Research On Unobservable Behaviour Of Discrete Event Systems Based On Petri Nets

Discrete event systems(DESs)are a kind of dynamic system driven by events and with discrete states.With the development of modern industrialization and intelligence,DESs play a more and more important role in human production,life and other activities,and the complexity of the system is gradually increasing.In some control systems,due to insufficient observation technology,limited observation cost or system failure,some unobservable behavior may inevitably appear in the process of system dynamic evolutions.Unobservable behavior is usually unable to be controlled,causing hidden dangers to the stable and safe operation environment of the system.It is of great academic value and engineering significance to study the problem of unobservable behavior recognition of discrete systems.In this thesis,Petri nets and automata are used as the modeling and analysis tools of a system,and the unobservable behavior is regarded as the unobservable places in Petri nets.By analyzing the input and output sequences of a system,we can find the unobservable behavior that may exist in the system,and finally construct a new model that can describe the system behavior with certain accuracy.The main contributions of this thesis are presented as follows:1.This thesis presents an analysis method of Petri net model accuracy based on reachable graph,which can calculate the additional sequences of different lengths that the model can produce,and verify whether the model is k-complete.An algorithm for identifying the observable behavior based on integer linear programming is presented.The algorithm uses observation sequences to construct automata describing the evolution of system states,the illegal transitions that need to be banned from launching are analyzed.The recognition of the observable behavior is transformed into linear programming described by mathematical language,and a k-complete Petri net model is finally obtained.2.In order to simplify the Petri net model,an optimization of structures is proposed to solve the problem of model complexity in the recognition algorithm based on integer linear programming.First,the calculation of illegal transition set is proposed,where the computation process of the observable database does not need to consider the dynamic changes of Petri nets.Then,the linear constraints and objective functions are improved,as a result an unobservable place can prohibit more illegal transitions fire and reduce the complexity of the resulting Petri net model.3.The advantages of the algorithm based on integer linear programming in model accuracy and the effectiveness of the structural optimization method for the complexity of the model are verified by a specific example.