Algorithmic Statistical Process Control: The Concept, An Elaboration

Abstract

SPC has traditionally been applied to processes in which successive observations are independently distributed, for the purpose of detecting “assignable causes” as a basis for making fundamental process improvement. Stochastic control, on the other hand, addresses processes in which observations are dynamically related over time. Its intent is to run the existing process well, as opposed to improving it, per se. SPC has been generally applied in monitoring product quality, whereas automated controls are most often applied to control process “direction.” Modern industrial systems often exhibit dynamic behavior at even the product quality level, suggesting the need to expand application of control theoretical principles into this realm, as well. Competitive realities dictate, however, that fundamental process improvement (and not merely process optimization) must remain a dominant consideration. This paper elaborates an understanding of how SPC and feedback control can be united into a system which exploits the strengths of both. Building upon past work by MacGregor, Box, Astrom and others, the paper covers the concept and, technical issues that arise.

The author would like to thank his colleagues in this effort: Fred Faltin, Scott Vander Wiel, Necip Doganaksoy, and Gerry Hahn.