On Consistency Of Diagnosability Of Composed Labeled Petri Nets Via Buffer Places

With the development of technology in modern industrial systems,system modeling,anal-ysis,reliability,and safety have become more and more complicated.Meanwhile,faults inevitably occur during the operation process.A fault is the deviation of system from ex-pected performance,which affects the safety and reliability of the system.In order to ensure the safe operation of the system,it is important to respond quickly to make appropriate de-cisions and controls.Fault diagnosis is a process of estimating and measuring the output information from the system used for determining fault type,size,location and time of oc-currence.Not all faults can be diagnosed in a finite step,which depends on the diagnostic strategy,time,system model and so on.Diagnosability is a system property that implies whether a fault can be detected after its occurrence within finite steps.It is crucial for a complex system to have the ability to deduce the unspecified behavior.Based on Petri nets(PNs),the diagnosability can be analyzed by a diagnoser.In recent years,there are many feasible methods for analysis of fault diagnosis and diagnosability,including centralized fault diagnosis,decentralized fault diagnosis,modular fault diagno-sis,and distributed fault diagnosis.Due to the state explosion problem,the design of fault diagnosers for large-scaled systems is computationally demanding and sometimes even im-possible.This paper focuses on complex systems where the components of the system have a very high level of concurrency and interact only via synchronizing condition.For such a dis-tributed system,it is necessary to establish a diagnoser for each part,and perform real-time interaction through established protocols,so as to analyze the global diagnosability efficient-ly.Specifically,modified basis reachability graph(MBRG)and basis reachability diagnoser(BRD)can be employed to provide some necessary and sufficient conditions of diagnosabili-ty.By analyzing the indeterminate cycles in different structures,we propose a new approach that can deduce diagnosability of global labeled Petri nets(LPNs)easily.New composition rules are proposed to compose the global LPN through some buffer places.Based on that,we can obtain enough information for the analysis of the diagnosability in the global LPNs by analyzing only one property of subnets.This work is devoted to reducing the complexity of diagnosability in a large system and providing new composition rules,which also gives a new solution for the diagnosability of complex systems.The main contributions of this work are summarized as follows.1.It provides a new composition approach.The main difference between our approach and other composition rules is that,in those works the authors assume that Petri nets are divided into different subnets or sites:each site is modeled as a subnet,which can interact with the other sites via shared places or transitions.In our approach,each site can interact with the other sites only via buffer places.It eliminates the limitation of common resources(including places and transitions),which can greatly increase the application of the fault diagnosis in the telecommunication and manufacturing systems.2.It gives an algorithm to analyze the possibility of breaking the indeterminate cycles in subnets.We obtain the diagnosability in a global net by analyzing whether the trouble traces forming the indeterminate cycles are survival in the net.The input are is added from buffer places to the observable transitions to analyze the possibility of destroying the indeterminate cycles.3.As an output of the algorithm,a set of observable transitions that are connected with the arcs from buffer place is provided.Based on that,we also propose a diagnosability function for obtaining diagnosability in the global net directly.Finally,we draw a conclusion and discuss future works which focus on the case where both subnets are non-diagnosable,and investigate the consistency of diagnosability.