数据驱动下磁性元件的磁芯损耗建模

Magnetic Core Loss Modeling of Magnetic Components under Data-Driven Conditions

摘要: 为实现磁性元件高效率、高功率密度的设计目标，深入探究其损耗特性是关键前提。本文以四种不同磁芯材料为研究对象，系统分析了磁芯损耗的关键影响因素。研究首先将磁通密度数据视为波形信号，提取其时域与频域特征，与原始数据中的温度、频率等变量共同构建特征集，建立XGBoost磁芯损耗回归模型。接下来基于经典的磁芯损耗模型斯坦麦茨方程，增加温度因素，并将其以幂指数形式引入，显著提高了磁芯损耗预测效果。最后通过方差分析探讨温度、励磁波形以及磁芯材料对磁芯损耗的独立影响和两两交互作用，并根据Tukey HSD事后检验结果，分析不同温度、励磁波形和材料组别之间的具体差异，找出影响损耗的主要驱动因素。

Abstract: To achieve the design goals of high efficiency and high power density in magnetic components, a thorough investigation of their loss characteristics is a crucial prerequisite. Taking four different core materials as the research objects, this paper systematically analyzes the key factors affecting core loss. First, the magnetic flux density data are treated as waveform signals, and their time-domain and frequency-domain features are extracted. These features are then combined with variables such as temperature and frequency from the original data to construct a feature set, based on which an XGBoost regression model for core loss is established. Next, based on the classical core loss model, namely the Steinmetz equation, a temperature factor is incorporated in the form of a power exponent, which significantly improves the prediction performance of core loss. Finally, analysis of variance is employed to investigate the independent effects of temperature, excitation waveform, and core material on core loss, as well as their pairwise interaction effects. Based on the results of the Tukey HSD post hoc test, the specific differences among different temperature levels, excitation waveforms, and material groups are further analyzed to identify the major driving factors affecting core loss.