(629a) Real Time Constrained Model Predictive Control for the Continuous Manufacturing of Pharmaceuticals

Manousiouthakis, V. - Presenter, University of California Los Angeles, Los Angeles
Mehta, N., University of California Los Angeles
Nocon, K., University of California Los Angeles
Sheikh, O., University of California Los Angeles
The cost of healthcare in the US is the fastest growing component of the inflation index. One method aimed at reducing these growing costs is in the development of processes for continuous pharmaceutical manufacturing. The continuous production of pharmaceuticals offers reduced costs, simplified and streamlined scale-up, and flexibility compared to batch process alternatives.

This research presents an approach to control system design for the continuous production of pharmaceuticals. First, a dynamic model of the continuous drug manufacturing process is developed that leads to a system of nonlinear, differential algebraic equations. This model is then linearized around typical process operating conditions. The linearized model is subsequently discretized and is used to develop a real time online model predictive control system that ensures the integrity of the pharmaceutical production process and helps satisfy the stringent FDA requirements typically required for safe drug production. Specifically, a constrained infinite time linear quadratic regulator control strategy is employed that can satisfy hard constraints in real time. For example, if there is variability in the quality of raw materials, the controller can ensure, in real time, that FDA requirements on drug purity are met.

The proposed approach is demonstrated with a case study based on a continuous Ibuprofen manufacturing process, as developed for steady-state conditions by Jolliffe & Gerogiorgis (2014), and references therein. In this work, a dynamic model of this process is developed, linearized, and controlled as described above.