Constrained CFD-RSM optimization for centrifugal pump head enhancement with minimal efficiency loss

This study tackles the persistent head–efficiency trade-off in centrifugal pump impeller design by developing a data-efficient, constraint-based optimization framework that integrates CFD with response surface methodology (RSM). A three-dimensional CFD model of the target pump is constructed and validated against experimental performance data, ensuring reliable prediction of head and efficiency. Eight geometric variables are then evaluated through a Plackett–Burman screening design, which identifies impeller outlet diameter, outlet width, and blade number as the dominant factors governing head variation. Because a full-factorial curvature check indicates strong response nonlinearity, a quadratic surrogate model is adopted. A Box–Behnken design is subsequently used to build second-order response surfaces for head and efficiency, supporting rapid sensitivity analysis and constrained optimization. The objective is to maximize head while limiting efficiency deviation to within ± 1.5 % of the baseline pump. The optimized parameter set is verified using additional CFD simulations at the rated condition. Relative to the baseline, the optimized impeller raises rated head from 52.38 m to 56.02 m, a 6.95% increase, while reducing efficiency by only 0.12%. This loss lies within typical CFD numerical uncertainty, suggesting practical near-decoupling of head and efficiency. Flow analysis indicates that separation and vortex dissipation near the blade trailing edge constrain efficiency. Further local sensitivity analysis shows that increasing the blade outlet angle from 24° to 25° suppresses separation and improves both head and efficiency. The CFD–RSM strategy achieves substantial head improvement with negligible efficiency loss, offering a practical route for pump upgrading in retrofit and redesign applications.