Adaptive control of continuous polymerization reactor

This study investigates the control of free-radical polymerization reaction of methyl-methacrylate initiated by azo-bis-isobutyronitrile in toluene solvent under isothermal condition. The dynamics of the polymerization reaction in the reactor is represented by a phenomenological mathematical model. The order of the model is reduced using the Hankel singular value decomposition (HSVD) to lessen the computational burden. However, process models are time-varying and exhibit different behaviors at different operating conditions. As a result, a recursive least squares with exponential forgetting factor (RLS-EFF) algorithm is employed to identify the parameters of the reduced-order model. Then, an indirect adaptive minimum degree pole placement control (AMDPP) which provides the desired closed-loop poles is implemented for the identified model. In addition, a continuous-time model reference adaptive control (MRAC) is developed for the reactor. Nevertheless, cost-effectiveness and constraint imposition cannot be achieved with both AMDPP and MRAC. Consequently, an adaptive model predictive control (AMPC) is designed for the identified model to overcome these limitations. Simulation results demonstrate the superiority of the AMPC.