On Structure-based Deadlock Control For A Class Of Petri Net Models With Varying Resource Numbers

The work in this thesis takes automated medical analysis systems as research objects and an application background.From the perspective of system resource allocations,it investigates formal models at the logical level of discrete event systems.It uses a subclass of Generalized Petri nets to model the automated medical analysis systems,in which numbers of system resources may vary.It analyzes logical behaviors of systems and proposes a new deadlock prevention strategy.The main contributions of this thesis are as follows:1.We simplify the structures of Liveness and Ratio-enforcing Supervisors(LRSs).The LRSs proposed in the previous work are based on the inherent live structures and initial states of systems.By adding LRS controllers,the allocations of shared resources are controlled,thereby preventing deadlocks caused by resource cycle waiting.This work analyzes the reachability graphs of LRS controlled systems and finds out that there are unnecessary parts in the structures of the original LRS controllers.Thus,the structures of LRS controllers can be simplified by removing some connecting arcs.In addition,this work gives an equivalent expression based on inhibitor arcs of the simplified LRS controllers.It also analyzes the situations,in which the simplified LRS controllers allow the controlled systems have more permissive behavior.2.We extend the control parameters(initial markings)of simplified LRS controllers from constant to variables.The initial markings of the simplified LRS controllors are set as variables;according to the control mechanism,the functional relationship between the controller parameters(dependent variables)and the initial markings of the resource places(independent variable)to be controlled is established.When the markings in the resource places abnormally change,the markings in the controllors can be adjusted accordingly,so that the status of the resource can be monitored in real time and a reasonable allocation can be made to achieve a certain degree of adaptive control.3.We proposed a control strategy for switchable controllers.The strategy performs structural analysis on the system models under consideration,adds control places to all resources in the system first,and then selects a group of control controls from all with the best control effect for actual control.The remaining control places are in a standby state,with no effect on the current system behaviors.When failures occur,and the number of a resource continues to decrease to a certain degree,the corresponding controller may not control the failure due to reaching the critical value.At this time,it should be taken out of control,and a standby controller should be chosen and activated,thereby making the system still maintain an operation to a certain degree.This thesis dwells upon the proposed control method and uses multiple examples and corresponding data to explain.It uses Tina,INA,CPN-Tools and other professional software tools to verify the proposed work,and preliminary shows the feasibility and effectiveness of the proposed deadlock prevention method in this paper.