Research On Performance Prediction Model For Multistage Centrifugal Pump Based On MATLAB

Multistage centrifugal pump is a mechanism that converts the mechanical energy into kinetic energy of the output liquid. The accuracy of predictive model directly affects the efficiency of the entire pump system. The consummate model for multistage centrifugal pump has not been formed. In recent years, with the development of computer technology, MATLAB software has been widely used in establishing and predicting the unknown model. In this paper, the black-box modeling method is proposed for the performance predictive modeling of multistage centrifugal pump. The main work of the paper is as follows:1) The principle of least-squares curve fitting method was introduced, and the measured data was used to acquire the relation of head, power and flow in multistage centrifugal. In addition, the fitting curves visualization was created by the MATLAB GUI toolkit, based on the least squares method.2) The prediction model based on BP neural network toolbox is formed using the data of flow, lift and efficiency as the input, the data of power as the output. The forecast results are also analyzed.3) The genetic algorithm is used to optimize BP neural network algorithm. Compared to the BP neural network model of multistage centrifugal pump, the genetic algorithm optimization model is more accurate.The work of the paper is supported by the national science and technology support Project of China "The research and application of large fluid mechanical energy-saving technology (2013BAF01B02)".