Effect of gradient approximations on aero-structural gradient-based wing optimization

Abstract

This paper presents several studies in the context of gradient-based aerostructural optimization. The studies are motivated by the computational limitations and constraints industry has when attempting to run such ptimization. The aero-structure wing optimization problem engages two parts; one that improves the aerodynamic performance and the other reduces the mass under the critical load cases. To setup a realistic test case, the aerodynamic critical load cases required for sizing the wing of the XRF1 configuration are identified, with the help of high-fidelity approaches and ROMs, and are afterwards employed to size the wing. Then the magnitudes of the disciplinary and the cross disciplinary gradients are investigated for the sized wing to check which of them drive the aerostructural optimization under the given parameterization. It was found that these gradients might differ in two to three orders of magnitude. At the end the configuration is aero-structurally optimized with different levels of gradient accuracy.