Batch-to-batch optimal control of nonlinear batch processes based on incrementally updated models

An iterative learning control (ILC) strategy for the tracking control of product quality in agile batch manufacturing processes is proposed. The ILC strategy is based on a linearised model of the controlled process that is initially identified from a minimum amount of historical process operation data and is updated after each batch run. The amount of historical process operation data in agile responsive batch manufacturing processes is usually limited and might not be sufficient for model identification if the model contains a large number of model parameters. To address this issue, the number of control intervals in a batch is set to be increased from run to run. Initially, the number of control intervals in a batch is set to a small value so that the linearised model contains a small number of model parameters, which can be estimated using historical process operation data. Control actions for a new batch run are calculated using ILC. Then the model is updated by using the augmented data set incorporating data from the latest batch run while the number of control intervals in a batch is increased by one until reaching its maximum value. The procedure is repeated from batch to batch. The proposed methodology is illustrated on a simulated batch reactor. The simulation results demonstrate that the proposed technique requires minimum historical process operation data and can improve control performance from batch to batch.