Efficient Impeller Optimization for Centrifugal Pumps Based on Kriging Surrogate Model With Multipoint Infill Strategy

Abstract To improve the operational efficiency of centrifugal pumps, optimization methods based on intelligent algorithms often incur high computational costs, while those relying on surrogate models are largely dependent on the accuracy of the surrogate models. This study proposes an optimization approach based on a Kriging surrogate model with a multisample infill criterion, aiming to achieve efficient pump design while reducing computational expense. Using the centrifugal pump model under investigation as the research object, the geometric shape of the impeller blades was parameterized, and a numerical simulation model was established. The proposed fast multipoint expected improvement and minimal surrogate prediction (Fq-EIO) criterion was employed to optimize the pump performance, improving efficiency while satisfying the head requirements. The results indicate that the optimized impeller structure significantly improves the internal flow characteristics by reducing low-pressure regions and vortices, leading to an approximate 4.4% increase in efficiency under the design operating condition, while maintaining stable head performance. These findings validate the effectiveness of the proposed optimization strategy in enhancing the performance of centrifugal pump equipment.