Performance and cavitation in automotive centrifugal pumps: experimental analysis and 3D-CFD modelling assessment

Centrifugal pumps are challenging components in several applications, including automotive cooling systems, where compact design, high efficiency and cavitation resistance are essential. This study develops and validates a predictive 3D-CFD methodology for the estimation of both performance and cavitation in complex-geometry centrifugal pumps adopted in high-performance vehicles. Two single-stage, single-suction pumps with comparable dimensions but distinct designs are investigated through a combination of experiments and CFD analyses. Experimental results are analysed using dimensionless coefficients, introducing a novel Performance Factor (PF) based on turbomachinery similitude to correlate cavitation onset with flow coefficient (φ) and cavitation number (σ). Pump X starts to exhibit cavitation for σ<0.5 at φ=0.20 and for σ<1.2 at φ=0.34. Pump Y shows cavitation at higher fluid temperature for σ<0.8 and 0.21<φ<0.24. As for the simulations, they compare three turbulence models (Realizable k-ε, k-ω SST, and Elliptic Blending Reynolds Stress Transport) and three rotational modelling approaches (frozen rotor, mixing plane, and sliding mesh), combined with the Schnerr-Sauer cavitation model. Quantitative comparison with the experimental data demonstrates that the k-ω SST turbulence model provides the best trade-off between accuracy and computational cost, with an average deviation of 4.3 % for pump X and 3.0 % for pump Y in predicting performance. The Elliptic Blending RST model reduces the deviation to 2.9 % but increases computational time by 70 %, limiting its practical use. Among the rotational models, the sliding mesh approach achieves the highest accuracy (4.3 % and 3.0 % deviation for pumps X and Y, respectively), while steady approaches (frozen rotor and mixing plane) show deviations up to 12.2 %, especially in off-design conditions. In cavitating regimes, sliding mesh and k-ω SST accurately capture the head losses, whereas alternative combinations significantly underestimate them. Additionally, mesh sensitivity analyses reveal that cavitating conditions require finer meshes than non-cavitating ones to accurately predict vapor formation. The adopted CFD framework thus provides a validated, computationally efficient, and predictive tool for the design and optimization of compact centrifugal pumps in automotive and other high-performance thermal management applications.