Multi-objective Optimization Of Low Specific Speed Centrifugal Pump Based On Intelligent Optimization Algorithm

Because of low specific speed centrifugal pump has a wide flow head and high working condition,which makes it widely used in petroleum,chemical and other fields.However,the low specific speed centrifugal pump has low efficiency,narrowing high efficiency range and which operation is extremely unstable under cavitation conditions.Therefore,how to design an high efficiency and opreation stable low specific speed centrifugal pump has been a hot research topic at Chinese and foreign.In this study,a low specific speed IS 80-65-310centrifugal pump with a specific speed of 30 is selected as the research object.The pump efficiency and NPSH_r are selceted as the optimization tragets.Combined with CFD numerical simulation andmulti-objective optimization based on RBF neural network and NSGA-II genetic algorithm to optimize this model pump.The main contents are as follows:1.In this paper,the computational domain is meshed by means of structural grid.The SST k-w turbulence model based on modified turbulent viscosity is numerically simulated.The reliability of the numerical simulation is verified by the comparison of the centrifugal pump experiment.The results show that the external curve of the model pump obtained by the simulation is basically the same as the curve obtained by the experiment,and error which is between experiment and simulation value is less than 5%.The comparison of model pump cavitation characteristic between simulation and experiment has indicated that experimental cavitation inception is smaller than the simulation,but the overall trend is basically the same.2.By analyzing the theoretical model of centrifugal pump efficiency and NPSH_r,we can get the mathematical expressions of each optimized target and the weighting coefficient values of each target is determined by the super-transfer approximation method.The Plackeet-Burman test screening method is used to select the four impeller geometric parameters that have the greatest impact on each optimization target.Then we chose the blade inlet angle?₁,the blade exit angle?₂,the blade exit width b₂,and the blade wrap angle?as the optimization variables,given the constraints of each impeller geometric parameters.Finally a multi-objective optimization form lays the theoretical foundation for later optimization.3.The optimal Latin subcubic sampling was used to design 25 cases of four optimized variables.The numerical simulation was used to obtain the optimal target value of each group as the initial training sample database of RBF neural network.Then we make use of RBF neural network to construct input variables and optimization approximation model.Finally using the NSGA-II genetic algorithm to perform extreme value optimization on the trained RBF neural network.After 500generations of iterative inheritance,the optimal target Pareto solution set frontier distribution is obtained and selected to meet the design requirements individuality as the final optimization.The numerical simulation of the optimization scheme shows that the optimized pump efficiency under the design condition is 2.63%higher than that of the original pump,and theNPSH_r value is decreased by 0.18m,which indicates that the optimized pump is more powerfrugal and steady than original pump.4.In order to better explain the optimization of the overall flow performance of the optimized pump is better than the original pump,by comparing the internal flow field characteristics and external characteristic between original case and optimized case.Firstly,we comparing the external characteristic curves,cavitation performance curves,matching characteristics of the volutes and the distribution characteristics of the respective blade loads between optimized pump and the original pump.The results show that the optimized pump is excellent in both efficiency and pump cavitation characteristic than the original pump.In addition,comparing with original pump the matching characteristic of optimized pump and the volute is more reasonable,the blade load distribution is more uniform,and the blade function is enhanced.Then the characteristics of the internal flow field distribution under non-cavitation and cavitation conditions are compared.The results show that under non-cavitation condition,compared with the original pump,the pressure field,velocity field and turbulent kinetic energy field in the optimized pump are improved,the flow in the pump is more uniform,the separation vortices are reduced,and the hydraulic loss is also reduced.Under cavitation condition,the anti-cavitation performance of the optimized pump is improved compared with that of the original pump,and the flow field structure in the pump is more uniform and reasonable.When cavitation is serious,the optimized pump can operate normally but original pump cannot operation even appear cut-off operation phenomenon.Finally,the transient characteristics of original pump and optimization pump under unsteady conditions are analyzed.Firstly,the efficiency fluctuation of the two pumps is compared.The results show that the internal flow of the optimized pump is more stable than that of the original pump in macroscopic performance.Secondly,the unsteady flow of the two pumps is compared by setting monitoring points in the pump.The results show that the pressure of the optimized pump is optimized at the front half of the impeller.The main frequency amplitude of the optimized pump is smaller than that of the original pump.In the second half of the flow passage of the impeller,the main frequency amplitude of the pressure fluctuation of the optimized pump is slightly higher than that of the original pump.In general,the flow stability of the optimized pump is better and the flow in the pump is more uniform and reasonable.