Applying Computational Fluid Dynamics Technology in Bioprocesses-Part 2

Computational fluid dynamics can resolve performance problems.
May 01, 2010

ABSTRACT

Computational fluid dynamics (CFD) has been widely applied as a trouble-shooting and Quality by Design (QbD) tool for bioprocess from downstream, to upstream, to fill–finish. Part 1 of this article series provided an overview of the technology, and the model setup. Here in part 2, we provide a detailed review of the applications of CFD in cell culture and fermentation processes, including flow characterization, mixing, resuspension, scale-up, and cell damage.


(AMGEN, INC.)
Computational fluid dynamics (CFD) is applied in bioprocesses to collect a wide variety of information from many diverse and complex systems comprising complex bio-fluids and particulate matter. Work has ranged from trouble-shooting to process design and optimization. Processes such as filtration, centrifugation, chromatography, product freezing, thawing, and drying have benefited from CFD analysis in many ways. Here, however, we will focus on how bioreactor operations can benefit from CFD. Reviews of the applications of CFD to other processes shared with pharmaceutical industry, including material transport and storage, blending, granulation, milling, compression, film coating, and spray drying, appear in the literature.1,2

Mechanically agitated or sparged tanks with single or multiple impeller systems are used extensively in the biopharmaceutical industry for various mixing processes, such as fermentation, cell culture, crystallization and emulsification, and resuspension. The empirical approach is still the prevailing practice for process scale-up and scale-down and technology transfer. This article shows how CFD has been used as a Quality by Design tool and for troubleshooting bioreactors.

Process development engineers frequently face biopharmaceutical performance problems such as cell damage, optimizing hardware configurations including sparger design and impeller choice, and system equivalency for technology transfer. CFD can address the critical operating parameters of mixing time, power consumption, particle size distribution, gas distribution, and the just-suspended agitation speed.

FLOW CHARACTERIZATION

The flow field in an agitated bioreactor under turbulent flow conditions is quite complex, consisting of several time-dependent flow systems, including discharge flow originating from the impeller, blade wake, trailing vortex systems, impinging jets on the wall, boundary layers on the wall, baffle-bound vortex systems, and large recirculation zones. Primary flow patterns depend on the impeller geometry, impeller clearance, the presence of baffles, liquid height, and blade pitch. Because the tanks and impellers used in bioprocessing come in all sizes and shapes, flow characterization is difficult.

CFD has been widely used to investigate the flow dynamics in agitated and sparged tanks because it can provide detailed three-dimensional information about velocity and pressure. It can reproduce the key flow pattern—the radial flow generated by a Rushton impeller and the axial flow caused by a pitched-blade impeller. Moreover, it has been used successfully to capture and predict important flow characteristics, such as the large recirculation patterns, the trailing vortex structure in the vicinity of the impeller and cavity formation, and even three distinctly different stable flow patterns (parallel, merging and diverging) in a tank equipped with dual Rushton impellers.3–10

Recently, numerous studies have verified the existence of the macro-instability (MI), a large-scale low-frequency phenomenon occurring in a baffle-agitated tank.11–18 The MI, when it appears, scales with the rotational speed of the impeller, N. Its occurrence is dependent on both the impeller geometry and the location of the impeller relative to the tank walls. MIs can cause practical concerns, such as massive vibrations that lead to significant damage to the tank interior wall, impeller shaft, and impeller, and also may be responsible for some observed phenomena in agitated tanks, such as large dissolved oxygen (DO) fluctuations, which can lead to serious control problems.

Three frequencies have been discovered, with the lowest being 0.02 N, the highest around 0.2 N, and various intermediate frequencies in between. These MIs are associated with the mean velocity as determined over a short time interval, rather than with turbulent velocity fluctuations. The MI of the lowest frequency, only under specific conditions and primarily in a region in front of the baffle, is induced by a precessing vortex,17 whereas the MI of the highest frequency may be related to circulation and is present for a wide range of conditions and geometries across a significant part of the vessel volume.16 The existence of intermediate MIs indicates that there is some cross talk between the low- and high-frequency motions for a wide range of frequencies.13 One intermediate frequency MI, 0.06 N, appears between the turbine and the liquid surface, characterizing an organized modification of the flow pattern that creates a transient violent flow activity in the upper part of the vessel.14

CFD simulations have identified all three MI frequencies.6,11,12,16,19–24 A low-frequency precessional vortex motion in a vessel stirred by a Rushton turbine has MI frequency values nearly identical to the lowest reported frequency.23 Large-eddy simulation (LES) models have revealed unsteady asymmetric flow patterns with intermediate frequency MIs in a pitched-blade turbine (PBT) impeller stirred-tank that is consistent with reported digital particle image velocimetry data.19,13 Frequency analysis of CFD results has revealed an MI with a frequency of 0.186 N in a PBT tank with a Reynolds number >20,000, also matching experimental results from different scales of tank.16

These studies may provide valuable insights for interpreting many previously unexplained hydrodynamic phenomena observed in stirred-tanks, such as the time-dependent settling of solids in dilute suspensions, bi-modal and tri-model velocity histograms in laser-Doppler experiments, and large fluctuations of measured DO and pCO2.