Advanced concepts for multi-objective evolutionary optimization in aircraft industry

Abstract

Evolutionary (multi-objective optimization) algorithms (EMOAs) are widely accepted to be competitive optimization methods in industry today. However, normally only standard techniques are employed by the engineering experts. Here, it is shown how these standard techniques can be completed and improved with respect to interactivity to other tools, runtime, and parameterization. The coupling with metamodels serves as an example for the interactivity to other tools, while the online convergence detection relates to runtime, i.e. stopping criteria. Finally, sequential parameter optimization improves results focussing on parameter tuning. We show that invoking all these methods on their own already enhances EMOAs for aerodynamic applications. It is concluded with an outlook on how these methods might come together to foster aerospace applications and, at a time, widen the application area to multi-disciplinary optimization tasks.