| 2007 Fault diagnosis in the state-based discrete-event system framework is studied. First, the resonating-cycle test for diagnosability of centralized diagnosers is derived, leading to an efficient algorithm which significantly reduces computational complexity. Second, design parameters of agents and communication protocols of decentralized state-based diagnosers are examined. Third, basic principles for state estimation are derived, and a simple condition to test consistency of state estimators is determined. Finally, three decentralized state-based diagonsers (DSBDs) are studied in detail. DSBD A involves communication of observations. It can be decoupled into subsystems which can be analyzed using earlier methods. Algorithms to optimize DSBD A networks are examined. DSBD B involves communication of observations and confirmed faults. DSBD C involves communication of observations and fault estimates. For both DSBD B and DSBD C, a consistent state estimator is specified, and diagnosability conditions stated. A structure-based diagnosability test is also derived for DSBD B. |
