Multi-objective optimization of double suction centrifugal pump

Abstract

Double suction centrifugal pumps are widely used for water supplying system. In this study, the original design of a double centrifugal pump lacked sufficient head at the design flow rate condition. Therefore, the most important objective was to optimize the design to improve the head. A strategy inspired by “liquid–gas cavitation process” is innovatively used for intelligent global search of better pump designs with both higher head and wider-higher efficiency. This strategy has advantages including flexibility, parallelism, and feasibility on overstepping the local-best. The computational fluid dynamics and artificial neural network are used. It helps this optimization to find unknown points in the non-linear and multi-dimensional searching space, and accelerate the optimization process. Candidates were found after search, and the best one was chosen using Pareto principle. Experimental and numerical studies verify that the optimized impeller meets the requirement of head. The efficiency is also significantly improved with higher best efficiency and wider high efficiency range than original design. The critical cavitation is also improved at design condition. This study provides an effective strategy and a good solution for multi-objective optimization of double suction centrifugal pumps. Moreover, this study provides references for the combination of optimizations with artificial intelligence especially in the pump’s design.