Modeling and optimization of bioreactor processes

This chapter focuses on the topic of dynamic modeling for bioreactor monitoring, optimization, and control applications. The first part of the chapter overviews mechanistic modeling across different scales, covering the concepts of structured/unstructured, segregated/unsegregated, and genome-scale modeling. The second part of the chapter covers machine learning methods for supervised, unsupervised, and reinforced learning in a bioprocessing context, with emphasis on building supervised bioreactor models that improve with process experience. Knowledge abstraction in the machine learning world is hardly compatible with the vast wealth of engineering and scientific knowledge accumulated over decades in the form of mechanistic models. The opportunities to develop hybrid mechanistic/machine learning models for bioreactors in the context of Industry 4.0 are finally highlighted. The vision is that machine learning should augment mechanistic bioreactor models rather than replace them. Several case studies are presented to illustrate the presented methods.