Particle image velocimetry and computational fluid dynamics applied to study the effect of hydrodynamics forces on animal cells cultivated in Taylor vortex bioreactor

• Autor(a) principal: Singh, Harminder
• Data de Publicação: 2016
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/7303
• topic: DNS; LES-WALE; RANS; Viscous energy dissipation rate; Turbulence; Dissipação de energia viscosa; Turbulência; ENGENHARIAS::ENGENHARIA QUIMICA::TECNOLOGIA QUIMICA

Taylor-Vortex reactor (TVB) is fast becoming the next bioreactor to culture animal cells due to milder shear and homogeneous flow structures through-out the bioreactor in comparison to the traditional stirred vessels. However, there is little information in the literature for the TVB on the viscous energy dissipation rate (VEDR), which is considered the ideal parameter to characterize the cell death, and its geometrical aspects, which may affect the culture of animal cells resulting in poor efficiency. Consequently, this work focuses on: the estimation of the VEDR of mean flow and turbulent kinetic energy (TKE) using an experimental 2D particle image velocimetry (PIV) method and a computational fluid dynamics (CFD) method using different turbulence models, principally the direct numerical simulation (DNS) model; and, the impact of the off-bottom clearance area and the external cylinder’s bottom shape on the flow structures of TVB. Both numerical and experimental methods confirm that the bulk zone comprising of the 80 % of the gap-width, where the cell cultures will spend most of the time, has a near constant velocity magnitude of around 50 % of the tip velocity and VEDR values which are around 10 times lower than at the walls. Qualitatively, the DNS model predicted well the flow structure of both mean and turbulence parameters in comparison with the experimental PIV predictions. However, quantitatively only the mean velocity predictions are in good agreement with the PIV data with certain amount of under-estimation of the turbulence parameters. Among different turbulence models, the large eddy simulation (LES) - wall adapting local eddy-viscosity (WALE) model presented best comparison with the DNS model data for all the flow parameters; while, the Reynolds stress model and the LES-Smagorinsky models were the poorest. On the other hand, the Reynolds averaged Navier-Stokes (RANS) based two equation models estimated well the mean velocity components in comparison with the DNS model data, but could not capture well the flow structures of the turbulence components. The geometrical features of curved surface of outer bottom and off-bottom clearance area which are of practical importance in stirred vessels, impact adversely the flow structures in the TVB due to poor axial velocity component. In comparison with the spinner vessel, a stirred tank type bioeactor but with lower shear, for similar Re/ReT ratio, the maximum and mean VEDR were always found to be of lower magnitude values, and due to much less difference between the maximum and the mean values, the TVB presents more uniform structures in comparison to the spinner vessel.