From Physics to Bioengineering: Microbial Cultivation Process Design and Feeding Rate Control Based on Relative...

Area of Science:

• Biotechnology
• Process Engineering
• Control Theory

Background:

• Industrial fed-batch cultivation traditionally relies on open-loop feeding control due to historical reasons, regulations, and the need for reproducibility.
• Existing methods often require continuous monitoring, such as glucose concentration measurements, for effective process management.

Purpose of the Study:

• To develop a numerical semi-global convex optimization procedure for calculating feeding rate time profiles in fed-batch cultivation.
• To enhance reproducibility, controllability, and process performance in industrial biotechnological processes.
• To create a method that reduces reliance on real-time measurements like glucose concentration.

Main Methods:

• Utilized a generic gray box biomass modeling procedure incorporating relative entropy as a key optimization metric.
• Introduced a multivariate path of Lagrange multipliers and a concept of nuisance time for posterior distribution approximation.
• Developed a numerical semi-global convex optimization algorithm for feeding rate profile calculation.

Main Results:

• The proposed optimization method, based on relative entropy, is applicable beyond specific gray box models and bioengineering contexts.
• The technique offers advantages for both feed-forward and adaptive control systems requiring biomass growth prediction models.
• Identified gray box model parameters enable exploration of alternatives in controllable industrial biotechnological processes.
• Demonstrated that glucose concentration measurements become unnecessary for process development post-model identification.

Conclusions:

• The developed numerical optimization procedure offers a robust and versatile tool for designing fed-batch cultivation processes.
• This approach significantly improves process control, reproducibility, and performance, while simplifying operational requirements.
• The method's independence from specific models and real-time measurements broadens its applicability in industrial biotechnology.