On-Line Monitoring of a Copolymer Reactor: A Cascade Estimation Design

Abstract

Abstract In this work the problem of on-line monitoring of product quality and production rate in a copolymer reactor is addressed, using an estimation scheme with secondary measurements of density, refractive index, temperature, and volume. Three different estimator structures are studied: (a) the nominal detectability structure that underlines the extended Kalman filter and Luenberguer observers, (b) a passive estimation structure with estimation degrees equal to one, and (c) a hybric structure that combines the detectability and passive structures in low and high gain, respectively. The nominal detector maximizes the reconstruction rate, the passive estimator maximizes the robustness, and the cascade (hybrid) design achieves a suitable compromise between them. The approach is illustrated with a case study and simulations.