A LV-MPC Strategy for Trajectory Tracking In Batch Processes

GODOY, J.L.; MARCHETTI, J.L.; VEGA, J.R.

Lamaca

Conferencia; 9th International Conference on Integrated Modeling and Analysis in Applied Control and Automation (IMAACA); 2016

A constrained latent variable model predictive control (LV-MPC) technique is proposed for trajectory tracking in batch processes. The controller allows the incorporation of constraints on the process variables and is designed on the basis of multi-way principal component analysis (MPCA) of a batch data array that is rearranged by means of a regularized batch-wise unfolding approach. The LV-MPC formulation includes a novel prediction stage and is offset-free. The controller parameters are calculated on the basis of the identified latent model. The main advantages of LV-MPC over other MPC techniques are: (i) a relatively small dataset is required (e.g., around 10-20 batch runs), (ii) nonlinear processes can efficiently be handled algebraically through MPCA models, and (iii) the tuning procedure is simple. The proposed constrained LV-MPC technique is numerically tested through a benchmark process that has been used in previous LV-MPC formulations.