(182g) Process Aware Data Driven Modeling and Model Predictive Control of Monoclonal Antibody Bioreactor

This manuscript addresses the problem of process control on a bio-reactor to maximize the production of a desired output while respecting the bio-process related constraints. A data driven model is first built and used to formulate a model predictive controller (MPC) with the results illustrated by implementation on a detailed monoclonal antibody production model (the test bed) created by Sartorius Inc. A recently developed data driven modeling approach using an adaptation of subspace identification techniques is utilized that enables incorporation of known physical relationships in the data driven model development (constrained subspace model identification) making the data-driven model process aware demonstrating further improved performance. The resultant controller implementation is compared to the existing PI controller strategy used in the monoclonal antibody production showing vastly increased production. Simulation results also demonstrate the superiority of the process aware or constrained subspace model predictive controller compared to traditional subspace model predictive controller. Finally, the robustness of the controller design is illustrated via implementation of a model developed using data from a test bed with a different set of parameters, thus showing the ability of the controller design to maintain good performance in the event of changes such as changed feed characteristics or a different cell line.