Multi-Objective Optimization of Two-vane Pump Based on NSGA-III Algorithm

Two-vane pumps, as the core equipment for sewage treatment, are subject to severe clogging and wear during operation. This paper proposes a multi-objective optimization strategy based on the CFD-DEM coupling method, random forest, and NSGA-III to improve the discharge performance of a two-vane pump. Firstly, the optimal Latin hypercube sampling is used to generate 60 samples. Secondly, the random forest algorithm is used to build an approximate model between the design variables and optimization objectives, and then NSGA-III is used to search the Pareto front. Finally, the TOPSIS with entropy weight is used to objectively select the optimal pump for the Pareto optimal solution set. Under clear water conditions, the optimized pump has a significantly higher head and efficiency. Under the rated solid-liquid two-phase flow conditions, the head and efficiency of the optimized pump are increased by 4.14% and 1.45%, respectively. The passing ability of the rag in the pump is significantly improved and the wear of the rag on the pump is greatly reduced by 36.4%.