Latent Variable Modeling of Batch Processes for Trajectory Tracking Control

Latent Variable Modeling (LVM) of batch processes is explored from the view point of its application to trajectory tracking model predictive controller design. The ability of the models to capture nonlinearity and time-varying properties of batch processes and to provide a well-behaved description of the process are important characteristics to be considered. Furthermore, the importance of requiring as few batches as possible in the modeling step is considered in the discussion of different models. Two previously proposed approaches for batch process modeling (Golshan et al., 2009b) are investigated from the above points of view and benefits of them as well as their drawbacks are specified. Then, a new approach is proposed to overcome the major shortcoming of each previous approach while capturing their major benefits. The impact of the different latent variable modeling approaches on MPC for trajectory tracking is illustrated using a simulation of a Nylon polymerization process.