Optimization Of Planetary Magnetic Gear With Modulation

Since the beginning of this century,many new types of magnetic gears have been proposed after the field-modulated coaxial magnetic gear(FMCMG)was introduced.With the various attempts of permanent magnet arrangements the torque density of magnetic gears has made a qualitative leap compared to the previous century,reaching the range of 100-200 N·m/L.A planetary magnetic gear with modulation(PMGM)uses planetary gears that passively rotate between the inner and outer rotors and actively enhance the modulation by using the rotating magnetomotive force.Compared with the traditional FMCMG that passively modulates the field by using silicon steel modulators,the PMGM significantly enhances the torque transmission capability and has become a new research hotspot.However,in order to find the rotation angles of the planetary gears,the electromagnetic finite element model has to be used together with an unconstrained optimization algorithm,which leads to high computational costs and difficulty for the analysis and design of PMGM.In addition,there has been no comprehensive and systematic optimization of PMGM since its birth.In the analysis of PMGM by using the electromagnetic finite element model together with the unconstrained optimization algorithm,researchers focus on the preconditioning of the optimization,i.e.initializing the angular positions of the planetary gears,in order to accelerate the convergence of the unconstrained optimization that attempts to search for the balanced positions of the planetary gears.The existing method relies on the speculation of the air gap field distribution for the preconditioning.This thesis proposes the use of statistical learning methods such as Gaussian process regression to assist in initializing the angular positions of the planetary gears.The angular positions of the planetary gears are initialized by collecting the balanced positions of the PMGMs with different structural parameters.This method does not require speculation and has been proven to outperform the traditional preconditioning method in practice.In order to design a PMGM with excellent torque performance,on the basis of finding the balanced positions of the planetary gears,the thesis optimizes the key structural parameters of three permanent magnet arrangements,i.e.surface-mounted,flux-focusing spoke type,and Halbach array for the maximum torque per unit volume of permanent magnet and for the maximum torque,respectively,and finalizes the optimal design by calculating its loss and stress.The final design achieves a torque density of 183.4 N·m/L while maintaining a large margin of air gap.The thesis also proposes the use of surrogate model as the future work on the optimization methods for PMGMs.