APPLICATION OF AN EXTENDED LUENBERGER OBSERVER TO THE CONTROL OF MULTICOMPONENT BATCH DISTILLATION

State estimators are used to predict the values of those variables that cannot be measured directly, using the state variables that are accessible for measurement. There is considerable interest in the application of observers, but very few applications to complex, nonlinear, high-order chemical engineering systems such as distillation columns have been reported. This paper explores the use of observers to predict compositions in multicomponent batch distillation through the use of temperature measurements. It is important in batch distillation to know the instantaneous composition of the distillate withdrawn in order to apply the optimum distillate policy, but usually perfect composition measurements are not available and we need to rely in some kind of inferential estimation of these compositions. Simulation results demonstrate the feasibility of applying extended Luenberger observers to multicomponent batch distillation. A design methodology is developed that addresses the important issues of the number and location of temperature sensors, the placement of the closed-loop poles of the observer, and the sensitivity of initial condition estimates.