Development of A Capacity-Filtering Method for Medium-term Scheduling of Semiconductor Fab

Seong KyungChan
The operational management of an electronics Fab such as semiconductor Fab, LCD Fab or PCB Fab requires long-term planning, medium-term scheduling and short-term control. Among the three areas, the weakest link is the medium-term scheduling or finite-capacity planning area which is responsible for generating mid-term schedule or finite-capacity plans. In such an electronic Fab, medium-term schedules such as release plan production plans are critical in achieving the goal of full capacity and on-time production. However, existing finite capacity planning methods, including time-bucking and time-shifted POR approachs which are most widely employed, are not able to accommodate the capacity fluctuations that can arise in the electronics Fab, especially the semiconductor Fab. In most semiconductor Fabs, there are workstations that are near-bottleneck, any of which may become a bottleneck at a given points in time due to changing product mixes and unexpected down time. Managing the interactions between these near-bottleneck stations to maintain high throughput and on-time delivery performance is a very challenging task. This is why existing finite capacity methods have difficulty planning and scheduling in semiconductor Fab. As other theoretical approaches also have shortcomings, it is anticipated at present that the capacity-filtering method proposed in this thesis is a viable solution.The comparative performance analysis of a real semiconductor Fab demonstrates that the proposed capacity-filtering method is superior to the existing ‘time-bucketing’ and ‘time-shifted POR’ methods. While a major drawback of this approach is that the fixed-size time-buckets do not adequately reflect the actual line capacity, the proposed method is able to be flexible enough to permit quick replanning. In addition they are able to produce high-quality production plans. This is one of reason why the proposed method is better than the others. What in more, this thesis represents a possibility that it can be used in ‘real-life’ release planning and Fab-out (production) planning.