Comparison of Gradient-Based and Genetic Algorithms for Laminar Airfoil Shape Optimization

Abstract

Reducing airfoil drag is a common objective to decrease fuel burn and emissions in aviation. Shape optimization tools have been shown to effectively design low drag airfoils. Two different approaches for laminar airfoil shape optimization are currently under investigation within the German research cluster SE2A (Sustainable and Energy Efficient Aviation). One is based on a gradient-free genetic algorithm and the other one uses a gradient-based adjoint algorithm. The performance of these two optimization toolchains is compared. In addition, a combination of the two methods by using the former in exploring the design space and the latter for further refinement of the solution is inspected. The RAE2822 airfoil was chosen as baseline airfoil.Two design conditions at transonic speeds and one at subsonic speed were investigated. The result of the adjoint method was found to be dependent on the initial reference airfoil, which limited the extent of drag reduction. In comparison, the genetic algorithm could explore diverse geometries and produce designs with longer laminar flow and weaker shocks, but was more computationally expensive. The combination of the two methods utilized the strengths of both and resulted in maximum drag reduction.