Approaching a modified adaptive swarm intelligence to energy efficiency enhancement of an inline pump

Pumps, as widely used general-purpose machinery, consume significant amounts of energy, making improvements in their operational efficiency and stability critically important. However, current mainstream intelligent optimization methods are limited by prediction accuracy and algorithm performance. This study proposes an adaptive improved particle swarm optimization algorithm with rapid convergence capabilities, integrating population classification, adaptive acceleration strategies, and self-learning mechanisms. Combined with computational fluid dynamics , the algorithm was applied to directly optimize a vertical inline pump with a specific speed of 107. The study performed statistical correlation analysis on 4000 design samples generated during the optimization process and utilized flow field loss visualization techniques to explore the causes of performance improvements. The results indicate that the proposed improved algorithm achieved approximately a 6 % performance improvement within 25 iterations and stabilized within 85 iterations. Experimental validation confirmed the enhanced performance of the optimized model, with a 10.71 % increase in nominal efficiency, a 5.09 % increase in head, and an improvement in Minimum Efficiency Index (MEI) from 0.39 to 1.09, indicating significant expansion of the high-efficiency operational range. The method proposed in this study significantly improves the optimization efficiency and accuracy for multi-parameter centrifugal pumps , providing theoretical support and technical solutions for high-efficiency and high-stability hydraulic designs. • Proposed a modified fast-converging swarm intelligence for pump optimization. • Experimentally validated significant improvements in pump performance. • Explored design-variable influence patterns via statistical correlation analysis. • Identified performance enhancement mechanisms through flow-loss visualization.