Model Identification Of Partially Observable Discrete Event Systems By Petri Nets

Nowadays,the high requirements for product functionality and quality,and short production cycles require that the models we build for production systems have a high degree of accuracy.However,in the actual production process,the model may be inaccurate due to realistic conditions or environment,i.e.,the currently constructed system model is not able to accurately describe some behavior of the system.It is significant to find the minimum implementation model of the system that cannot describe the behavior on the basis of the current model.As a powerful modeling tool,Petri nets are widely applied in the modeling and analysis process of automated manufacturing systems.We call the behavior that can be described by the current system model as observable behavior,and the behavior that cannot be described by the model as unobservable behavior.Petri nets are used to identify the unobservable behavior in a system,which is usually described as a place or transition.According to the dynamic evolution process of a system,the unobservable behavior in the system is treated as transition sequences.We assume that a partial Petri net model that represents the observable behavior of a system is given in which all the transitions are observable.According to the observed sequence of observable transitions,this thesis uses integer linear programming to identify the unobservable behaviors in the system.The main contributions of this thesis are as follows:1.Due to the actual production process,some behavior in a system cannot be separated due to factors such as process or realistic conditions,i.e.,each variation in the sequence of observable changes cannot be considered individually.In order to better fit the actual production process,the algorithm proposed in this thesis adopts the approach of considering the observed sequence of observable transitions,instead of considering each transition in the sequence of observable transitions individually.Since we consider the whole transition sequence as a whole,we add fewer unobservable transitions to the same sequence of observable transitions.2.In the process of identifying the unobservable behavior in a system,we should not destroy the relevant properties of the original system(capacity of the storage,maximum capacity of the conveyor per delivery,maximum gripping capacity of the manipulator,etc.).And these properties are expressed in the Petri net model as the boundedness of the Petri net system.The algorithm proposed in this thesis can guarantee that the boundedness of the original system does not change after the addition of the unobservable transition to the bounded net system.