abstract
- © 2016 IEEE.This work addresses the fault detection and isolation problem in systems modeled by Timed Continuous Petri Nets (TCPN) under the infinite server semantics. The proposed approach assumes that the normal behavior TCPN model is available and partially observed (i.e., only the marking of a subset of places is available). Potential product rejection faults and speed reduction faults are represented by transitions, whose firing characterize the occurrence of the corresponding faults. Based on this model, a novel structural characterization of diagnosable TCPN is presented. This characterization is based on the structural observability of places belonging to directed paths, between measurable places, containing information about the firing of the fault transitions. In order to isolate faults, a diagnoser is also proposed in this work. The diagnoser is based on a set of robust reduced order finite-time observers. The faults are isolated using internal signals in the observers that are different from zero when a specific fault occurs.