Optimal Scheduling Of A Multi-cluster Tool With Process Modue Failure

Multi-cluster tools are widely adopted as wafer fabrication equipment in semiconductor manufacturing.Due to a great number of multi-cluster tools running at a wafer fab,process modules(PMs)are prone to failure from time to time.Such failure could decrease productivity and scrap wafers if it were not handled properly.Thus,to increase the throughput of a wafer fab,it is economically significant to address this issue.In a multi-cluster tool system without any parallel modules,when one process module at a step fails,the entire process of any raw or work-in-process wafers cannot be completed.The system must be forced into a close-down process to be empty.However,some wafers cannot follow the predefined processing route,which may cause a deadlock due to the limited capacity of a buffer module(BM)and wafer residency time constraints.Consequently,the order of the robot activities cannot be determined simply by a traditional backward strategy.Owing to the random location of a failure module,it is rather challenging to find an effective and uniform schedule for different failure cases.However,the existing methods cannot solve such a difficult problem.Up to now,there is no report on how to schedule a multi-cluster tool properly after a PM fails so that it can be closed as soon as possible.Therefore,this research work focuses on the transient process for a multi-cluster tool without any parallel PMs after a PM failure subject to wafer residency time constraints.This work first analyzes different operation routes of the wafers under various failure situations.The robot task sequences and the synchronization conditions are analyzed for robot activities avoiding deadlock in the shared buffer modules between the cluster tools.Then an algorithm is presented to obtain the wafers' distribution information of the moment when the transient close-down process begins.Based on the output parameters of this algorithm and the analysis of robot activities,this work proposes another two algorithms to decide the robot task sequences for the two different kinds of the failure situations,respectively.Although in one of the failure situations,the order of some robot tasks is still undetermined,the start time of these tasks can be decided by setting the time interval between them.Therefore,a pure integer linear program model and a pure integer non-linear program model are built for such two different kinds of the failure situations to determine the start time of the tasks and minimize the makespan,so that a feasible and optimal schedule of the special close-down process with PM failure is found.With the multi-level nested lower subscript,for the first time,this proposed method can be the induction of the solutions for all different failure situations in the process-dominant2-cluster tool.Three examples from different failure situations are given to illustrate the application of this method.