Optimal experimental design for process optimization with stochastic binary outcomes

ESTEFANÍA COLOMBO; MARTIN F. LUNA; ERNESTO C. MARTINEZ

Rosario

Simposio; Argentine Symposium on Industrial Informatics SII 2015 44th Argentine Conference on Informatics (44 JAIIO); 2015

Effective control of the end-use properties in order to guaranteeproduct quality is of paramount importance. This is especially valuable forproducts such as medicament, polymers and nanomaterials. There are plenty ofinnovative processes whose principles of operation are unknown and for whichis not possible to develop reliable models due to time and cost. Despite this, it isimportant to know the optimum operating zone where the process should workto ensure that the end product meets the required properties. To address this issue, a run-to-run optimization approach is proposed to find the reduced regionof operation which guarantees obtaining end-use properties with high probability when a first-principles model for the batch process is not available a priori.In order to evaluate the effectiveness of the methodology to find optimal policies for runs involving stochastic binary outcomes, the well-known example ofthe emulsion polymerization of styrene has been addressed. Results obtaineddemonstrated that the proposed method is a powerful tool for optimal design ofexperiments aiming to guarantee end-use product properties.