高速永磁电机由于其功率密度大、体积小、可以省去变速机构从而实现直驱等优势，逐渐成为国内外电机领域的研究热点。然而，由于高速永磁电机的转速高、体积小等特点，导致了高速永磁电机的设计需要同时考虑转子应力强度、转子动力学、电磁与损耗、散热及传热等方面的问题，且不同物理场之间还存在非线性、强耦合的关系。因此，如何对高速永磁电机的多物理场综合性能进行分析与优化设计成为了该领域的难点。

本论文围绕高速永磁电机多物理场优化设计展开深入研究，包括：转子长径比对高速永磁电机多物理场综合性能的影响规律；高速永磁电机不同代理模型的拟合精度；以及如何简化高速永磁电机多物理场优化设计流程。

高速永磁电机多物理场优化设计由于多物理场之间的非线性、强耦合关系变得非常困难，而现有文献中针对该问题尚未出现理想的解决方法。由此，本论文提出了一种高速永磁电机多物理场优化设计流程，该流程简化了由于多物理场之间相互耦合导致的高速永磁电机多物理场优化设计困难的问题。通过结合序贯优化策略，实现了多物理场之间耦合关系的简化；同时，通过结合子空间优化策略，实现了高维多目标优化空间的降维优化，并有利于提高代理模型的拟合精度。

转子长径比对高速永磁电机多物理场综合性能影响非常大，由此本论文针对转子长径比对高速永磁电机多物理场综合性能的影响规律进行深入研究。对比、分析并总结了转子长径比对转子应力、转子动力学、电磁与损耗特性以及电机温升等多物理场综合性能的影响规律，为转子结构尺寸参数方程中转子长径比的选择以及高速永磁电机的多物理场优化设计提供了参考。

现有文献中针对高速永磁电机不同代理模型拟合精度的相关研究尚少，而代理模型的拟合精度直接影响优化设计结果，因此需要对不同种类代理模型的拟合精度进行研究。本论文选取代表参数化模型的响应面代理模型、代表半参数化模型的Kriging代理模型以及代表非参数化模型的神经网络代理模型，对这三种代理模型的拟合精度进行对比分析，为后续高速永磁电机的多物理场优化设计提供精确的优化目标模型。

最后，基于本文的高速永磁电机多物理场优化设计流程，设计并制造了一台60kW，30000rpm的高速永磁电机实验样机。实验样机测试数据与有限元分析结果表明，本论文提出的高速永磁电机多物理场优化设计方法能够有效优化高速永磁电机性能，其中包括电机效率、输出转矩、电机热负荷以及永磁体涡流损耗等关键性能。

High-speed permanent magnet motors (HSPMMs) have advantages such as high-power density, small size, and direct drive, it become a research hotspot in the field of motors. However, due to the high speed and small size of HSPMMs, the design of HSPMMs needs to consider the multi-physics performances, such as rotor stress, rotor dynamics, electromagnetic and loss, heat dissipation, and heat transfer, etc. And there are nonlinear and strong coupling relationships between different physical fields. Therefore, how to analyze and optimize the comprehensive multi-physics performance of HSPMMs has become a difficult task in this field.

In this thesis, a multi-physics optimization design method for HSPMMs was proposed, which was based on the multi-physics comprehensive design process of HSPMMs. Including: the influence of rotor Length-Diameter ratio (L/D) on the comprehensive multi-physics performance of HSPMM. The fitting accuracy of different surrogate models of the HSPMM. And how to simplify the optimal design process for multi-physics optimization of HSPMM.

The multi-physics optimal design process of HSPMMs is very difficult due to the nonlinear, strongly coupled relationship, and in the existing literature, many solutions were proposed, but ideal solution to this problem has yet appeared. Thus, this thesis proposes a multi-physics field optimization design process for HSPMMs, which simplifies the problem. By combining sequential optimization strategy, the simplification of the multi-physics coupling relationship is achieved. Meanwhile, by combining subspace optimization strategy, the dimensionality reduction of the high-dimensional multi-objective optimization space is achieved and the fitting accuracy of the surrogate model is facilitated.

The influence of rotor L/D on the multi-physical comprehensive performance of HSPMM was studied. The multi-physics performances were compared and analyzed, which include rotor stress, rotor dynamics, electromagnetic and loss characteristics, and motor temperature rise. And the influence law is summarized, which provides a reference for the selection of rotor L/D in the HSPMMs design process and the optimal design process of HSPMMs.

There are few studies in the literature on the fitting accuracy of different surrogate models for HSPMMs, and the fitting accuracy of the surrogate models directly affects the optimization design results, so the fitting accuracy of different kinds of surrogate models needs to be studied. In this thesis, the response surface surrogate model, Kriging surrogate model, and neural network surrogate model are selected to compare and analyze. And the compared results of these three surrogate models to provide accurate optimization target models for multi-physics optimization design of HSPMM.

Finally, a 60kW, 30,000rpm HSPMM experimental prototype was designed and fabricated based on the multi-physics optimization design process proposed in this thesis. The experimental prototype test data and finite element analysis (FEA) results show that the multi-physics optimization design method of HSPMMs proposed in this thesis can effectively optimize the performance of HSPMM, which includes key performance such as motor efficiency, output torque, motor thermal load and eddy current loss of permanent magnets.