Projection methods for process analysis and statistical process control: examples and experiences from industrial applications

Multivariate projection methods based on PLS and PCA can be utilized to analyze process and product performance in cases where a large number of variables are involved. Two examples from industrial processes are presented in detail: one with 440 process variables from a petrochemical company and one from the semiconductor industry with 80 product quality variables. References are given to applications of projection methods to a variety of processes from diverse industries such as petrochemical, polymer, steel, pulp and paper, mining, and experiences from such applications are discussed. The objective is to illustrate the kind of industrial problems that can be solved and the diversity of industries that can benefit by utilizing these methods.