Trajectory tracking in batch processes using latent variable models

The set-point tracking of certain process variable trajectories is often needed for the lower level control in batch processes so as to achieve desirable final product quality for the higher level control. In order to realize trajectory tracking successfully, process models should be known in advance. In fact, process models play an essential role in trajectory tracking. Due to the difficulty for developing first-principle models for batch processes, empirical models such as multi-way principal component analysis (PCA) and multi-way partial least squares (PLS) are increasingly used in practice. Trajectory tracking using multi-way PCA models has been proposed in the literature, where the underlying optimizations are performed in the latent variable space. This paper explores the corresponding application of multi-way PLS models for the task of trajectory tracking and compares it with the existing multi-way PCA model-based methods through benchmark case studies.