Department of Industrial Engineering & Decision Analytics [Joint IEDA/ISOM] seminar - Data-Driven Operations Management for Semiconductor Supply Chains

     Semiconductor manufacturing combines capital-intensive equipment, complex production flows, and long lead times, yet operational decisions often must be made with limited data. This seminar examines how data-driven operations management can support such decisions and enable adaptive manufacturing models.

     The main part of the talk considers the qualification of newly manufactured lithography systems. For each system, the duration of an initial monitoring phase must be selected before it is known how long the system will take to meet the required qualification criteria. A short monitoring phase can trigger a lengthy and labor-intensive follow-up procedure, whereas a long phase can unnecessarily delay systems that meet the criteria quickly. Because the resulting observations are censored and the underlying distribution is initially unknown, we formulate the problem as a sequential decision process that combines Bayesian learning and optimization.

     The talk concludes with a brief research outlook on the future of modeling semiconductor manufacturing systems, beginning with data-driven aggregate models as lightweight alternatives to manually constructed, detailed simulation models and extending toward self-scoping digital twins that can automatically learn both system dynamics and the information needed to represent them.