| dc.description.abstract | Global demand for food, pharmaceuticals and energy is constantly increasing, and the transition from a historically hydrocarbon-based economy to a green and renewable energy based-economy is accelerating due to the demands of society and the constraints of sustainability. The emerging technologies are based around carbon capture and storage (CCS), conservation of the ozone layer, electrocatalytic routes for hydrogen production, and bio-based renewable processes.
Biochemical technologies and systems biology present a reliable alternative to the traditional hydrocarbon processes due to lower energy demand requirement, and their selective biochemical routes. Industrial microbial biocatalytic processes include activities related with strain development, microbial cultures in different reactor configurations, bioprocess control and optimization, separation and purification. Microbial cultures are by nature multiple-input, multiple-output (MIMO) systems, and their processes must ensure a controlled and stable operation, to reach the desired growth and green production of chemicals. Model-free based approaches can be implemented for control and optimization. However, understanding of the process and its limitations require modelling techniques that consider conservative equations and kinetics of the biochemical process. Moreover, understanding of nature, modelling, control and optimization of biochemical processes can only promote, and aim at desirable levels of concentrations, titers, rates and yields.
This Ph.D. thesis presents modelling techniques applied in the industrially relevant Corynebacterium glutamicum, a gram-positive, rod-shaped bacteria widely used as industrial workhorse in the production of high-value chemicals and amino-acids.
Firstly, dimensional analysis was studied. The method corresponds to a traditional chemical engineering approach, that despite being common knowledge, it is normally forgotten as a model development technique. The thesis has a derivation of the dimensional analysis method from the analysis of the null space of the dimensional matrix in a CSTR and in two Monod models, and the implications in model identification, structural identifiability, control, and optimization is studied through two conference papers and a tutorial review research article.
Secondly, an experimental study using different sensors was addressed. The sensors had a different measurement principles, frequency and sensitivity, and their signals were correlated with optical density and cell dry weight. It was possible to compare their performance through scaling, and the sensitivity of the sensors to air flow and stirring rates was evaluated, with the latter affecting all the sensors. In this thesis, algorithms to evaluate their performance are presented to obtain an online estimation of growth and microbial growth rate in our bioreactor set-up. The studied sensors exhibited steady state noise that posed a drawback when computing the time derivatives, and the problem was addressed with the implementation of a Gaussian filter.
Finally, a proposal to link process and metabolic model is presented; the connection is possible through an optimization routine. The method considers the experimental measurements from a complex mixture of sugars, to develop a process model that is then linked with a metabolic model. The results provide an estimation of flux distribution in the microbial culture. | en_US |
