Design Of Robust Optimal Controller For Manufacturing Systems Based On Petri Nets

As an efficient and mathematical modeling and analysis tool,Petri nets have been widely used in the theoretical and applied research of automated manufacturing systems(AMSs)because of their advantages in model construction,state analysis,and deadlock control.A system is prone to deadlocks due to improper allocation of resources,which cause the system to crash.In order for AMSs to operate smoothly and safely,the deadlock problem in systems must be properly addressed.The existing researches based on Petri nets have given some analysis methods and control strategies to study the deadlock problem of AMSs.These methods are based on structural analysis or reachability state space.In the real world,the state explosion problem is inevitable for large-sized net models.In order to solve the problem,this thesis uses partial ordering techniques to simplify the state space of a model.The main contributions are introduced as follows:This thesis deals with partial order techniques to analyze the state space of Petri net based on persistent step graph,and a deadlock control policy is developed to design a liveness-enforcing supervisor.First,the partial order reduction rules are used to analyze the persistent step graph of model to reduce the state space.The number of states in the persistent step graph is less than the number of states in the reachability graph.Then,the state space of net model after reduction can be divided into two parts:dead-zone and live-zone.In order to retain all legal markings in the live-zone,the deadlock controllers can be designed by place invariants to forbid the markings in the dead-zone and retain all legal markings.Finally,after repeated iterations,we propose a deadlock control policy based on partial order techniques.It can compute the case where the legal reachability space is convex,and the controllers can be computed by using a self-loop method if the reachability space is non-convex.In order to obtain a more efficient control strategy to address the problem that resources of the system may fail,this thesis proposes a robust deadlock control method based on structure reduced Petri net and persistent step graph.For a GS3PR(generalized system of simple sequential processes with resources)with unreliable resources,the markings in state space of persistent step graph are divided into robust legal markings and forbidden markings.By adding monitors,all robust legal markings are reachable and the rest of markings are unreachable.Hence,the resource failure problem is transformed to a prohibited marking problem and the robust liveness-enforcing supervisor can be designed by place invariants.Regardless of whether the resources break down,the controlled net is live.Finally,several examples are presented to illustrate the developed control methods,which are able to keep the controlled net live by adding monitors while retaining more states of live zone through multiple iterations.At the same time,this method can keep a GS3PR with unreliable resources live regardless of whether the unreliable resources fail or not.Com-pared with the existing classical methods,this method has the advantage in computational complexity.