Grasping extreme aerodynamics on a low-dimensional manifold

Area of Science:

• Aerodynamics
• Fluid Dynamics
• Machine Learning

Background:

• Modern air vehicles face challenges in gusty environments like urban canyons and mountainous terrains.
• Increasing extreme weather events due to global warming necessitate stable flight capabilities for aircraft, especially smaller ones.
• A lack of theoretical fluid-dynamic foundation exists for describing extreme vortical gust impacts on wings.

Purpose of the Study:

• To investigate the fundamental physics of extreme aerodynamics and gust-wing interactions.
• To develop a method for simplifying complex, high-dimensional aerodynamic flow fields.
• To enable stable flight for next-generation air vehicles in challenging atmospheric conditions.

Main Methods:

• Utilized a lift-augmented autoencoder, a machine learning technique, to compress nonlinear vortical flow fields.
• Analyzed the parameter space of gust-wing interactions.
• Demonstrated the ability to represent complex flows with a low-dimensional manifold.

Main Results:

• The study reveals that extreme aerodynamic flows are fundamentally simpler and lower-rank than previously assumed.
• Nonlinear vortical flow fields can be compressed into just three key variables using the developed autoencoder.
• The machine learning approach allows for real-time sparse reconstruction, dynamical modeling, and control of unsteady gusty flows.

Conclusions:

• The findings provide a theoretical basis for understanding and managing extreme aerodynamic phenomena.
• The developed low-dimensional manifold enables real-time control strategies for gusty flow conditions.
• This research supports the stable flight of future small air vehicles in previously unflyable atmospheric conditions.