Unsteady fluid mechanics applications of neural networks

The capability to harness or alleviate unsteady aerodynamic forces and moments could dramatically enhance aircraft control during severe maneuvers as well as significantly extend the life span of both helicopter and wind turbine blade/rotor assemblies, Using recursive neural networks, time-dependent models that predict unsteady boundary-layer development, separation, dynamic stall, and dynamic reattachment have been developed, Further, these models of the flow-wing interactions can be used as the foundation upon which to develop adaptive control systems. The present work describes these capabilities for three-dimensional unsteady surface pressures and two-dimensional unsteady shear-stress measurements obtained for harmonic and constant-rate pitch motions. In the near future, it is predicted that such techniques will provide a viable approach for the development of six degree-of-freedom motion simulators for severe vehicle maneuvers as well as a foundation for the active control of unsteady fluid mechanics in a variety of systems.