Explicit pH and Temperature Control of Complex Pharmaceutical Bioprocesses

With the recent impetus on advanced biotechnical processes in the pharmaceutical industry, it is critical to explore methods that can improve the efficiency and performance of the real-time operation for these sophisticated processes. The advanced control of these systems represents a challenging task due to the highly complex dynamics. The majority of in-use control techniques for pharmaceutical production use single-input single-output (SISO) and error reactive methods which can lead to suboptimal operations. The use of Model Predictive Control (MPC) provides an approach for optimally controlling systems. However, in pharmaceutical bioprocesses, the systems may be too complex to model accurately, or intricate models may lead to computational inefficiencies in the online implementation of MPC. Alternatively, multi-parametric model predictive control (mp-MPC) can solve the control problem offline to obtain optimal control laws as linear piecewise affine functions of system parameters (e.g. process states, setpoints, etc.). Understanding the effects of pH and temperature in pharmaceutical processes is critical in safely and reliably controlling these systems to produce high-grade pharmaceuticals. To facilitate the application of model-based controls, system models need to be validated in their ability to capture the effects of manipulating system variables while predicting accurate operational trajectories.

In this work a model for the batch lactic acid fermentation production is adopted and validated from literature for its robust and generalizable formulation featuring dependencies on temperature, pH, and substrate concentration. In validating this model, utilizing parameter values taken from literature, accuracy errors are present. This is rectified by employing non-linear parameter regression techniques to minimize modeling errors. To this end, we present a new set of regressed modeling parameters that more accurately fit the experimental data. The high-fidelity model with optimized parameters is linearized to enable to use of mp-MPC, thus avoiding computational complexity involving the elaborate system model. An explicit MPC problem is formulated utilizing the linearized models to maintain the pH and temperature of the biotechnical batch reactor at desired setpoints by manipulating the addition of acids or bases and the heating/cooling utility flowrates to the reactor. Control studies are presented and validated against the true nonlinear system to demonstrate the efficacy of advanced control techniques on these pharmaceutical bioprocesses. The improved high-fidelity model derived in this work and the application of mp-MPC stand as foundational steps for optimizing the operations of future large scale industrial implementations of pharmaceutical bioprocesses.