Industrial applications of product design through the inversion of latent variable models

Abstract

A methodology, presented by Jaeckle and MacGregor [C.M. Jaeckle and J.F. MacGregor, Product Design Through Multivariate Statistical Analysis of Process Data, AIChE J. 44 (5) (1998) 1105–1118.], for finding a window of process operating conditions within which a product with specified quality characteristics can be produced is applied to two industrial polymerization processes. The approach uses historical data available on the process operating conditions and on the corresponding product quality for a range of existing product grades. Latent variable models — built using the existing data — are inverted to obtain a window of operating conditions, which are not only capable of yielding the desired product but are also consistent with past operating procedures and constraints. A semi-batch emulsion polymerization process for manufacturing various grades of a copolymer latex, and a batch solution polymerization process for manufacturing a range of polymer resins are considered.