SPC for a data-rich environment using profile techniques

In modern manufacturing processes, huge sample size scenarios are becoming popular. Serious failure patterns that posses some local features, like partial mean or variance shifts, raise unique challenges for statistical process control (SPC). Conventional SPC methods monitoring the simple summary statistics are not an efficient way under such circumstances. Based on the study of a machine vision system example, we suggest characterizing each sample by a quantile-quantile (Q-Q) plot, and monitoring the linear profile generated from it. Profile monitoring techniques are implemented and studied. Simulation results show this an effective approach to handle this problem.