Liang Tang, Gregory J. Kacprzynski, Kai Goebel, Johan Reimann, Marcos E. Orchard, Abhinav Saxena, Bhaskar Saha
The term Automated Contingency Management (ACM) has been used to describe intelligent systems capable of mission re-planning and control reconfiguration based on information from health state diagnosis. While a diagnostics driven ACM capability designed to optimize multi-objective performance criteria remains a significant technical challenge, it is fundamentally a reactive paradigm. This paper introduces an automated contingency management paradigm based on both current heath state (diagnosis) and future health state estimates (prognosis). Including Prognostics in the control loop poses at least two additional challenges to ACM. First, future state prediction will, in general, have uncertainty that increases as the prediction horizon increases so adaptive prognosis routines that manage uncertainty are critical. Secondly, the larger time horizon afforded by prognosis requires ACM to be split into a real-time reactive component and a non-real time planning component. The latter considers temporal parameters and the potential impact of being proactive with mitigating action. The proactive ACM paradigm was developed and evaluated in the context of a generic mono-propellant system model in Simulink/Stateflow with diagnostics, prognostics and an optimal reconfigurable control system. Applications of Artificial Intelligence (AI) technologies in prognostics enhanced ACM system are briefly discussed. Preliminary results from on-going research work are presented.
Subjects: 1.5 Diagnosis; 15.3 Control
Submitted: Sep 10, 2007