Control Policies for Scheduling of Semiconductor Manufacturing Plants

Two important goals in scheduling semiconductor manufacturing plants are to minimize the mean and variance of the manufacturing cycle time. The manufacturing cycle time is the time taken by a job to traverse through the manufacturing system and become a finished product. In addition to the obvious economic and competetive adavantages, reducing the cycle time is also beneficial in minimizing the time that a wafer is exposed to contamination during the manufacturing process, thus providing higher yield. On the other hand, a small variance of the cycle time allows tighter production planning, and more consistent on-time deliveries of the end product. To meet these two goals, good input release policies and job scheduling policies have to be determined. The former determines job release control at the system input, while the latter allows proper job sequencing at each processing station. In a recent paper, Wein [3] has reported on the comparative performance of a number of release and scheduling policies on a simulated research and development fabrication line. In this paper, besides the policies studied by Wein, we propose two new scheduling policies, the Least Slack (LS) and the Least Optimized Slack (LOS) policies, and report on their performance in a discrete event simulation. Our preliminary results appear to demonstrate that the LOS scheduling policy can reduce both the mean and standard deviation of cycle times by perhaps significant amounts, especially under heavy traffic.