A dynamic process model of an industrial crystallizer train for para-xylene production, which consists of five scraped surface crystallizers, two hydrocyclone separators, and two centrifugal separators, is developed for control system design, and the model parameters are estimated from real plant data. Based on the identified model, steady state optimal operation policy, which considers maximization of paraxylene recovery and load distribution among the crystallizers, are derived. Multi-loop and multivariable controllers are designed to realize the operation policy and their performances are compared through simulations to show a clear advantage of the multivariable control system. (C) 2010 Elsevier Ltd. All rights reserved.