A Review of Data-Driven Models for Electromagnetic Devices Design and Analysis

In recent years, the design and optimization of electromagnetic devices have grown increasingly complex, driven by the demand for higher efficiency, greater power density, and cost-effectiveness. Traditional approaches such as finite element analysis (FEA) offer precise simulations but can be time-consuming and computationally intensive. To address these challenges, data-driven methods have gained traction as efficient alternatives. This review discusses the application of data-driven models in the design and optimization of electromagnetic devices, summarizes the statistical models such as Response Surface Methodology (RSM), and recent popular machine learning (ML) methods in handling multiple variables, as well as the deep learning (DL) models, in predicting various electromagnetic device parameters and optimizing electromagnetic models. This paper highlights the latest advances in DL models for electromagnetic device applications, including motors, transformers, and electrical wires. It discusses their potential to assist FEA to accelerate design and optimization. Future key directions are proposed to improve efficiency and expand the versatility of data-driven models.