Research On The Scheduling Of Batch Switching Of Single-Armed Cluster Tools With Wafer Residency Time Constraints

Semiconductor manufacturing industry is the foundation of modern information industry,and has important strategic significance for promoting national economic development and technological progress.As a key part in semiconductor manufacturing industry,wafer manufacturing is characterized by complex processing technology and strict environmental requirements.To meet these requirements,cluster tools are widely used in wafer manufacturing.With the improvement of customers' personalized requirements,wafer manufacturing tends to adopt multi-varieties and small-batch processing mode,and cluster tools need to switch from processing one batch of wafers to another frequently,which reduces the utilization of cluster tools and prolongs the processing period.For some wafer processing technology,there is a strict constraint called residency time constraint,which requires wafer stay in a processing module(PM)within a limited time.Therefore,in order to improve the working efficiency of cluster tools and ensure the quality of wafer manufacturing,the batch switching scheduling problem of single-armed cluster tools with wafer residency time constraints was addressed.Firstly,the switching process of two batches of wafers was analyzed,especially the sequence of robot activities.A Petri net was adopted to describe the batch switching process in detailed,and the control strategies were established to avoid deadlock for the Petri net model.The transitions and places in the Petri net model are associated with time,and the sufficient and necessary condition was obtained for the live system through the analysis of activities time.Secondly,during the evolution from a system marking to another,according to the transitions firing sequences in the Petri net model,the analytical expressions for robot tasks time and wafers sojourn time were obtained.The relationship of workloads among the steps of two batches of wafers were divided into two situations,namely relative balance and relative unbalance.For the first situation,the theorems for judging the system's schedulability and the corresponding algorithm for solving the robot waiting time were obtained,and it was proved that the algorithm can acquire a feasible and optimal scheduling scheme under the condition of the theorems.For the second one,we established the linear programming models(LPM)for system operations.LPM can also judge the feasibility of the scheduling,and obtain the optimal scheduling if it exists.In order to improve the solving efficiency of LPM,the corresponding LINGO models were established.We analyzed several examples to verify the validity and correctness of the theorems and LPM.Finally,due to the large running cost of cluster tools,we used the eM-Plant establish simulation model for single-armed cluster tools,and controled the flows of wafers and the movements of robot in the simulation operations through the embedded programming language.In the simulation operations,the time when wafers enter and leave PMs were recorded in a table,so as to judge whether it violates the wafer residency time constraints and verify the feasibility of scheduling scheme.The validity of the scheduling method and the simulation model were verified by several examples.The above results showed that,for the single-armed cluster tools with the given robot scheduling strategy,the methods proposed in this paper can judge the feasibility of batch switching scheduling,and obtain the optimal switching makespan if the system is schedulable.