Centrifugal pump impeller and volute shape optimization via combined NUMECA, genetic algorithm, and back propagation neural network

Abstract

This paper presents a fast, efficient, and convenient shape optimization design method for the centrifugal pump impeller and volute. A meridional curve, stream surface, blade stacking, and two-dimensional blade profile are obtained for impeller parameterized fitting by NUMECA, which substantially decreases the parameters to be optimized. A combination of genetic algorithm (GA) and back propagation neural network (BPNN) is then employed to optimize the impeller design while preventing prematurity or stagnation due to the GA. The head and efficiency of the optimized impeller under the designed flow rate condition increase by 7.69% and 4.74%, respectively, while power decreases by 2.56% post-optimization. Static pressure in the optimized impeller middle span is more uniform post-optimization, and the hydraulic performance of the centrifugal pump with the optimized impeller exceeds that of the original centrifugal pump under low and designed flow rate conditions. Head increases by 2.69 m and efficiency increases by 4.32% under the designed flow rate condition as well. The base circle diameter, volute inlet width, and volute baffle tongue can be modified to optimize the volute shape design. The head of the centrifugal pump with the optimized volute and optimized impeller increases by 4.83 m and 6.35 m and efficiency increases by 9.12% and 18.65% under 1.2 and 1.4 times the designed flow rate compared to the pump with the original volute and optimized impeller. Vortices in the optimized volute are reduced significantly and particularly relative energy losses. Under low flow rate conditions, compared with the original centrifugal pump, the head and efficiency of the experimental centrifugal pump with optimized impeller and optimized volute increase by 1.56 m and 1.12%; under the designed flow rate condition, they increase by 4.34 m and 5.23%; and under the high flow rate condition, they increase by 3.71 m and 8.54%, respectively. Compared to the traditional optimization method, as evidenced by numerous shape optimization design cases, NUMECA-GA-BPNN produces better optimized shapes with stronger hydraulic performance more quickly and efficiently.