Model Identification and Model Predictive Control of Biopharmaceutical and Biomedical Systems

This paper demonstrates the use of model predictive control (MPC) formulations for uncertain time-varying biopharmaceutical and biomedical systems implemented using measured data without prior knowledge of an accurate model. Furthermore, we demonstrate how prior knowledge can be incorporated in the identification of the model either through constraints or as regularization of the system identification procedure. We demonstrate the use of system identification to develop a model of the fed-batch Chinese hamster ovary mammalian cell bioreactor process and the implementation of model-based control to maximize therapeutic product yields. We also use a time-varying nonlinear biomedical system to demonstrate improvements due to incorporating prior information in the learning of the models and reidentification of the models when prediction accuracy deteriorates. We propose a new partial least squares algorithm that incorporates regularization from prior knowledge and can handle missing data in the independent covariates. Simulation case studies involving a biopharmaceutical production process and automated drug delivery demonstrate the capabilities of the proposed techniques.