Application of the Jiles—Atherton Hysteresis Model for Crack Defect Detection in Power Electronic Transformers of Iron Core

The advancement of power electronics has elevated the importance of transformer cores within high-frequency converters, rendering them both crucial and susceptible components. Conventional methods for detecting defects are often either invasive or lack the necessary sensitivity to internal degradation. In this research, a novel hybrid diagnostic framework is proposed, integrating the inverse Jiles—Atherton (J-A) hysteresis model with Particle Swarm Optimization (PSO) for the extraction of magnetic parameters. The assessment of faulty cores is based on coercivity, remanence, and B-H loop distortion. Through a comprehensive statistical analysis spanning frequencies of $200 ~\text{Hz}, 600 ~\text{Hz}$, and 1000 Hz, it is determined that 600 Hz exhibits the highest sensitivity to defects. Subsequently, the corresponding B-H data is utilized to train a CNN-BiLSTM-Attention network for automated classification. This amalgamated methodology enhances detection accuracy and offers a scalable approach for intelligent condition monitoring of power transformer cores.