Compared to traditional batch manufacturing routes, continuous biopharmaceutical manufacturing is a fast-approaching paradigm with the potential to save time, money, and lives. Achieving these gains, however, requires a suite of predictive, diagnostic, and corrective tools which are formalized in the Process Systems Engineering (PSE) field. Throughout my PhD training, I focused on understanding and applying these PSE tools to a variety of continuous upstream and downstream unit operations in biomanufacturing applications. My efforts can be summarized in three ways:
Answering questions of process model utility and process observability through model selection and state estimation-based soft-sensing. These tools require real-time data processing coupled with fast numerical simulations of dynamic models for intrinsically distributed operations, i.e., which are described by population balance models (PBMs), or for strongly timescale-separated operations, i.e., which are described by differential algebraic equations (DAEs). Projects included pH estimation in bioreactors and imaging-based growth kinetics estimation of morphologically-varied, crystallizing or precipitating pharmaceuticals;
Answering questions of process uncertainty and mismatch with model behavior through robust open-loop (i.e., uncontrolled process) analysis. These techniques involve using current and previous experimental studies to perform parameter estimation, uncertainty quantification, and identifiability studies on proposed models. Projects included viral vaccine production in two-stage stirred tank reactors and modeling the shallow temperature-cycled conversion of a monoclonal antibody from one solid form to another;
Answering questions of process controllability through closed-loop (i.e., controlled process) synthesis. Here I applied a variety of optimization-based controls methods (e.g., dynamic optimization, nonlinear model predictive control) for optimizing objectives relevant to biomanufacturing, e.g., production process costs and critical quality attributes (CQAs). Projects included oscillatory yeast cell manufacturing optimization and stochastic model predictive control of a dynamic tubular precipitator.
Altogether my focus points have revealed some bigger gaps holding back current biomanufacturing practices as a whole. With so many devils lurking in these (sincerely non-academic) details, I hope to continue engaging the biopharmaceuticals sector to help close some of these gaps in my future work.
