Gas−Liquid Flow Generated by a Pitched-Blade Turbine: Particle Image Velocimetry Measurements and Computational Fluid Dynamics Simulations

Validation of a CFD model for cell culture bioreactors at large scale and its application in scale-up

Among all the operating parameters that control the cell culture environment inside bioreactors, appropriate mixing and aeration are crucial to ensure sufficient oxygen supply, homogeneous mixing, and CO2 stripping. A model-based manufacturing facility fit approach was applied to define agitation and bottom air flow rates during the process scale-up from laboratory to manufacturing, of which computational fluid dynamics (CFD) was the core modeling tool. The realizable k-ε turbulent dispersed Eulerian gas-liquid flow model was established and validated using experimental values for the volumetric oxygen transfer coefficient (kLa). Model validation defined the process operating parameter ranges for application of the model, identified mixing issues (e.g., impeller flooding, dissolved oxygen gradients, etc.) and the impact of antifoam on kLa. Using the CFD simulation results as inputs to the models for oxygen demand, gas entrance velocity, and CO2 stripping aided in the design of the agitation and bottom air flow rates needed to meet cellular oxygen demand, control CO2 levels, mitigate risks for cell damage due to shear, foaming, as well as fire hazards due to high O2 levels in the bioreactor gas outlet. The recommended operating conditions led to the completion of five manufacturing runs with a 100% success rate. This model-based approach achieved a seamless scale-up and reduced the required number of at-scale development batches, resulting in cost and time savings of a cell culture commercialization process.

A Comparative CFD Study on Gas-Liquid Dispersion in A Stirred Tank with Low and High Gas Loadings

The computational fluid dynamics (CFD) combined with a population balance model (PBM) was applied to simulate gas-liquid dispersion in a stirred tank with low and high gas loadings. The model predictions were validated by using the data in the literature. The simulation results show that the flow patterns and gas dispersion characteristics are very different in the stirred tank for low and high gas loadings. A typical two-loop flow pattern forms as that in single-phase stirred tank for low gas loadings, while a triple-loop flow pattern, with two recirculation loops above and one below the impeller is found in the tank for high gas loadings. Shaft power input of impeller agitation plays a major role for gas dispersion with low gas loadings. For high gas loadings, the potential energy due to gas sparging has significant effect on gas dispersion and can not be neglected. Compared to low gas loading, high gas loading causes average gas holdup increased in the stirred tank, while relative local gas holdup in the lower circulation-loop region and near-wall region reduced. The ability of impeller agitation for gas dispersion reduces with high gas loadings, and mean bubble size becomes larger and the volume-averaged bubble size distribution is wider.

Optimizing the rotor design for controlled-shear affinity filtration using computational fluid dynamics

Controlled shear affinity filtration (CSAF) is a novel integrated processing technology that positions a rotor directly above an affinity membrane chromatography column to permit protein capture and purification directly from cell culture. The conical rotor is intended to provide a uniform and tunable shear stress at the membrane surface that inhibits membrane fouling and cell cake formation by providing a hydrodynamic force away from and a drag force parallel to the membrane surface. Computational fluid dynamics (CFD) simulations are used to show that the rotor in the original CSAF device (Vogel et al., 2002) does not provide uniform shear stress at the membrane surface. This results in the need to operate the system at unnecessarily high rotor speeds to reach a required shear stress of at least 0.17 Pa at every radial position of the membrane surface, compromising the scale-up of the technology. Results from CFD simulations are compared with particle image velocimetry (PIV) experiments and a numerical solution for low Reynolds number conditions to confirm that our CFD model accurately describes the hydrodynamics in the rotor chamber of the CSAF device over a range of rotor velocities, filtrate fluxes, and (both laminar and turbulent) retentate flows. CFD simulations were then carried out in combination with a root-finding method to optimize the shape of the CSAF rotor. The optimized rotor geometry produces a nearly constant shear stress of 0.17 Pa at a rotational velocity of 250 rpm, 60% lower than the original CSAF design. This permits the optimized CSAF device to be scaled up to a maximum rotor diameter 2.5 times larger than is permissible in the original device, thereby providing more than a sixfold increase in volumetric throughput.