Comparative study of turbulence models for hemolysis prediction and hydraulic performance in centrifugal blood pumps.

BACKGROUND: Accurate prediction of hydraulic performance and hemolysis in centrifugal blood pumps is essential for improving efficiency and reducing blood damage. Computational fluid dynamics (CFD) predictions are highly sensitive to turbulence modeling, yet systematic validation for complex rotating blood flows remains limited. METHODS: Four turbulence models: RNG k-ε, k-ω SST, RSM-ω, and IDDES were evaluated for predicting hydraulic performance and hemolysis in the FDA benchmark centrifugal blood pump. Predicted pressure head and velocity fields were compared with experimental data. Hemolysis was quantified using plasma-free hemoglobin measurements at two operating conditions: 2500 RPM at 2.5 L·min-1 and 3500 RPM at 6 L·min-1. FINDINGS: RANS models (k-ω SST and RNG k-ε) showed strong agreement with experimental pressure head data, with errors of 1.1-4.7%. All models predicted hemolysis within the experimental uncertainty, with accuracy depending on operating condition. RSM-ω and k-ω SST performed better at 2500 RPM, whereas RNG k-ε and IDDES showed improved agreement at 3500 RPM. Scale-resolving models captured high-shear regions and vortex structures associated with blood damage more effectively. INTERPRETATION: Turbulence model choice strongly influences hemolysis prediction. While RANS models reliably predict hydraulic performance, scale-resolving approaches better represent flow features relevant to hemolysis, particularly at higher rotational speeds.