Neural-Network-Based Adaptive Flow Control for Maneuvering Vehicles

Conventional flow control methods employ actuation frequencies that are the same order as the characteristic frequency of the flow. Georgia Tech has demonstrated novel high-frequency synthetic jet actuation to modify the apparent aerodynamic shape of aerosurfaces and achieve quasi-steady flow reattachment over an otherwise separated airfoil. Practical application of this technology in dynamic maneuvers, however, will require closed-loop flow control that does not depend on an exact model of, or detailed measurement of, local flow phenomenon. In the USAF RESTORE program, GST, Georgia Tech and Boeing flight demonstrated neural-network adaptive control algorithms that can accommodate gross errors in the aerodynamic models used for design. Georgia Tech researchers have since extended this adaptive method to the case of output feedback. This program will demonstrate adaptive output feedback for control of synthetic jets to produce desired control moments without requiring detailed models of local flow phenomenon. In phase I, GST and Georgia Tech will produce a dynamic model of a 2-D airfoil equipped with synthetic jet actuators, and demonstrate in simulation adaptive closed-loop control of airfoil attitude. Phase II will provide a real-time wind-tunnel demonstration of this adaptive closed-loop flow control technology for a 3-D aerodynamic configuration that both rotates and translates at various low-speed conditions.