Development of a 2d discrete turbulent adjoint solver using automatic differentiation

This paper presents an implementation of discrete adjoint solver using an automatic differentiation tool for a two-dimensional Reynolds Averaged Navier-Stokes (RANS) finite volume solver. Moreover, automatic differentiation tool is also utilized calculating flux Jacobian required by implicit finite volume solver. The developed RANS adjoint solver is computationally efficient and accurate. In the present implementation, the time required for the calculation of adjoint variables is comparable to the time required for one iteration of the flow solver. The RANS adjoint is verified with a brute force method. Finally, the aerodynamic shape optimization capability of the developed RANS adjoint solver is demonstrated by optimizing a RAE 2822 airfoil in terms of drag at constant lift coefficient