Online Fault Diagnosis of Discrete Event Systems Modeled With Labeled Petri Nets Using an Overall Fault Status

In this paper we present a fault diagnosis approach using labeled Petri nets, where the faults are modeled by unobservable transitions and the unobservable subnet is acyclic. In contrast to detecting the individual faults separately, a new specification called an overall fault status is introduced, which indicates the occurrence of faults from a global system perspective. Due to the introduction of the overall fault status, a more precise and informative diagnosis result can be provided and in some cases, the occurrence of some faults in a system can be detected before the actual faults are isolated, i.e., we are certain about the occurrence of faults but which faults have not been ascertained. An integer linear programming (ILP) problem is built according to the observed word. We prove that all transition sequences determined by solutions to the ILP problem constitute the set of sequences consistent with the observed word. By specifying different objective functions to the ILP problem, the diagnosis results of each individual fault and the overall fault status can be obtained. An online diagnosis algorithm is developed to implement the proposed diagnosis process, which reports the diagnosis results after the occurrence of every observable event.