Application of Derivative - Free Estimator for Semi Batch Autocatalytic Esterification Reactor: Comparison Study of Unscented Kalman Filter, Divided Difference Kalman Filter and Cubature Kalman Filter

• Published in: Computer Aided Chemical Engineering Vol. 37; pp. 329 - 334
• Main Authors: Rohman, F.S., Abdul Sata, S., Aziz, N.
• Format: Book Chapter
• Language: English
• Published: 2015
• Subjects: auto catalytic esterification; CKF; DDKF; Nonlinear estimation; semi batch processes; sUKF; sUKF; DDKF; semi batch processes; auto catalytic esterification; Nonlinear estimation; CKF
• ISBN: 9780444634290; 0444634290
• ISSN: 1570-7946
• DOI: 10.1016/B978-0-444-63578-5.50050-5

The online or real time optimization system required an on-line monitoring for important state variables and parameters in order to provide feedback to the algorithm implemented. The unmeasured states and unknown parameters can be determined by estimator. The estimator used must be feasible and has several functionalities such as bias-free parameter and state estimates, high speed of convergence from initialization errors, perfect tracking and short computational time. To address those issues, the derivative free stochastic observer is applied in this work. Different derivative-free techniques implemented which are based on the Gaussian approximation are scaled Unscented Kalman Filter technique (sUKF), Divided Difference Kalman filter (DDKF) and Cubature Kalman filter (CKF). Here, an autocatalytic esterification of Propionic Anhydride with 2-Butanol is considered as a case study. The performance of estimator which is indicated by the accuracy, speed convergence and robustness, is represented by the root mean squared error (RMSE). While, the efficiency of the estimator’s computation are evaluated by CPU time consumed. The derivative free-estimators are then evaluated within two case studies which are normal condition and effect of uncertainty of initialization parameter. The results of overall state and parameter estimator study show that the CKF outperform the sUKF and DDKF for both cases. Moreover, computational load consumed for the CKF is feasible for online practice However, in term of CPU time DDKF results to the shortest among all techniques implemented.