Probability-Based Design of Experiments for Batch Process Optimization with End-Point Specifications

Abstract

Consistently complying with end-point specifications, mainly end-use product properties, is a key issue to thecompetitiveness of batch processes. To maximize in some way the probability of observing successful runs, a series ofexperiments is specifically designed to pinpoint the smallest operating region that guarantees that end-point conditions can meettheir desired targets. On the basis of data-driven modeling of the underlying binomial probability of success, the proposedmethodology seeks to trade off improving parameter precision with experimenting in a reduced region where there is a highprobability of satisfying end-point specifications. Two case studies are used to demonstrate the efficacy of the probability-basedoptimal design of experiments to find optimal policies for runs involving stochastic binary outcomes. Run-to-run improvement ofthe success rate for the operating policy in the acetoacetylation of pyrrole with diketene is first discussed. Results obtained foremulsion polymerization of styrene are also presented to illustrate how end-use properties, such as the tensile strength and meltindex, can be maintained in their desired target regions by the proper choice of the operating policy.