Modelling of gradients in industrial aerobic fed-batch fermentation processes

The bio-based manufacturing of many products of interest (e.g. enzymes and biomass) is performed with submerged aerobic fed-batch fermentation processes, with bioreactor volumes typically ranging between 20 and 1000 m3. With increasing scale, mixing may become limiting, and consequently gradients in relevant cultivation variables (e.g. substrate concentration) may arise.

These local non-homogeneous conditions have been shown to potentially affect cell physiology, ultimately leading to decreased process productivity and profitability. Knowing this, investigating the impact of gradients on large-scale fermentation processes has become a relevant subject of study from both academic and industrial viewpoints.

Numerical modelling of gradients is a valuable approach to study the magnitude and occurrence of gradients at industrial scale, as it provides a local description of the fermentation environment, and it can be applied to any fermentation process (i.e. those with different microorganisms, bioreactor types, volumes and operational conditions).

Gradient modelling is a challenging task, since the combination of both microbial kinetics and fluid dynamic models is required. This is typically done using Computational Fluid Dynamics (CFD) models combined with kinetic models, which result in a highly-detailed description of the broth environment for both flow (e.g. pressure, velocity) and fermentation process (e.g. substrate concentration, specific rates) variables.

However, CFD modelling in combination with microbial kinetics entails a high computational demand, leading to very long elapsed real times (several weeks) for the simulation of fermentation process snapshots (hundreds of seconds), which limits the use of CFD modelling. To overcome the high computational burden, compartment models can be developed.

These consist of collections of ideally mixed volumes with such flows and connections between them that resemble the flow pattern and magnitude of the large-scale bioreactor. Microbial kinetics can be implemented, and snapshot simulations can then be performed very quickly (several seconds). Up to now, CFD and compartment model simulations are restricted to fixed-volume processes, i.e. batch processes or snapshots of fed-batch processes.

Furthermore, the current compartment models do not have volume addition implemented to simulate entire fed-batch processes. This PhD thesis is a step forward in the field of gradient modelling, with focus on developing and utilising several modelling tools to assess the magnitude and occurrence of gradients in fermentation processes and estimate their impact on process performance.

First, a five-regime model describing the different metabolic behaviour of the industrially-relevant microorganism Bacillus licheniformis under glucose and dissolved oxygen concentration gradients has been experimentally calibrated and tested. Following the same model structure, kinetic models for Escherichia coli and Saccharomyces cerevisiae with model parameters from the literature have also been defined.

Subsequently, six case studies with varying reactor types (a bubble column and two stirred tanks with different impeller configurations) and operational conditions have been constructed for three stages of industrial aerobic fed-batch fermentation processes (40, 60 and 90 m3). CFD modelling has been used to describe the fluid dynamics and oxygen transfer in the different case studies.

Furthermore, compartment models have been developed based on the CFD results. The kinetic models previously developed and/or extracted from the literature have been implemented to all CFD and compartment models for the simulation of fermentation process snapshots. Finally, a methodology to develop dynamic compartment models that can account for volume addition in aerobic fed-batch fermentation processes has been constructed.

The resulting dynamic compartment models have been tested by simulating the entire aerobic fed-batch fermentation processes of B. licheniformis, E. coli and S. cerevisiae. The results have been analysed in terms of local glucose and dissolved oxygen concentrations, by-product levels, local metabolic regimes that the cells experience and influence of gradients on process metrics.

Two main findings arise from these results. First, it has been concluded that the current main limitation in the field of gradient modelling is the development of kinetic models that can describe product formation under different environmental conditions and glucose uptake at low rates more comprehensively, and that can take the cell culture history and the length and magnitude of fluctuations into account.

Secondly, glucose starvation (i.e. the glucose concentration level is such that the maintenance requirements are not fulfilled) has been determined to be the most frequent metabolic regime found in all simulations. The reason is the feeding strategy used, which manipulates the feed rate according to the oxygen transfer rate level by keeping the dissolved oxygen level at a constant value.

This leads to very low glucose concentration levels in the broth, therefore increasing the chances to develop glucose starvation if mixing limitations arise. Consequently, situations that prevent glucose starvation also prevent the development of significant gradients in most cases. Besides working with short mixing times, operating at high oxygen transfer rates has been found essential to prevent the development of significant gradients with all microorganisms.

From an operational perspective, the utilisation of bubble columns in comparison with stirred tanks has shown to lead to significantly lower mixing times and seems a promising bioreactor type for the prevention of gradients if sufficient oxygen transfer levels can be achieved. Overall, this PhD thesis contributes to the characterisation and prevention of significant gradients in industrial aerobic fed-batch fermentation processes, helping to guarantee adequate process performance in spite of potential non-homogeneous cultivation conditions.