Iterative learning control of a reactive batch distillation process

Quadratic criterion-based iterative learning control (Q-ILC) was applied to a numerical reactive batch distillation process, in which methacrylic anhydride (MAN) is produced through the reaction of methacrylic acid with acetic anhydride. The role of distillation is to shift the equilibrium conversion toward the direction of the product by removing acetic acid (AcOH), a by-produt of the reaction. Two temperatures at both ends of the column were controlled by individual control loops. A nonlinear PID control manipulating the reflux ratio was employed to regulate the top temperature at the boiling point of AcOH. A constrained Q-ILC was used for the tracking of the reactor temperature. A time-varying reference trajectory for the reactor temperature that maximizes the product yield was obtained through repeated simulations. The Q-ILC achieves satisfactory tracking in several batch runs with mild control movements, while the PID control in compative studies exhibited unacceptable performance.