Timed residuals for fault detection and isolation in discrete event systems

Abstract

In this paper a new attempt for fault detection and isolation in discrete event systems is proposed. An identified model constitutes a timed observer of the fault-free system behavior. Non-acceptable plant operation is detected by comparing the behavior of the model with the observed system output. For fault isolation, timed residuals and generic fault symptoms — early and late events — are introduced. Time bounds are composed using Boolean conditions and statistical analysis. In case of a fault, timed and untimed residuals are concluded in order to refine a set of potential faulty candidates. The method is applied to the given benchmark system of a virtual production plant with an external controller.