Diagnosability Analysis Of Labeled Petri Nets Based On Basis Markings

In complex industrial systems,no matter how safe a system design is,how advanced a control strategy is and how comprehensive the automation is,faults are almost unavoidable(which are deviations of a system from predefined specifications).Broadly speaking,faults include any unusual phenomenon in systems.Faults will not only reduce the reliability of a system,but also bring economic losses and even catastrophic consequences.Therefore,fault detection and diagnosis technologies have very important theoretical and practical significance.However,faults in a system can be detected and diagnosed only if the system is diagnosable.This work addresses the problem of diagnosability for discrete event systems,i.e.,if it is possible to reconstruct the occurrence of fault events by observing words of finite length.As a prior condition,diagnosability is crucial for modeling and deducing the unspecified behavior.Based on labeled Petri nets,the diagnosability is analyzed in a single-fault system and a multifault system.Diagnosability of a system can be categorized into centralized-diagnosability and decentralized-codiagnosability.The former describes diagnosability in the case of centralized structures.Centralized structures mean that only one site S exists such that it receives all information from a system.The latter describes diagnosability in the case of decentralized structures such as communication networks and manufacturing systems,where the information in these systems is distributed.Such a system contains multiple sites Si,i=1,2,...,which observe the system behavior based on the information provided by the sensors connected to it.The main contributions are introduced as the following:1)To avoid the state explosion problem during the procedure of obtaining reachability states,we use the notion of basis marking to develop diagnosability analysis of labeled Petri nets.Two oriented graphs are constructed,called an extended basis reachability diagnoser and a basis reachability diagnoser,respectively,which allow us to decide the diagnosability of a labeled Petri net.The former is an automaton that contains basis markings tagged with diagnosis flags and the latter is an automaton that contains diagnosis nodes and symbols.The structure of the two diagnosers depends on the net,and the languages of the two diagnosers are related with that of the labeled Petri net.Each node of the basis reachability diagnoser contains diagnosis decisions for each fault transition.2)The notion of uncertain cycle and the diagnosis value associated with uncertain cycle are defined.The diagnosis value is computed by a state equation in an extended basis reachability diagnoser,which is important to this research.3)Necessary and sufficient conditions for diagnosability and codiagnosability of bounded labeled Petri net systems are proposed.Specifically,we have to check whether each cycle in a basis reachability diagnoser satisfies particular conditions.For the decentralized case,the coordinator receives diagnosis result from each site and makes the final diagnosability decision.