Mixing studies in an unbaffled bioreactor using a computational model corroborated with in-situ Raman and imaging analyses

• Published in: Chemical engineering journal advances Vol. 9; p. 100232
• Main Authors: Vivek, Vasudevan, Eka, Fitriani Nur, Chew, Wee
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.03.2022; Elsevier
• Subjects: Bioreactor; Image analyses; Mixing time; Multi-physics computational model; Multivariate chemometrics; Raman spectroscopy; CFD; MRF; Mixing time; BTEM; CAD; Raman spectroscopy; RGB; RoI; Bioreactor; RANS; Image analyses; Multivariate chemometrics; Multi-physics computational model
• ISSN: 2666-8211; 2666-8211
• DOI: 10.1016/j.ceja.2021.100232

•Designed a computational model of a non-standard 10 L bioreactor.•Single-phase mixing simulations validated against literature.•Tracer mixing experiments performed at various stirrer speeds & liquid viscosities.•Mixing times calculated from image analyses & inline Raman spectroscopy.•Simulated mixing times validated with experimental mixing data. A computational model of an unbaffled 10 L bioreactor fitted with a unique serpentine tubing structure and two standard six-blade Rushton impellers was developed using Reynolds Averaged Navier-Stokes (RANS) turbulence models (standard k-ε and k-ω SST) and multiple reference frame (MRF) CFD modeling. Flow and mixing time simulations were validated with literature findings for baffled stirred tanks ranging from laboratory to commercial scales, viz. 5 L, 38 L, 22 m3 and 98 m3. Global quantities such as power number (Np), impeller flow number (NQ) and single-phase mixing times (t95) were verified with literature values. Flow and mixing simulations in the 10 L bioreactor were performed in full domain and a reduced 60° domain models. The reduced 60° domain model was selected for calculating mixing times at stirrer speeds from 50 to 500 rpm and fluid viscosities of 1 to 6 cP. This viscosity range corresponds to a typical high cell density E. coli fermentation run. For stirrer speeds more than 150 rpm in the 10 L bioreactor (Re > 3000), the mixing pattern in such high-density cultures could be reliably simulated with the standard k-ε model. Mixing time simulations in the 10 L bioreactor were validated against tracer mixing experiments using an inert red dye with HD video recording and inline Raman spectroscopy. Analyses of video images, univariate calculations and multivariate chemometrics on Raman data yielded experimental mixing times to compare with simulation results at four distinct spatial positions in the unbaffled 10 L bioreactor.