Performance Analysis And Optimization Design Of High-Speed Permanent Magnet Machine For Electric Turbocharger

Compared with conventional-speed electrical machines(EMs),high-speed EMs have the advantages of higher power density,smaller volume,lighter weight and so on.It can be directly connected with high-speed load.It is an important direction of future motor development and has broad application prospects in major equipment and intelligent manufacturing.At the same time,the analysis,design and control of high-speed EMs are also more complicated than those with conventional speed,from the basic electromagnetic design to electricity,magnetism,heat,force and other physical field coupling design,promoting the motor design towards multi-field,systematization and automation development.Under this background,the optimization method of high-speed EMs is deeply studied in this paper.Firstly,the operation and design characteristics of high-speed EMs are analyzed.Based on the Python optimization design tool,the multi-physics automatic simulation and analysis platform for high-speed EMs is improved.Then,the surrogate model-based machine optimization design method is studied in view of the serious computation and time consumption defect of large-scale optimization.Finally,aiming at the problem that deterministic design cannot guarantee the quality of batch products,the reliability-based robust optimization method of high-speed EMs is studied in this thesis.The main research contents and achievements of this thesis are as follows:1.The coupling relationship between multiple physical fields of high-speed motor is sorted out,the interaction between electromagnetic,temperature rise and stress is analyzed from the aspects of magneto-thermal and thermal-stress,the calculation method of machine performance in large-scale optimization is summarized,and the automatic optimization design platform is improved.2.Taking a 3.2k W,60000r/min surface-mounted high-speed permanent magnet machine as the research object,a multi-objective optimization framework for magnetic-thermal bidirectional coupling and thermal-stress unidirectional coupling is designed.Through the correlation analysis of machine parameters,the dimension of optimization variables is reduced,and through the correlation analysis of machine performance,cost minimization and torque maximization are taken as optimization objectives.Then,based on the automatic optimization design platform,the multi-objective optimization design of large-scale high-speed machine is completed.3.The application of multilayer perceptron,support vector regression,generalized regression neural network and Kriging model in machine optimization is studied.A comprehensive model evaluation method considering both prediction accuracy and time consumption is proposed,a detailed surrogate-based high-speed machine optimization process is refined,and the optimization results of different surrogate models are compared.4.Aiming at the problem that the optimziaiton results usually need further verficication by finite element method,the information in the initial data set is made full used of in this paper,and besides the regression model,the classifier is innovatively introduced to identify and filter out invalid designs in the optimization results,improving the accuracy of large-scale surrogatebased machine optimization,reducing the workload of the secondary verfification.5.In order to accurately calculate the mean and variance of robust optimization objectives,as well as the probalitity of satisfactory of reliability constraints,the polynomial chaos Chebyshev interval method and the weighted index Monte Carlo method are combined.The basic principles and implementation methods of these two methods are analyzed in detail,and aiming at the disadvantage of insufficient accuracy of surrogate models in the uncertainty interval,a local surrogate model strategy is proposed,which significantly improves the accuracy of machine robust optimization.