Fault Diagnosis Of Discrete Event Systems Using Petri Nets

Discrete event system(DES)is a dynamic system driven by a sequence of events.With the increasing level of modern industrialization and intelligence,DESs exist in lots of important systems.When the scale of a DES increases,any fault can cause a devastating consequence.For example,fault can lead to property damage and casualties.Thus researching the fault diagnosis problem of DESs has significant impact on both academic and engineering.The character of DESs‘state space explosion',which means the number of the reachable states increases exponentially with the scale of a DES increasing,would make the fault diagnosis problem even worse yet.In order to avoid this explosion effectively,three different methods of fault diagnosis are put forward,and there is no need traverse the entire state space of the system.Furthermore,diagnostic efficiency of DESs can be improved.In this thesis,fault diagnosis problem of DESs is studied using the modeling tool—Petri net.Three methods are obtained on the basis of the structural characteristics of Petri net models.The main contents are summarized as follows:1?Aiming at a class of DESs modeled by acyclic Petri nets with partially observed transi-tions,a method is proposed to diagnose fault on the basis of the maximum number of tokens in special places.Firstly,by defining diagnostic network structure within the Petri net model,the di-agnosable range of the fault can be confined.Subsequently,three rules for simplifying the network structure are raised to extract the information of fault.Finally,in order to infer whether the fault event occurs or not,the flow pattern of tokens in the Petri net need be described.Meanwhile,the maximum number of tokens in the special places can be calculated benefiting from the structural information of the Petri net.From the perspective of the network structure,this method focuses on the information of a number of nodes in Petri net.Meanwhile,the computational complexity of fault diagnosis can be reduced through simplifying the structure of the Petri net.2?Aiming at a class of DESs modeled by labeled Petri nets with partially observed transi-tions,a method is propounded to diagnose fault on the basis of path constraints within the Petri nets.Firstly,four transformation rules for changing the network structure are derived with the use of the inherent properties of the labeled Petri net.Soon afterwards the multiple path constraints in the transformed Petri net can be obtained through the information of paths.Meanwhile,the path constraints can convert to the path constraint rules combining with the situation of tokens flowing into the paths.Finally,three different results of diagnosis can be inferred utilizing the path constraint rules.Furthermore,the possibility values of tokens flowing into the faulty paths can be calculated when the diagnosis result is‘uncertain'.This method can diagnose the fault event effectively by extracting the multiple path constraints within the labeled Petri net.3?Aiming at a class of DESs modeled by bounded Petri nets with partially observed tran-sitions,a fault diagnosis method is put forward to calculate the probability value of a fault on the basis of the information within a Modified-Access-State-Graph(M-A-S-G).Firstly,the rela-tionship of nodes in Petri net is reinterpreted owing to concept of the transition group,and an Access-State-Graph(A-S-G)can be constructed on the basis of it.Secondly,an A-S-G can be reconstructed to a M-A-S-G utilizing the structural information of the partially observed Petri net and the information of fault in system.Finally,the probability value of the fault can be calcu-lated through classifying the accesses and states of the M-A-S-G into different sets and deducing the flow of tokens backwards.This method maps the information of fault within a bounded Petri net into a M-A-S-G,which can avoid the interference of redundant information and reduce the computational complexity of fault diagnosis.The three methods of fault diagnosis mentioned above can avoid the problem of f state space explosion' in DESs with the use of the structural characteristics of Petri nets.Comparing with the traditional fault diagnosis methods,the best advantages of our methods are reducing the computa-tional complexity and improving the efficiency of fault diagnosis.Furthermore,our methods can meet the real-time requirement and have a certain academic value and engineering significance.