Model-Based Fault Detection and Identification System for Increased Autonomy

摘要

This paper describes the Model-based Off-Nominal State Detection and Identification (MONSID) diagnosis system and proof of concept demonstration in a simulation environment. Nominal models of the application system stimulated by sensor and command data are processed by the diagnostic engine. Faults are detected when inconsistencies arise between the modeled behavior and sensor data. The prototype system was applied to satellite attitude control system hardware and evaluated with simulated nominal and fault data. MONSID correctly diagnosed all seven fault scenarios demonstrating the soundness of the fault detection and isolation algorithms. Improvements to MONSID's performance include techniques to reduce the amount of processing required to correctly identify faulty hardware components and reducing the potential for false negatives during fault detection.