Product transfer between plants using historical process data

Abstract

( ) Based on the concepts laid out in an earlier article Jaeckle and MacGregor, 1998 , this

Advanced Process Decision Making Using Multivariate Latent Variable Methods

“…The idea of latent variable model inversion, first introduced by Jaeckle and MacGregor [41][42][43], has been recently generalized by Tomba et al [46]. The following sections provide an overview of the theoretical aspects beyond this modeling technique.…”

“…However, if this solution is moved along the directions of the latent space that do not affect the data in Y, the same set of desired response variables y DES is obtained through the model. The space identified by these directions is called null space, and represents the locus of the X projections not affecting the quality space Y [41,43]. In the following subsection, details are given on the evaluation of the null space.…”

“…(16) can be applied [41][42][43]. If the values of some of the elements in y DES are not assigned and/or in the presence of a null space, the inversion problem can be defined as follows (SCENARIO 2):…”

“…The effectiveness of LVMs in such activities as process understanding, process monitoring and troubleshooting [16], process control [37][38][39][40], process design [41][42][43], product design [44], and optimization [45] has been proved in several industrial sectors.…”

Editorial Professor John F. MacGregor: a pioneer of multivariate statistical process control and recipient of the fourth Herman Wold medal

“…Missing dataÐmajor analysis of missing data methods in PCA/PLS [23] . Novel uses of databases and multivariate methods & Multivariate specs (Shewell award for best paper at Fall Technical Conference) [24] & Product design/transfer [25,26] . varying rapidly moving images) for on-line monitoring and feedback control.…”