Using on-line process data to improve quality: Challenges for statisticians

In the process industries measurements on a large number of process variables are routinely collected at regular intervals by on-line computers. This paper makes a case for incorporating these process variables into Statistical Process Control (SPC) schemes. Multivariate statistical methods such as Principal Component Analysis (PCA) and Partial Least Squares (PLS) can be used to project these data down into low dimensional spaces where analysis, monitoring and diagnosis are easily performed. Strong justifications for taking this approach are presented and examples are given, The statistical process control community has been slow in adapting to the data explosion brought about by the computer era, It has continued to stick with traditional control charts on the quality variables and ignored this rich source of additional information on the process. This paper explores some of the reasons for this and argues that the SPC community must adapt rapidly or lose control of the field to scientists and engineers. The paper also tries to induce statisticians into looking more seriously at the many unsolved problems in this area of reduced rank multivariate statistics.