As a key apparatus for cell culture in the biopharmaceutical industry, the stirred bioreactor holds irreplaceable value in mammalian cell cultivation due to its ease of operation and low maintenance costs. However, the growth environment required for mammalian cells is highly demanding. Ensuring high mass transfer efficiency in a stirred bioreactor while simultaneously maintaining conditions suitable for cell growth remains a persistent challenge. Against this backdrop, computational fluid dynamics (CFD) is employed to numerically simulate the stirred bioreactor for structural optimization. Furthermore, a novel impeller optimization method integrating machine learning is proposed to enhance reactor performance. A novel method was proposed for optimizing the impeller structure of stirred bioreactors suitable for animal cell culture by integrating machine learning techniques. Specifically, a data prediction model based on CFD and the Crested Porcupine Algorithm-optimized BP neural network (CPO-BP) was developed. The NSGA-II multi-objective optimization algorithm was employed to enhance the bioreactor’s performance. Key impeller parameters—impeller diameter ( D ), blade width ( B ), number of blades ( N B ), and blade inclination angle ( θ )—were selected as optimization variables. A multi-objective optimization problem was formulated to balance the volumetric mass transfer coefficient ( k L a ) and minimize fluid hydrodynamic stress ( F ). The final optimized parameters were D = 144 mm, B = 22 mm, NB = 6, and θ = 55°. Compared to the original design, the optimized configuration improved mass transfer performance by 54.7%, while the fluid hydrodynamic stress within the reactor increased by only 18.6%.
Wang et al. (Sun,) studied this question.