Multivariate Statistical Process Control And Optimization (pp. 209-227)
Authors: Pomerantsev, A. L. and Rodionova, O. Ye. (Institute of Chemical Physics, Moscow, Russia)
Abstract: The multivariate statistical process control (MSPC) is an approach that helps to run the real-world processes. The MSPC concept is to apply the historical process X-data for construction of linear model, which explains how the final results (i.e. quality, y) depend on X-variables. Apparently, studding this model, one can suggest the program of actions that could improve the performance in general. However, the most important thing in production is the local optimization, which means the immediate actions that are performed in the course of process (i.e., in-line) in order to improve the quality of the final product or/and to reduce the production costs.. The paper contains new ideas, which (we believe) are interesting in process control. The first one comes within the conventional MSPC approach. We suggest constructing a series of MSPC models that describe the process in order of its cycles, stage by stage. This Expanding MSPC (EMSPC) method can be used for prediction of future final product quality at the early process stages while production cycle is not over yet. This may be considered as a passive optimization as we do not suggest the correcting actions. The second idea implies the novel SIC method to find the best in-line corrections. This technique may be called the Multivariate Statistical Process Optimization (MSPO) and it also applies the idea of EMSPC. The MSPO approach may be considered as an active optimization as we always suggest the correcting actions in the course of the process.