Iterative learning control of a crystallisation process using batch wise updated linearised models

An iterative learning control strategy with batch wise updated linearised models identified using principal component regression (PCR) is proposed in this paper for the supersaturation control of a batch crystallization process. Taking the immediate previous batch as the reference batch, the linearised model relates the deviations in the control profiles with the deviations in the quality variable trajectories between the current and the reference batches. The linearised model is used in calculating the control policy updating for the current batch. Simulation results show that the proposed method can overcome the effect of disturbance and improve the process operation from batch to batch.