Sequential Screening in Semiconductor Manufacturing, II: Exploiting Lot-to-Lot Variability
研究了半导体制造中在晶圆层面进行序贯筛选的问题，采用经验贝叶斯方法建模，并用工业数据测试最优策略，发现芯片层面筛选比晶圆层面更有利可图。
This paper addresses the same quality management problem as Longtin, Wein and Welsch (Longtin, M., L. M. Wein, R. E. Welsch. 1996. Sequential screening in semiconductor manufacturing, I: Exploiting spatial dependence. Opns. Res. 44 173–195.), except that here screening is performed at the wafer level, rather than at the chip level. An empirical Bayes approach is employed: The number of bad chips on a wafer is assumed to be a gamma random variable, where the scale parameter is unknown and varies from lot to lot according to another gamma distribution. We fit the yield model to industrial data and test the optimal policy on these data. The numerical results suggest that screening at the chip level, as in Longtin, Wein and Welsch, is significantly more profitable than screening at the wafer level.