Multi-objective optimization design in a centrifugal pump volute based on an RBF neural network and genetic algorithm

Abstract

The hydraulic and acoustic performance of centrifugal pumps are interrelated and contradictory in hydraulic structure design. An optimization design method of a volute based on a radial basis function (RBF) neural network and genetic algorithm (GA) is proposed to solve this problem. The efficiency and total sound pressure level of the centrifugal pump are used as the optimization objectives. The diameter of the base circle, the width of the base circle, the installation angle of the volute tongue, and the height of the volute diffuser tube are used as the optimization variables. Latin hypercube sampling (LHS) is used to establish the sample space, the RBF neural network is used to build the agent model between the optimization variables and objectives, and the GA is used to multi-objective optimization. For the Pareto solution set obtained, two extremal individuals and initial individuals are selected for comparative analysis of hydraulic and acoustic performance under different working conditions. The result demonstrates that under the rated working condition, compared with the initial individual, the efficiency of the optimal individual of efficiency increases by 3.79%; the internal noise of the optimal individual of sound pressure level decreases by 5.5%, and the external noise decreases by 2.3%.