Optimal Control for Multi-variable Problems

Abstract

Chemex Laboratories measures concentrations of various minerals in land samples provided by their clients. The problem presented by Chemex is that they would like to be able to detect when their measurement apparatus needs re-calibration due to a shift in process parameters, so as to minimize rework.

Three very contrasting approaches to the problem were taken during the workshop. They are: considering the process as a Markov Chain; use of Bayesian Belief Networks; and Multivariate control chart implementation. Preliminary data analysis for data provided by Chemex was also performed in order to initially assess some of the models considered.