Abstract
This paper presents a new decentralized fault diagnosis strategy for event-driven controlled systems such as Programmable Logic Controller (PLC). First of all, the controlled plant is divided into some subsystems and the time interval between observed successive two events in each subsystem is modeled as a random variable by Timed Markov Model. Second, the Bayesian Network, which represents the causal relationships among faults and observations in each subsystem, are introduced. By exploiting the Bayesian Network, the comutational burden for the diagnosis can be reduced. As a result, large scale diagnosis problems can be solved for practical situations. Finally, the usefulness of the proposed strategy is verified through some experimental results of an automatic transfer line.
