Petri Net-based Modeling And Control Of Cluster Tools For Concurrently Processing Two-type Wafers

In recent years,single crystal silicon that is the key crystal material and some related advanced industries have become the strong power for rapid development of information technology.Meanwhile,it has become an important industry to support social economy and ensure national security.Due to the high cost,weak technology and the low rate of products passing in domestic enterprises,the development of wafer manufacturing industry is restricted.In order to improve the technology of wafer fabrication,cluster tools are widely used in semiconductor manufacturing.Such tools are highly integrated and automated equipment with unique single-wafer processing technology.It can effectively deal with serious manufacturing requirements and other complicated process.With limited space between the process modules(PMs)in cluster tools and mass customization,wafer fabrication tends to arrange different wafer types with similar recipes in a lot.When multiple wafer types are processed in the tools,the system is deadlock-prone.Thus,it is great of importance on the processing of multiple wafer types in cluster tools to find an efficient method.Firstly,the deadlock analysis of the system is conducted,and then two deadlock avoidance rules are developed.Secondly,this thesis uses Petri net with time to model cluster tools as well.With the asynchronous concurrency property of Petri nets model,it can describe multiple products concurrent processing well in discrete event systems.Thus,it has been widely used to model manufacturing systems,such flexible manufacturing systems.Subsequently,it has been proved deadlock-free.This Petri net model not only can be applied for single-arm cluster tools with concurrently processing two wafer types in a lot,but also can be applied for scheduling problem with multiple wafer types.Finally,this work develops a scheduling algorithm according to earliest starting strategy.We call it PN based-earliest starting strategy.Moreover,particle swarm optimization(PSO)has been applied to solve the scheduling problem.Based on ROV strategy,the sequential of wafers can be obtained by the value of particles in solution spaces.In additional,based on the developed two scheduling algorithms,programming is finished by Matlab software and some experiments have proved the availability of two algorithms.Under the same experimental conditions,the results show that makespan from PSO is much smaller than using the algorithm of PN-based earliest starting strategy.But,the PN-based earliest starting strategy can be used for real-time scheduling.Additionally,the latter can apply for other scheduling problems in discrete manufacturing system(DMS)according to the universality of this PSO.It can provide effective scheduling method for DMS.