Machine Learned Surface Dynamics for Drag Reduction of Turbulent Compressible Airfoil Flow

• Applicants: Professor Dr.-Ing. David Emory Rival, since 10/2023; Professor Dr.-Ing. Wolfgang Schröder
• Subject Area: Fluid Mechanics
• Term: since 2016
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 298994276

In this follow-up project, a machine-learned surface dynamics (MacLSD) approach is developed to reduce the aircraft drag in particular over the wing sections. A key enabler is spanwise propagating transversal surface waves whose impact on lowering friction drag has been thoroughly analyzed in the previous period. Due to the direct link between drag and fuel consumption or energy conversion, this is an essential contribution to protect the environment independently from the type of propulsion system for any aircraft. In the first successful period of the project, a massive drag reduction of over 30% was achieved for generic subsonic turbulent flat plate flow using large-eddy simulations (LES). Even the net power savings accounting for actuation costs were around 10%. This result in drag reduction and net power saving prevailed for non-swept and swept flow conditions. A novel machine-learned response model revealed the essential actuation parameter setup and accurately predicted the best achievable drag reduction outside the hitherto considered range of actuation parameters. In addition, the first control-oriented reduced-order method for coherent structures associated with wall drag reduction was developed, a cluster-based network model. In the second period, the investigation will be extended from a turbulent flat plate to a turbulent wing flow. More precisely, the MacLSD approach is used to optimize drag reduction of turbulent transonic wing flow. Thus, besides friction drag effects caused by compressibility, pressure gradient, i.e., pressure drag, and wave drag will be considered. The variation actuation parameters of the moving surface, i.e., wavelength, wave period, and wave amplitude, will be optimized by a machine-learned response model and physical insights from control-oriented reduced-order models. Finally, the benefits of closed-loop control are explored in a cluster-based network optimizing nonlinear infinite-horizon control with a tiny fraction of first-principle linear optimal control. The model hierarchy from LES, reduced-order models, and response models will result in physics-founded engineering design principles of airfoil flow drag reduction extending the knowledge of the first funding period by the additional phenomena compressibility, pressure gradient, and wave drag.

• DFG Programme: Research Grants
• International Connection: China
• Cooperation Partner: Professor Dr. Bernd Rainer Noack
• Ehemaliger Antragsteller: Dr. Richard Semaan, until 9/2023