I completed my Ph.D. in May 1996 at the Operations Research Center, an interdepartmental program at Massachusetts Institute of Technology. My research centred on the analysis and control of manufacturing systems. Here is an executive summary.
In my thesis, I developed and analyzed a particular class of control policies. These control policies are called hybrid or two-boundary policies, and combine the finite buffers (and thus low inventory) of kanban and similar schemes with the overall robustness and responsiveness of the basestock and hedging point policies. The hybrid policies are quite attractive for high volume production in environments with high inherent variability, such as semiconductor manufacturing. The thesis proposal is here. I presented the proposal for my committee on Oct 3, 1995, and right here are the overheads from that talk. This is a seminar I gave at the Laboratory for Manufacturing and Productivity on Nov 17, 1995. My thesis advisor was Dr. Stanley B. Gershwin.
These policies are implemented in the CIDM scheduling and simulation testbed at the MIT Microsystems Technology Laboratories. This is a factory floor dispatch program integrated with the CAFE computer integrated manufacturing system, but it can also be executed as a discrete event simulation to estimate the consequences of adopting particular policies in the factory. In this mode, it can read factory description datasets distributed by Sematech, a semiconductor industry consortium, which allows us to see how the policies perform in other factories than our own. Here is an introduction to the testbed, and here are the results from a study of various control approaches in semiconductor manufacturing.
We also explored the application of these policies in other manufacturing contexts, such as automobile manufacturing. Here is a paper that demonstrates how these policies outperform kanban control with respect to inventory levels, even if the production system belongs to Toyota. The study was done by a distributed simulation model.In the spring of 1994, I analyzed the distribution of lead times in semiconductor manufacturing (they can be closely approximated by a gamma distribution). The paper is here.
From Aug 12, 1996, I work in the Oslo office of McKinsey & Co, Inc.
Since I'm not around, please send any comments to
gershwin@mit.edu