An Improved Method For Think-Globally-Act-Locally Approach In Generalized Petri Nets Modeling FMSs

Flexibility manufacturing systems(FMSs)have been widely used in manufacturing industries.In an FMS,different processes utilize the limited number of system resources to finish different kinds of jobs.In this case,it may lead to deadlocks if the system resources are not properly assigned.When some processes wait endlessly for resources to be released by other processes,the system will fall into a deadlock.Deadlock means system blockage and a low efficiency of system,so it is a highly undesirable situation.Petri nets are widely used as an important tool for modeling,analysis,and control of FMSs.Based on Petri nets,deadlocks are usually controlled by adding a set of control places to enforce some necessary constraints on the system.The advantage is that a supervisor can be obtained off-line in a static way.Thus,there are a series of deadlock control policies proposed for FMSs based on their Petri net models.Based on the Petri net models,deadlock analysis and control techniques are divided into two categories: structural analysis and reachability graph analysis.The former is always based on some special structures of Petri nets,such as siphons and resources circuits.It is usually so conservative that some legal markings are forbidden.The later can always obtain optimal or near-optimal supervisors but it requires generating all reachable states.Therefore,it has to encounter the state explosion problem when dealing with complex systems.This thesis focuses on developing an efficient method to design optimal supervisors in complex systems.The following research contributions are developed in this thesis.1.For complex systems,the think-globally-act-locally method(TGAL)and the thinkglobally-act-locally method with weighted arcs(TGALW)have been developed to prevent deadlocks in an iterative way.On the other hand,a vector covering approach is developed,which can significantly reduce the number of considered first-met bad markings(FBMs)and legal markings.Meanwhile,it can obtain a set of optimal supervisors by solving integer linear programming problems(ILPPs).This thesis develops a method to synthesize TGAL and the vector covering approach.A temporarily global idle place(GP)is introduced to prevent the deadlocks in an iterative way.At each iteration,the vector covering approach is applied to design a set of optimal control places.Thus,all legal markings are reachable but no FBM survives at each iteration.As a result,the proposed method can lead to more reachable markings than TGAL and TGALW.2.By using the previous iterative method,there exist some Petri net models that are not optimally controlled.The reason is that some legal markings do not appear but they are prevented by the supervisor obtained in previous iterations.For this case,we provide an improved method that is also based on TGAL and the vector covering approach.For each FBM,the markings in the set of FBM-related legal markings are modified to guarantee that every place has the maximal the number of tokens as a legal marking.Thus,we can obtain a set of supervisors with the maximally permissive since no legal markings of system are prohibited.Finally,a generalized Petri net model can be optimally controlled.