Iterative learning strategy for a class of nonlinear controllers applied to constrained batch processes

In this paper, we apply an iterative learning strategy to improve the performance of a class of nonlinear controllers, when they are applied to constrained batch processes. The idea is to exploit the control error information from the previous batches so that the corrected control inputs will iteratively improve the control performance. In this iterative learning scheme, we provide the convergence proof of the feed-forward input correction strategy as the batch cycle progresses. Furthermore, we extend the proposed strategy for handling input constraints, which in some cases the constraints may result in an accumulated error during the iteration process. To deal with this problem, we propose a segmented reference trajectory, where the learning strategy is applied for each segment with the assumption that a smooth transition between segments is established. Throughout the paper, a batch reactor control problem is used to illustrate how the proposed methods work in practice.