Real-Time Health Management Supervisors for Autonomous Fault Detection/Accommodation and Decision Support

"Impact Technologies, in cooperation with the Georgia Institute of Technology, propose the development of an autonomous health management supervisory software capable of fault detection, isolation and accommodation utilizing adaptable models implementedthroughout all regimes of vehicle operation. Utilizing an integrated and hierarchical software architecture that is suited for adaptive modeling and control, and can simultaneously support health management reasoner development, a novel system healthsupervisor will be demonstrated for autonomous optimization of vehicle operations. Specifically, through accurate vehicle regime recognition and fusion of health management and controls logic, adaptable data-driven and physical models will be utilizedwith intelligent reasoners to provide automated response analysis to changing environmental and system degradation/failures in real-time. The proposed program will develop a generic health management (HM) supervisor architecture with a modeling paradigmthat can be extended to various air/ground, manned/unmanned vehicle applications. The end goal is to enable a hierarchical supervisor and design paradigm that allows for full operational coverage of system/vehicle faults and degradation, but also supportrigorous system testing/validation. A realistic demonstration of the HM supervisor will be developed for an unmanned rotorcraft application under a threat engagement scenario. With the successful developments and implementation of this proposed Phase Ieffort, it is strongly anticipated that the health management supervisor technologies will be able to be applied in a variety of DoD ground/air vehicle applications. Hence, during the follow-on Phase II effort, commercial-grade HM Supervisor software willbe available for specific design applications. Although the developed HM supervisors will first be tested utilizing a UAV application, the potential for use of the developed technologies is much broader. With the generic supervisor modeling architectureproposed, the developed software modules can