Review of Machine Learning-Based Electromagnetic Design Optimization for Electrical Machines

The electrification of transportation is seeing rapid expansion. As an indispensable component, the electrical machine (EM) plays a pivotal role in this context. To enhance the optimization design of EMs efficiently and precisely, techniques based on artificial intelligence have emerged as a new research trend. This paper overviews the application of machine learning (ML) in the electromagnetic optimization domain for EMs. The involved approaches are categorized according to the input object of the ML model, the employed ML paradigm, and the type of metamodel. Furthermore, the principles of the different metamodels and methodologies, together with their respective advantages and limitations, are articulated in a detailed and comprehensive manner. In the end, the forthcoming research challenges and potential pathways for improvement in this field are discussed.