CFD simulation and Pareto-based multi-objective shape optimization of the centrifugal pump inducer applying GMDH neural network, modified NSGA-II, and TOPSIS

Abstract

Inducer is an important device which is mounted upstream of the inlet to the main impeller of the centrifugal pump and rotates at the same rotational speed as the impeller. The main purpose of the inducer is to improve the suction performance of the pump, but this improvement is dependent on the geometrical parameters of the inducer. Therefore, it is essential to optimize these parameters. In the present study, the performance of an inducer is optimized by considering the inlet tip blade angle, the outlet tip blade angle, and the ratio of the outlet hub radius to inlet hub radius as design variables and the head coefficient, the hydraulic efficiency, and the required net positive suction head (NPSHR) as objective functions. The inducer performance is simulated using 3-D computational fluid dynamics (CFD) and compared with experimental data, which shows the validity of the used method and assumptions. Then the group method of data handling (GMDH) algorithm is used to model the objective functions with respect to design variables. Using the modified non-dominated sorting genetic algorithm II (NSGA-II) approach, Pareto fronts are then plotted and trade-off optimum points are obtained using the technique for order of preference by similarity to ideal solution (TOPSIS). Using multi-objective optimization, the head coefficient, the hydraulic efficiency, and NPSHR are improved 14.3%, 0.3%, and 30.2%, respectively. Recommended design points unveil significant optimum design principles that can be obtained only by using a multi-objective optimization approach.