A just-in-time-learning based two-dimensional control strategy for nonlinear batch processes

Abstract

In this paper, we study the problem of two-dimensional (2D) integrated model predictive iterative learning control for nonlinear batch processes. A nonlinear process is presented by using local models together with a just-in-time-learning (JITL) method. In order to deal with the problem of massive computation for identifying JITL models, a three-layer searching strategy and an updating mechanism for databases are proposed. First, a dynamic model whose parameters vary with time and batch is established for batch processes. Then, a JITL-based 2D control strategy is devised to realize comprehensive control by combining iterative learning control in batch-axis with model predictive control in time-axis. As a result, not only the closed-loop control performance can be improved, but also the optimization procedure can be simplified. Finally, performance analysis verifies the effectiveness of the proposed methods.