Gradient free aerodynamic design of laminar airfoils and wings using high dimensional Bayesian optimization

One major focus of enhancing the fuel efficiency in aviation lies in implementing drag reduction technologies. There exists a significant potential to reduce the viscous drag to facilitate this objective by delaying the laminar to turbulent transition of boundary layers using hybrid laminar flow control. This study evaluates the potential of novel Monte Carlo Bayesian optimization techniques used in conjunction with trust regions for high-dimensional aerodynamic design of transonic laminar airfoils and wings using black-box transition models that lack gradient information. The aerodynamic analysis considered in this study is based on coupled Euler-boundary layer equations with a mid-fidelity conical 2.75D approach, which is more suited for the objective to balance the reduction of viscous and wave drag, in comparison to the conventional low-fidelity panel methods used for fully turbulent designs. A comparative analysis of the underlying acquisition functions governing the Bayesian decision policy is also presented in this work. The Bayesian design framework adapted in this study delivers a 98-dimensional laminar wing with very high aerodynamic efficiency and robust performance within a wide design range. Laminar extensions close to 50% chord lengths are achieved at both upper and lower surfaces of the laminar airfoils and wings when assisted with optimized leading-edge suction distributions, thereby also exhibiting significantly lower viscous drag, while maintaining the wave drag at a practically negligible level, which is a major challenge in laminar aerodynamic designs.