Diagnosis of timed patterns for discrete event systems by means of state isolation

Abstract

This paper deals with the diagnosis of timed patterns for discrete event systems that are modeled by a particular class of timed automata. A timed pattern is a set of behaviors characterized by a sequence of events, occurring in a given order, and a group of time intervals, during which the underlying events occur. The problem addressed in this research is to detect the occurrence of such timed patterns by the analysis of the timed observations. For this purpose, we define a completely synchronous composition of a timed automaton and a given timed pattern. Then, a timed pattern recognizer and a diagnoser are designed, which are suitable to solve the timed pattern diagnosis problem as a state isolation problem. A necessary and sufficient condition for timed pattern diagnosability is also proposed from a timed-logical perspective.