Intelligent Optimization Design Platform Of Centrifugal Pump Impeller Based On Machine Learning

Centrifugal pump as one of the most widely used fluid transmission device,according to the statistics of relevant departments,the power consumption of centrifugal pump is very huge,accounting for about 10% of China’s annual power consumption,so improving the efficiency of centrifugal pump is conducive to achieve the national energy conservation and emission reduction goals.Centrifugal pump impeller is the core of the centrifugal pump flow components,impeller design affects the centrifugal pump internal flow condition also determines the efficiency of the centrifugal pump,for the centrifugal pump impeller design method and optimization method of the research has never been interrupted.Under the continuous efforts of scholars,a variety of new design methods and optimization methods began to be applied to manufacturing and production,but in terms of the development cycle of new products,to design an efficient centrifugal pump impeller still needs a lot of time and cost.The iterative process of design,simulation,experiment,and then back to design is still the most mainstream design idea.In recent years,with the continuous development of machine learning theory,how to shorten the development time of new products with new technologies has attracted the attention of researchers.This paper discusses the possibility of transforming qualitative empirical design into scientific quantitative design in the iterative design of centrifugal pump impeller by combining artificial intelligence technology and genetic algorithm.Finally,the intelligent design optimization platform of centrifugal pump impeller based on machine learning is developed and verified by simulation calculation and experiment.The main research contents of this study include:(1)Summarized the research status of centrifugal pump design at home and abroad,including hydraulic design methods and related flow channel optimization,blade optimization,etc.At the same time,the design process of centrifugal pump impeller depends on the experience of designers,and the whole iterative optimization process of design,modeling,simulation and modification is high in labor cost and time cost.Therefore,based on the development of artificial intelligence-related technologies in recent years,this paper analyzes the possibility of transforming empirical design which is difficult to be quantified into intelligent design which can be quantified.(2)A neural network for the design of the main size of centrifugal pump impeller is established and used to predict the main size.The change curve of the flow section area of the impeller channel is used as the basis of the flow channel optimization and an evaluation algorithm is written according to this theory.The algorithm is used as the fitness function of genetic algorithm to determine the flow channel optimization method.The effectiveness of the optimization method is verified by the flow analysis of a group of optimization examples.At the same time,a neural network of inlet and outlet angle of centrifugal pump impeller blade was established and used to predict each angle of the blade,and the blade drawing was completed by point by point calculation and recursive binary search method combined with Bezier curve library.(3)In order to break through the limitations of neural network database and reduce labor costs,this paper combined CFturbo,Pumplinx and Isight software to build an intelligent simulation platform,and carried out numerical calculation on the pump model designed by neural network prediction parameters,and optimized the pump model by simulated annealing algorithm.A group of optimization examples were used to illustrate the co-simulation automatic optimization method,and the optimized impeller model had good internal flow state and high hydraulic efficiency.(4)Latin hypercube sampling is used to sample the specific speed range allowed by the intelligent design method.Choose one of the samples with CFturbo design by default,the internal flow field in comparison with traditional experience design result external characteristic comparison,results show that the intelligent optimization design method can indeed realize from dependence on the designers design experience,to automatic prediction parameters of intelligent design,compared with the traditional method has the advantages of higher design efficiency,The hydraulic efficiency and internal flow of the designed model are also guaranteed.Finally,one group of samples was processed into a model,and a test-bed was built to test.The accuracy of numerical calculation was verified by comparing the test results with the simulation results.(5)Based on Python language and Qt framework,the intelligent design optimization platform of centrifugal pump impeller was built,and the basic architecture and main development methods and functions of the intelligent design platform were described.Finally,the design process of centrifugal pump impeller is simplified,the design efficiency of centrifugal pump impeller is improved,and the traditional method is separated from the experience of designers.