A Comparison of Nonlinear State Estimators for Closed-loop Control of Batch Crystallizers

This study investigates the effectiveness of various nonlinear estimation techniques for output feedback model-based control of batch crystallization processes. Several nonlinear observers developed under deterministic and Bayesian estimation frameworks are applied for closed-loop control of a semi-industrial fed-batch crystallizer. The performance evaluation is done in terms of closed-loop behavior of the control strategy and its ability to cope with model imperfections and process uncertainties such as measurement errors and uncertain initial conditions. The simulation results suggest that the extended and the unscented Kalman filters perform best in terms of fulfilling the control objective. Adopting a time-varying process noise matrix, which is particularly suited for batch processes, further enhances the accuracy of state estimates at the expense of a slight increase in computational burden. The results also indicate that model imperfections and process uncertainties rather significantly deteriorate the closed-loop performance of the controller due to inaccurate state estimation.