Dispatching rules using flow time predictions for semiconductor wafer fabrications

Abstract

Reducing flow times is a major goal for wafer fabrication facilities. It could take more than a month to finish a wafer fabrication process. Reducing such a long flow time implies a quicker customer response, better marketing strategy, higher profit, lower factory WIP level, and, sometimes, higher yield rates. This study examines two dispatching rules—the modified least slack rule and the shortest remaining flow time rule. Both of these rules require future flow time predictions. We experiment with two flow time prediction methods—an exponential smoothing method and an empirical queueing approach. These new rules are compared with some other commonly used rules by simulation experiments. For the performance measure of mean flow time, the shortest remaining processing time rule performs best, whereas for the performance measure of flow time standard deviation, the modified least slack rule with the flow time prediction by the empirical queueing approach performs best. © 2002 Taylor & Francis Group, LLC.