Abstract: Based on the boosting cell parameters of independently developed third generation muhiphase pump YQH—100, orthogo—nal experimental design was used to design the optimization scheme.Using relative head and efficiency as the evaluation targets for the optimization design, the relative head and efficiency of the multiphase pump were predicted by software FLUENT.BP neural networks and GRNN model were adopted to construct the response relation between the design variables and the objective function.The results of prediction show that the prediction performance of GRNN model Was much better than that of BP neural network.The optimization resuits show that relative head and efficiency of the muhiphase pump were increased by 0.74%and 0.83%, respectively.