Multi-objective Optimum Design Of Screw Centrifugal Pump Based On Rbf Neural Network And Differential Evolution Algorithm

Because of its good non-clogging and high efficiency,the screw centrifugal pump is widely used to transport fluid with large particles and high mixing ratio and low viscosity.However,its special blade structure makes the overall performance of spiral centrifugal pump lower than that of ordinary centrifugal pump in clean water mediu.Therefore,it is of great significance to find a design method that can improve the overall performance of the screw centrifugal pump.On the basis of the relevant research and optimization design academic achievements at home and abroad,in this paper,the screw centrifugal model pump is studied systematically.The method of combining radial basis function?RBF?neural network with MATLAB is proposed.The head and efficiency of the design point are optimized as a new optimal design method of the screw centrifugal pump.The main contents of this paper are as follows:1.According to the two-dimensional drawing of the initial model pump,referring to the design method of the blade and volute of the spiral centrifugal pump,the three-dimensional solid pump hydraulic model was generated by using Pro/Engineer5.0 software.ICEM was used for grid division,and then the grid independence was verified,and a suitable grid division scheme was selected.2.ANSYS-CFX software was used to perform steady calculations on the internal flow field of the model pump and compared with the test data to ensure the reliability of the simulation method.The performance difference of the model pump in different working medium was also analyzed.3.The parameters of the impeller hydraulic structure that affect the objective function were studied.The multi-factor analysis of variance in MATLAB was used to screen the significant test parameters.It is determined that the factors that have significant influence on the design flow point head and efficiency are the inlet angle of impeller hub?_1b,the outlet diameter of impeller D₂,the outlet width of impeller b₂and the blade wrap angle.4.Combined with Fang Kaitai's uniform test table,the training samples for RBF neural network were designed.The mapping relationship between significant influencing factors and performance was obtained.The performance prediction model was constructed.Ten groups of structural parameters were randomly generated by the rand function in MATLAB,and the relative errors between the predicted value and the calculated value of CFX were compared and analyzed.5.The trained artificial neural network prediction model and differential evolution algorithm were used to optimize the head and efficiency.In combination with Pareto solution set,the internal characteristics of the individual with the best head and efficiency were studied.The results show that the performance of the model pump after optimization is better than before.