Digital twin based feedback control of biomass specific rates in a Fab-fragment producing E. coli fed-batch process

Abstract

To ensure stable product quality and batch to batch reproducibility of fermentation processes, bioprocess control is the matter of choice. A majority of recombinant protein production processes are carried out in fed-batch mode, which enables control by the feed addition. Models are needed for the establishment and the investigation of more complex control objectives as a representation of the real process. These models could be extendedto applicable Digital Twins if they are integrated in the monitoring and control loop of the fermentation plant.The aim of the thesis is to find control laws to keep biomass specific rates of a recombinant protein producing E.coli fed-batch process constant and investigate the impacts and the usability of this new control approaches. Therefore an already existing process model was extended and fitted to historical datasets. Subsequently the model was used to carry out a nonlinear feedback linearization to obtain the demanded control laws. In addition to thewell known and often used control of the specific growth rate control laws were derived to control the biomass specific substrate uptake rate, the recombinant protein production rate as well as the biomass specific oxygen uptake and carbon dioxide production rate.To judge the applicability of these control laws an implementation study was carried out. Therefore an independent verification plant was simulated as a substitute for real experiments. To apply the control laws in a meaningful way a setpoint optimization was performed beforehand and the controllers were tuned.The control laws could successfully be used to keep the biomass specific rates at constant setpoints. Additionally the implementation study showed an theoretical potential to improve process performance using the determined optimized setpoints. Eventually the potential to apply this novel control approaches and the model to real experiments is discussed with regard to an applicable digital twin.