(65a) Industrial Application of Multivariate Modeling Methods for Process Forecasting and Model Predictive Control

Multivariate (MV) methods are commonly used for characterization and statistical process control (SPC) monitoring of batch processes. Extensions of these modeling methods can also be used for predictive purposes to estimate future trajectories of important process parameters or final conditions of batch type processes. This talk provides a brief overview of the multivariate forecasting methods, referred to as imputation in literature, as well as the introduction of a new regression based method. Application of these imputation methods within an optimization framework is provided to demonstrate their use for process optimization and model predictive control (MPC). This is particularly useful in batch processes where traditional MPC technologies are not suitable.