Hydraulic Optimization Of High-flow Centrifugal Pump Based On BP Neural Network And NSGA-Ⅱ Algorithm

Centrifugal pump as a simple structure of fluid transmission equipment has been widely used in nuclear power,chemical industry and aerospace and other fields,China’s independent research and development of small and medium-sized centrifugal pumps in the domestic accounted for a relatively large,and large-flow centrifugal pump domestic bias to import,resulting in less research on large-flow centrifugal pump in China.There are structural differences between large flow multistage centrifugal pump and small and medium-sized multistage centrifugal pump,and this structural difference has a great impact on the flow pattern in the pump,so the traditional optimization method of centrifugal pump is not suitable for large flow multistage centrifugal pump.Large flow multistage centrifugal pump needs to design large flow area in order to reduce the impact of water flow and other losses,and needs to use three-dimensional twisted blades.In order to optimize the head,head curve slope,efficiency and NPSHr of the first stage impeller of multistage centrifugal pump,this paper selects various parameters of the three-dimensional twisted blades of the first stage impeller as the optimization parameters.The head and efficiency and slope of head curves are taken as the optimization objectives,and BP neural network and improved NSGA-Ⅱ algorithm are used to optimize them.Then the cavitation calculation is carried out on the optimized impeller and the difference between the cavitation law of multistage centrifugal pump and volute centrifugal pump is analyzed.According to the cavitation law of multistage centrifugal pump,the cavitation optimization method applicable to multistage centrifugal pump is proposed and verified by calculation.The main research contents of this paper are as follows:1.NX is used to model the inlet section,suction chamber,first stage impeller,interstage flow channel and outlet section of multistage centrifugal pump,and Fluent meshing is used to divide the calculation domain into polyhedral meshes,and then the independence of the mesh is verified and the minimum number of meshes is determined as the number of meshes used for numerical calculation under the condition of meeting the calculation accuracy.2.Carry out numerical calculation of the pump through Fluent software,select the standard k-ε turbulence model and use the real inlet pressure as the inlet boundary condition to calculate the first stage head and efficiency of the multistage centrifugal pump.Compare the head and efficiency obtained by numerical calculation with the test,and find that the performance obtained by numerical calculation is higher than that of the test because the clearance and leakage of the centrifugal pump are not considered in the numerical calculation,However,it is within the allowable error range of the project,and the numerical calculation can accurately predict the performance change trend of the pump,so it is concluded that the numerical calculation has certain accuracy and can be used as a tool for impeller preliminary optimization.3.Optimize the efficiency and the head and slope of head curves of the first stage impeller,select the impeller outlet width,inlet hub placement angle,inlet shroud placement angle,outlet placement angle and wrap angle as the optimization parameters of the impeller,take the efficiency and the head and slope of head curves as the optimization objectives,and select the Latin hypercube experimental design method to design 40 groups of experiments.BP neural network is written in Python,and 40 groups of experimental data are fitted.The fitted BP neural network is used as the fitness function of the NSGA-Ⅱ algorithm,and then the NSGA-Ⅱ algorithm is improved by adding the command to remove the individuals whose slope of the range curve does not meet the use requirements.Set the number of population as 50,the number of iterations as 2000,the crossover rate as0.9,and the variation rate as 0.1.After optimization,the head of the optimized pump under standard working conditions is increased by 2m,and the efficiency is increased by 1.7%.4.Calculate the NPSHr of the optimized pump and analyze the cavitation distribution.It is found that the cavitation will occur at the baffle of the suction chamber of the multistage centrifugal pump due to the difference between the suction chamber structure and the volute pump.If the traditional method of optimizing the cavitation performance(blade forward extension)is used,the cavitation at the baffle will enter the blade with the liquid flow and aggravate the cavitation on the blade.Analyze the flow field and then choose the opposite way(moving the blade inlet backward)to keep the blade inlet away from the unstable fluid at the water baffle.At the same time,moving the blade backward can increase the flow area between the inlet blades,so as to reduce the flow rate and increase the hydrostatic pressure.Under the condition of large flow,the NPSHr of the optimized pump is large,and the blade inlet placement angle is increased,so that the cavitation area in the impeller is concentrated near the back of the blade inlet.After adjusting the blade inlet,the NPSHr of the modified pump meets the design requirements of the pump.