Improving adjoint-based aerodynamic optimization via gradient-enhanced Kriging

Gradient-based optimization using adjoint method has proved effective for automatic aerodynamic shape optimization via high-fidelity Computational Fluid Dynamics (CFD) methods. Past experience suggests that its optimization efficiency and robustness are crucially affected by the step size along a direction of descending. In this paper, a surrogate modeling method based on gradient-enhanced Kriging is exercised to determinate the step size of adjoint-based optimization and a routine for adjoint-based aerodynamic design has been proposed. Representative results are presented for the inverse design of a RAE 2822 airfoil. It is found that the efficiency as well as robustness of adjoint-based aerodynamic optimization can be dramatically improved.