(195a) Efficient Numerical Optimization of an Aerated Mixing Vessel for Fermentation

Oxygen is a critical component of aerobic fermentations and is sparingly soluble in aqueous solutions. Due to the slow mass transfer of oxygen from the bubbles into the liquid the rate of cell metabolism depends on the rate of oxygen supply from the gas phase. Mass transfer across gas-liquid interfaces plays also a vital role in many other reaction systems. The mass transfer is a function of the gas-liquid surface area which is a function of the coalescence and breakup rate which is again a function of the local velocity and turbulence. The local flow condition are finally determined by the reactor geometry, the fluid properties and – in this study – by the rotational speed of the impeller. So from a reactor performance standpoint, high aeration rates and high agitation rates are favorable.

On the other hand the operating costs are mainly determined by the total power consumed for the aeration and by the impeller of an aerated mixing vessel and should be as low as possible.

Traditionally dimensionless correlations and expensive experimental investigations are used to design stirred tank fermenters in production scale. However, design methodologies available for stirred tank fermenters are often insufficient and are using simplifications which affects e.g. the mass transfer significantly. Using computational fluid dynamics (CFD) based on validated models enables using fewer assumptions on the geometry and physics resulting in a better insight into the system.

In this work the performance of a stirred tank fermenter is going to be optimized using CFD simulations in combination with design explorations. After validating some implemented models especially to incorporate bubble size distribution including coalescence and breakup in the finite volume code STAR-CCM+, HEEDS is used to explore the design space. HEEDS is a multi-objective optimization tool using hybrid, adaptive search strategies to find better designs, faster and couples seamlessly to STAR-CCM+, so that no model simplifications are needed. Details how this approach works and how it can be applied to other problems will be shown. Finally the obtained Pareto front w.r.t. power consumption and mass transfer rate will be presented.