Quality Quandaries: The Application of Principal Component Analysis for Process Monitoring

Abstract

When monitoring industrial processes, it is increasingly common that a large number of process parameters are observed simultaneously over time. In that case, the observations will most likely be correlated with each other (contemporaneously) and in time (temporarily). To gain understanding of the behavior of a system, especially when troubleshooting a process, graphical and statistical tools appropriate for such multivariate situations can be helpful. This column demonstrates a few techniques, mostly graphical, that are useful when dealing with such situations. We focus primarily on contemporaneous correlation. Indeed, one of our goals is to demonstrate principal component analysis (PCA), a method akin to a Pareto analysis. We assume no prior knowledge of PCA, and rather than taking a typical mathematical approach, we focus on the geometry of PCA to enhance intuition.