Diagnoser reduction of discrete-event systems

Abstract

In order to ensure the reliable operation of Discrete-Event Systems, several works in the literature propose the implementation of diagnosers with the view to detecting and isolating unobservable fault events within a bounded number of event occurrences. The efficiency of the diagnosis method can be measured by the delay of diagnosis, defined as the largest number of event occurrences after the fault until its detection. The main drawback of implementing the diagnoser proposed in the literature is that its state set can be very large, requiring a great amount of memory to be implemented in complex systems. In this work, we propose an algorithm for the computation of a deterministic reduced diagnoser, that preserves the diagnosability of the system language and the same diagnosis delay than the original diagnoser. We show that the proposed reduction strategy can lead to a reduced diagnoser with fewer states than by using other strategies proposed in the literature.