Comparing Different Off-the-Shelf Optimizers' Performance in Conceptual Aircraft Design

This paper covers the comparison of multiple optimization algorithms applied to SUAVE, a conceptual aerospace vehicle design tool. Using an aircraft similar to an Embraer E-190 as the initial conditions, we compare the optimum aircraft created and the convergence criteria of different optimizers. Our comparison is not only to find what optimizer works best for a specific problem, but to better understand the convexity, multi-modality, and discontinuities of the aerospace vehicle design space. We look at four different two variable optimization exploring aircraft geometry-geometry, geometry-mission, and mission-mission variable optimizations. In our studies, the surfaces did not have many local minimas so gradient-based optimizer found the same optimum point in approximately an order of magnitude less function evaluations than a population-based method. The full aircraft was also optimized. A little less than thirty percent reduction in design mission fuel burn was found by a gradient-based optimizer when changing only the geometry and an additional one percent was found by exposing the cruise segment mission parameters.