Hypothesis-driven compartment model for stirred bioreactors utilizing computational fluid dynamics and multiple pH sensors

The ability to understand and predict mixing behaviour in stirred bioreactors through the use of mathematical models is an effective strategy to improve or ensure the performance of a process. Historically, compartment models have been developed to facilitate the prediction of hydrodynamics and in the last decade, computational fluid dynamics (CFD) has become the standard tool to assess this in detail. Despite the continuous increase in computational capacity, performing CFD simulations of bioreactors remains a challenge in industry, where the need for simplified models is apparent due to the complex phenomena occurring at the industrial scale. The combination of CFD and compartment models has been investigated in this work using a pilot scale stirred bioreactor equipped with three Rushton disc turbines and multiple pH sensors as a case study. A hypothesis-driven compartmentalization strategy, comprised of five steps, has been developed and applied to the pilot scale stirred bioreactor. The compartmentalization resulted in 56 compartments with unidirectional flow between adjacent compartments. The performance of the methodology was evaluated against a data-driven compartment approach and the full CFD simulation, in terms of its ability to recreate transient tracer profiles following top and bottom feeding of the tracer. The data-driven compartment model proved to be the most accurate of the three investigated methods, whereas the hypothesis-driven compartment method had a 10-12% higher error on prediction. Nevertheless, the developed methodology should be considered a viable alternative to conventional CFD methods, especially when models of complex phenomena - for example multi-dimensional population balance models - are to be incorporated in the final model as well.