Analysis of Control Mechanisms in a Re - entrant Manufacturing System

Modern semiconductor wafer fabrication facilities need to run at high levels of utilization in order to absorb the huge investments made in facilities and equipment. Simultaneously these facilities desire to minimize the cycle times of products in the factory for a variety of reasons including improved customer response, increased yield, and support of product development efforts. However, probabilistic models argue against the full realization of both of these goals at the same time. Digital Equipment's Fab 6 facility had implemented a series of factory rules aimed at achieving these dual purposes. However this facility faced some unique challenges based on dual the mission of the fab to make multiple products and to cooperate with product development direction. Available publications and theory were considered inadequate to evaluate the effectiveness of the policies in use. This study was conducted to assess the contribution to cycle time and cycle time variability of three operational policies in use in Fab 6. The three policies include start policy, dispatch policy and kanban-type control system. Through a design of experiments style investigation conducted on a simulation model of the factory, it will be shown that the fab has opportunities to improve its cycle time performance by adopting different policies. Moreover, it will be demonstrated that certain combinations of policies do not work well together and should be avoided. Thesis Supervisors: Assistant Professor David Cochran, Mechanical Engineering Professor Larry Wein, Management Acknowledgments I gratefully acknowledge the support and resources made available through the Leaders for Manufacturing Program, a partnership between the Massachusetts Institute of Technology and major United States manufacturing companies. This research project would not have been successful without the resources and support of the Fab 6 organization of Digital Equipment Corporation. In particular, I would like to thank Ellen Mager and Karl Koch for their direction and input. Professors Dave Cochran and Larry Wein provided important research insights during the internship period and valuable writing direction during the thesis phase. This paper would not exist without their assistance. My MIT experience could not have happened without the prayerful and loving support of my parents Duane and Marilyn Gifford and my new parents Hae-Yong and Song-Cha Park. Nor would I have survived the hot summers and cold winters without the unwavering commitment of my dear wife, Kyoung. Table of