Cross Validation of Aerodynamic Shape Optimization Methodologies for Aircraft Wing-Body Optimization

Abstract

This paper presents the cross validation of two aerodynamic shape optimization methodologies in order to validate, characterize, and compare the two methodologies. Both methodologies use gradient-based optimization based on the Reynolds-averaged Navier–Stokes equations driven by the discrete adjoint method. The first methodology uses a B-spline surface representation of the geometry coupled with free-form deformation and axial curve-based geometric control, as well an efficient linear elasticity-based mesh deformation method. The second methodology uses class-shape transforms for airfoil geometry control and either in-CAD or out-of-CAD surface modeling. Two benchmark geometries are used for comparison: a wing-body configuration representative of a large transport aircraft, and a wing-body configuration with aft-mounted nacelles typical of a small business jet. Both single-point and multipoint optimizations are performed. Cross validation is conducted such that flow evaluations are performed on the geometries produced by the two methodologies using a single flow solver and consistent meshes. It is found that both methodologies provide similar geometries and performance improvements. The primary differences that exist are a result of the distinct geometry parameterization and control methods that, in some cases, allow for more localized geometric control. The degree of agreement is indicative of the maturity that has been achieved by modern state-of-the-art aerodynamic shape optimization methodologies.