Animal cell culture: macroscopic modeling, estimation and control

Animal cell cultures are targeted namely at the production of biopharmaceutical products such as vaccines, recombinant proteins and antibodies. The industrial production of these promising target-specific drugs is fairly recent. There are largely unexplored aspects that could be improved on so that a lower end-user price can be sustainable, namely, a more rational use of the culture medium.

In order to explore this, a model needs to be considered. In this thesis, we propose a practical approach of identifying model parameters with two experimental case studies (CHO-S and interferon-gamma producing CHO-320 cultures): a step-by-step parameter identification procedure that uses gradually more complex models and takes its inspiration from the analysis of the sensitivity functions.

The model is then useful for the control of a bioreactor. A focus is put on the continuous perfused production regime that allows a small volume to be cultured for a longer period and allows for a faster downstreaming. A controller can generate a manipulation of flowrates such that concentrations remain close to setpoints.

This research work then addresses the mathematical possibility of estimating some concentrations that cannot be measured in real life from the knowledge of other concentrations that can be measured online through probes currently available on the market. An illustration is provided via an implementation of the extended Kalman filter algorithm.

Finally, model-based automatic control is studied and the usefulness of nonlinear model predictive control highlighted. A model-plant mismatch case study serves to pinpoint the importance of how well biochemical kinetics are captured by the model in order to prevent persistent waste of culture medium.