grid has had good results in other solvers.
When I switch to the Cont. Adjoint, I get poor convergence and varying numerical scheme and CFL does not seem to improve the convergence. Has anyone had a similar problem? Any advice? My config and residuals history is attached.
(Note: This case was a restart, I raised the number of linear solver iterations)