(301ab) Latent Variable Methods for Process Design: Dealing with Specification and Operational Constraints

Previous work has been presented and published in the used ot latent variable models for process operational design and optimization [1,2,3]. Such approaches address the problem in two steps: 1) Estimate a score vector given a specification for the desired final properties of the material, which includes at least one equality specification; and 2) Reconstruct the corresponding operating conditions with the possibility of an optimal criterion.

This work presents a combined formulation that addresses the problem in one optimization problem. This approach is particularly usefull when the specifications of the new material consists of only inequality constraints and an optimization criteria; and still operational constratints and operational optimal criteria have to be kept.

This new formulation is presented and illustrated with an application to a pharmaceutical process.

[1] C.M. Jaeckle and J.F. MacGregor, Product Design Through Multivariate Statistical Analysis of Process Data, AICHE J., 44 (1998) 1105 1118.

[2] C.M. Jaeckle and J.F. MacGregor, Industrial applications of product design through the inversion of latent variable models, Chemom. Intell. Lab. Syst., 50 (2000) 199 210.

[3] Garcia, S. Neogi, D., Metha, S., MacGregor JF., Kourti, T., "Optimization of Batch Processes using Data Driven Latent Variable Models". AICHE Annual Meeting, Cincinnati, OH. 2005