Abstract
Because of residency time constraints and activity time variation of cluster tools, it is very difficult to operate such integrated semiconductor manufacturing equipment. This paper addresses their real-time operational issues. To characterize their schedulability and achieve the minimum cycle time at their steady-state operation, Petri net (PN) models are developed to describe them, which are very compact, and independent of wafer flow pattern. It is due to the proposed models that scheduling cluster tools is converted into determining robot wait times. A two-level operational architecture is proposed to include an offline periodic schedule and real-time controller. The former determines when a wafer should be placed into a process module for processing, while the latter regulates robot wait times online in order to reduce the effect of activity time variation on wafer sojourn times in process modules. Therefore, the system can adapt to activity time variation. A cluster tool derived as a not-always-schedulable system by the existing methods is shown to be always-schedulable by using the proposed novel method.
Note to Practitioners-This paper addresses the real-time operational problem of dual-arm cluster tools with residency time constraints and activity time variation. A formal PN model is developed to describe the system. Then, based on the model, a two-level architecture is presented to deal with the problem. At the upper level, we have a periodic schedule generated offline such that, under deterministic task times, feasibility is guaranteed and cycle time is minimized. The real-time controller at the lower level is used to offset the task time variation. This paper presents such a real-time controller that makes the system adapt to task time variation. Hence, it is applicable to semiconductor manufacturing applications.