| Double channel pumps are widely used for transporting liquids containing complex components due to their characteristics of wide overflowing passage and little possibility of blockage.However,because its hydraulic structure is extremely simple and the number of blades is extremely few,the overall performance of the double channel pump is often lower than the multi-blade pump of the same specific speed.Therefore,it is of great significance to find a method to optimize the design of double channel pumps.In this paper,a double channel pump with a model of 80QW50-15-4 is used as the research object,taking the head and efficiency of design flow points as the optimization objective,combining radial basis function(RBF)neural network and multi-objective particle swarm optimization algorithm,providing a new method for multi-objective optimization design of double channel pump.The major research works of this dissertation are as follows:1.The previous research results of the double channel pumps' internal flow principle,experimental simulation and optimization methods are summarized,and the feasibility of applying the intelligent optimization method to the optimization design of double channel pumps is analyzed.In addition,according to the actual engineering application,it is determined that the working medium of the dual channel sewage pump are water solution,solid-liquid two-phase fluid and Bingham fluid,and the water-like solution is the main working medium.2.The basic formula of computational fluid mechanics is extracted to explain the theoretical basis of the numerical simulation.According to the two-dimensional hydraulic drawings of the initial model pump,Pro/Engineer5.0 is used to shape the water body of the impeller and the volute.The grid is divided by ICEM,and a reasonable grid partition scheme is selected through grid independence verification.3.The calculation formula of the main performance parameters of the pump is described.Experiment and simulation performance data of model pump in clean water are analyzed and compared to ensure the reliability of simulated data.Apart from that,the difference between performance and internal flow field of the model pump when transporting different working media are also analyzed.4.Part of the initial model impeller's structural parameters are selected to carry out influence significance analysis of performance.The blade wrap angle,exit angle and the outlet width of the impeller are selected to be factors which affects head and efficiency significantly at design flow point through Plackett-Burman test.RBF neural network's training samples of significant factors are arranged according to Fang Kaitai's uniform design table.A performance prediction model between significant structural parameters and performance is established through RBF neural network.In order to detect the reliability of the performance prediction model,five sets of random structural parameters are used for test and error analysis.5.The trained radial basis neural network performance prediction model is used as fitness evaluation model of multi-objective particle swarm optimization,and the Pareto solution set and corresponding structural parameters of head and efficiency are obtained.The optimal head individual and optimal efficiency individual's performance and internal flow field are researched,and the results of numerical simulation show that the performance of the optimized individuals is better than the initial model regardless of media.In addition,through the clear water test,it is verified that the head of the optimal head individual at the design flow point increases by 0.96 m compared with the initial individual,and the increase is 5.5%.The efficiency of the optimal individual is 10.11 percentage points higher than that of the initial individual.It shows that the result is remarkable. |
PDF Full Text Request