An Optimal Design Research Of PMSM Magnet Steel Based On Neural Network

With the continuous exclamations of rare earth ore,many high performance Permanent Magnet Synchronous Machines(PMSMs)have been invented,and they have accelerated the step on the road of modern industrial development in China.PMSM has been widely applied in many circumstance such as servo system and robot field because of its high efficiency and superior control performance.With the improvement of the performance requirements of PMSM,the torque ripple problem of permanent magnet motor,which is becoming more and more prominent,has become a hot research topic in current motor development and design,and it is urgently needed to be solved.Based on the analysis of the cogging torque of PMSM,the influence of the shape of the magnetic steel and a special radial magnetization method on the cogging torque of surfacemounted PMSM is studied in this thesis.A method of Fourier decomposition analysis is shown for analyzing the shape of the motor magnetic steel in this thesis,and then the torque ripple of the motor is reduced by changing the coefficients of the Fourier series of the magnetic steel and adjusting the magnetization parameters.Finally,a neural network is designed to optimize the parameters above,for optimizing the motor magnetic steel on PMSM,weakening the torque ripple of the motor and improving the speed performance of the motor.In this thesis,the analytical method is used to decompose the shape of the magnetic steel of the motor,and the expression of the cogging torque corresponding to the magnetic steel is deduced.And using the neural network model built by MATLAB,the probability of the coefficients of the magnetic steel is designed to increase monotonically according to Boltzmann distribution while the torque ripple is decreasing,and then the optimal size parameters of the magnetic steel are optimized by sampling.The design method is applied to the optimization design of magnetic steel for an 115kw/3000 rpm PMSM,and the magnetic steel's Fourier decomposition coefficient and the magnetizing parameter that is corresponding to a low torque ripple are obtained.Through the finite element simulation,the torque ripple is effectively reduced.The motor performance of optimizing the motor magnetic steel is compared with the performances of the eccentricity design and the skewed slot design.Through theoretical calculation and simulation analysis,the thesis has found factors of torque ripple by analyzing the shape of the motor magnetic steel which is decomposed into Fourier series.And this thesis propose a method of optimizing the size and magnetization parameters of magnetic steel by using the neural network model and Boltzmann distribution.This method can effectively reduce the torque ripple of the motor.It opens up a new idea for the development of the new shape magnetic steel for large power AC machines in order to improve their performances.