Batch process control from practice to 2D model predictive control

Abstract

Owing to the natures of batch processes, such as high nonlinearity, time-varying, and limited batch time duration, their control remains as a challenge to modern industries. This paper takes a typical batch process, injection molding, as an example to present a set of control schemes for batch processes. Advanced control algorithms such as adaptive control and model predictive control have been adopted to deal with the inherent process nonlinear and time-varying characteristics. These control algorithms are all focused on single cycle control performance. A multi-cycle two-dimensional model predictive learning control has been developed for batch processes control to take advantages of batch process repeatability. In this presentation, besides showing the control results/methods, the authors wish to illustrate the development evolution with their understanding of the natures of batch processes in general, injection molding in particular.