Application of neural network to performance predicting of multiphase rotodynamic pump

Based on the third generation YQH-100 multiphase rotodynamic pump independently developed as the research object, The models of BP and RBF neural network were established respectively to predict the Multiphase Rotodynamic pump energy characteristics. The 27 groups of sample data of neural networks were created by FLUENT numerical calculated. The performance data of 20 multiphase rotodynamic pumps were used to train the two models and the data of the other 7 multiphase rotodynamic pumps were used to test the two models. The predict results show that the predicted of two models were favourably accorded with experiment. So it is possible to use BP and RBF neural network for predicting performance of multiphase rotodynamic pumps, which can reduce cost and shorten experimental time.