Muti-Objective Nonlinear Model Predictive Control of Semibatch Polymerization Reactors Pilar Morales, Antonio Flores-Tlacuahuac* 1. Introduction With the ever increasing competition in the chemical and process industry the margin for revenues has been shrinking. One of the ways of raising process profits consists in the search of novel operating procedures leading to the reduction of raw material and utilities consumption. Of course, other approaches such as process retrofit could also be worth to explore. Because of the complexity of polymerization systems it turns out that relying only on physical experience and heuristics could make hard to come out with improved operating policies especially when several conflicting objectives ought to be considered. On the other hand, systematic techniques for the computation of better operating policies do not depend on intuition and they may lead to discover unusual operation procedures that would be hard to find out by using only process expertise. The deployment of advanced optimization techniques constitutes one of the best available systematic tools for increasing process profit. Dynamic optimization is one of the systematic techni- ques that could be deployed for stretching revenues. In fact, dynamic optimization techniques have been widely used for approaching the optimal control of polymerization reactions [1,2] especially under grade transition scenar- ios. [3,4] Polymerization reactors have also been used for assessing the impact of simultaneous scheduling and control techniques [5] with the long range objective of incorporating planning activities as well. In fact, this is an open research problem [6,7] worth to explore. However, until now, most of the addressed optimal control problems in polymerization reaction engineering have dealt with single objective functions. Of course, this is not necessarily wrong providing that only one objective should be optimized. The problem comes out when several conflicting and contra- dictory optimization objectives must be considered simul- taneously. Under this scenario the common way of addressing multi-objective optimization problems has Full Paper P. Morales, A. Flores-Tlacuahuac Departamento de Ingenierı ´a y Ciencias Quı ´micas, Universidad Iberoamericana, Prolongacio ´n Paseo de la Reforma 880, Me ´xico D.F., 01210, Me ´xico E-mail: antonio.flores@uia.mx A novel control system for semibatch polymerization reactors based on nonlinear MPC is proposed that is aimed to deal with more than one control objective. Commonly, multi- objective control problems are reduced to single objective problems, but better control can be achieved by solving the problem as a true multi- objective optimization problem because the inter- actions among the control goals are taken into account. Moreover, the selection of subjective weighting func- tions is avoided. The procedure is based on computing a trade-off solution among the control objectives that features the minimum distance from a given point on the Pareto front to the Utopia region. To illustrate the application of the multi-objective nonlinear MPC strategy two complex reaction systems are deployed. 252 Macromol. React. Eng. 2012, 6, 252–264 ß 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com DOI: 10.1002/mren.201100074