ABSTRACT
State estimation techniques have been well developed for dynamic processes described by ordinary differential equations. Among them the Kalman filter (KF) is an optimal estimator for linear dynamical systems in the presence of state and measurement uncertainties [1, 2]. Many techniques have been reported for inferential estimation of compositions in continuous distillation columns [36]. State estimator design for batch distillation has to deal with the time varying nature of the batch column. Quintero-Marmolet al. [7] applied an extended Luenbergerobserver to predict compositions in multicomponent batch distillation from temperature measurements. In order to compensate for limited state variable measurements, the extended Kalman filter (EKF) has been extensively employed to estimate the unmeasured states and unknown parameters from measured states. Composition profiles and operating conditions may change over a wide range of values during the entire operation and the state estimators must be designed to deal with the time-varying nature of the batch columns [8, 9].
