Multi-objective optimum design of high specific speed mixed-flow pump based on NSGA-Ⅱgenetic algorithm

Abstract: With the vigorous promotion of the national strategy for hydraulic engineering, a demand for mixed-flow pump in infrastructure has been increasing for years. As a kind of high-performance pump, the high specific speed mixed-flow pump has the advantages of stable performance and wide application area. With the increasing of specific speed, the high specific speed mixed-flow pump has begun to replace the axial-flow pumps in some application areas in recent years. It is necessary to optimize the impeller performance for it affects the pump performance directly. The velocity torque distribution along mixed-flow pump impeller is a significant parameter, which plays an important role in energy conversion between impellers and fluid. In the design of dualistic theory of mixed-flow pump impeller, the regulation of velocity torque distribution should be defined. But, there is no uniform method of expression as well as specific rule for designers to follow, and too much experience is needed to rely on. It has far-reaching meaning for the promotion of the level of mixed-flow pump design and the performance by establishing the optimization parameter model of velocity torque distribution along impeller. In order to further research the hydraulic performance of optimization method for high specific speed mixed-flow pump, a mixed-flow pump whose specific speed is 803 was chosen as the research object, and the commercial software CFX and the shear stress transport turbulence model were applied to compute the interior flow field within the pump. In this paper, the efficiency and the head were chosen as the optimization objectives, and 3 parameters that describe the velocity torque distribution were chosen as the optimization parameters, which were used to parameterize the impellers. The uniform design was adopted to arrange the sample space, the RBF (radial basis function) neural network was used to fit the relationship between the variables and objectives, and finally the NSGA-Ⅱ genetic algorithm was used for multi-objective optimization. Moreover, the difference of the internal flow field was obtained by comparing the initial linear distribution individual with the optimal efficiency individual and the optimal head individual respectively, which were selected from the Pareto solutions. The velocity torque distribution and the change trends of the wrap angel along the axial plane streamline between the initial and optimal individual were analyzed. The wrap angel values of optimal efficiency and optimal head were 75.15° and 67.85°, respectively. The variation trends of the wrap angel of micro-element based on optimal efficiency individual were contrary with the initial one, while the variation trends based on optimal head individual were the same as the initial one, but the change range was enlarged. Compared with the initial linear distribution individual, the efficiency of the optimal efficiency individual was improved by 1.12% through experimental verification. Distribution maps of the relative streamline, velocity and static pressure of the impeller guide vane at 0.5 times blade height and the static pressure of the corresponding chord length at 0.2, 0.5 and 0.8 times blade height were showed. And, the possible reasons for the differences between initial individual and optimal individuals were given. It was found that, utilizing the RBF neural network combined with NSGA-Ⅱ genetic algorithm, the effect of optimizing the impeller hydraulic performance of high specific speed mixed-flow pump was remarkable. The research provides a certain theoretical reference for further improvement of the performance of high specific speed mixed-flow pump.