Probabilistic Fault Diagnosis of Event-driven System Focused on Sequential Order of Processes

Abstract

This paper presents a fault diagnose method for event-driven systems based on probabilistic inference, Timed Markov Model (TMM). The TMM is one of semi-Markov process models wherein state transition probabilities depend on stay time at a state. The TMM requires numerous amount of learning data to establish a diagnosis system, and it is difficult to collect enough learning data in a large and complex system. This paper focuses on a diagnosis strategy based on sequential order of processes in the event-driven system, and the method can diagnose the fault even if the amount of learning data is not rich. Finally, the usefulness of the proposed strategy is verified through some experimental results of an automatic transfer line.Key Words: event-driven system, fault diagnosis, Timed Markov Model, sequential order of processes 1.