We study a framework of dynamic sampling in a wafer
production line in order to control the operational costs which include costs
of inspections and revenue loss due to the production of “out of control” (OOC)
batches. We incorporate elements from statistical process control (SPC) and the
machine learning theory to develop a dynamic sampling method which improves the
total operation costs in comparison with static sampling commonly implemented in
the industry.
The research aims at (1) online estimating the current production
line state to foresee the quality of wafers, and (2) adapting the sampling plan
so that the long run total cost is minimized.
