Abstract

.This chapter presents the salient features of a general methodology for fault diagnosis in partiallyobserved
finite-state automata and its application to automated manufacturing systems. The system of
interest is modeled as a set of interacting automata coupled by common events. The total event set
comprises observable and unobservable events, reflecting those events that can, or cannot, be perceived
by the sensors attached to the manufacturing system. Fault events are inherently unobservable and the
diagnostic task is to infer their occurrence from the sequences of observable events and the system model.
On-line diagnosis is performed using diagnoser automata, which are constructed from the system model
either off-line or on-the-fly. The analysis of the diagnosability properties of the system is done off-line
using verifier automata, which are also constructed from the system model. The algorithms presented
are illustrated with relevant examples. The chapter concludes with a discussion of sensor selection for
diagnosability and of cooperative diagnosis for systems with decentralized information.