High-Fidelity Aerostructural Gradient Computation Techniques with Application to a Realistic Wing Sizing

Abstract

Aerostructural optimization is a keystone process to concurrently improve aerodynamic performance and reduce the structural mass of an aircraft. However, gradient-based multidisciplinary design optimization is efficient only if the computation of gradients is fast and accurate. To this end, two high-fidelity aerostructural gradient computation techniques are proposed for strongly coupled aeroelastic systems. In the specific context of this work, the focus is on design variables affecting structural stiffness only. Scalar objective functions like aerodynamic performance criteria are considered as well as a field of structural grid forces. The most intrusive technique includes well-established direct and adjoint formulations that require substantial implementation effort. In contrast, an alternative uncoupled nonintrusive approach is proposed that is easier to implement and yet capable of providing accurate gradient approximations. The accuracy of these methods is first demonstrated on the ONERA M6 wing test case. Their efficiency and applicability are then illustrated via a mass minimization problem applied to the Common Research Model. Both methods lead to very similar optimal designs, and the detailed analysis of results promotes the nonintrusive approach as a promising gradient computation alternative.