Toward a Real-time Scheduler and Controller for Semiconductor Fabrication Systems
Hyeung-Sik Min, Albert T. Jones, Yuehwern Yih
The complexity of a semiconductor fabrication system is high because of complicated product flows, uncertain operation times, variable yields, changing products, and evolving technologies. Effective scheduling, which attempts (1) to predict accurately the start and completion times, and (2) to regulate the work-in-process inventory levels, can be quite challenging in this environment. The challenge grows when there are multiple performance objectives that vary over time. In this paper, we describe a competitive, neural-network-based approach that runs in real-time and at predetermined, fixed, time intervals. At each such interval, the network uses the current performance objectives and the current system status to generate a new schedule. We also describe briefly the simulation approach we used to train the network. Finally, we discuss our current efforts to include robustness computations in the decision-making and to use this approach as part of a real-time controller for the system.
Proceedings of the International Conference on Product Research
, Jones, A.
and Yih, Y.
Toward a Real-time Scheduler and Controller for Semiconductor Fabrication Systems, Proceedings of the International Conference on Product Research, , USA, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=821979
(Accessed May 27, 2023)