Predicting Grain Size of Fe Si Alloys Steel Sheet through Magnetic Properties and Machine Learning with XAI

Measuring the grain size of electrical steel is important for understanding its properties, but until now studies have attempted this using experimental equations or machine learning (ML), resulting in limited models. This study applied three ML and Explainable-AI (XAI) to analyze magnetic properties and identify key the factors most relevant to grain size prediction. Among the three tree-based models, B4_Br/Bm, B10_P_cm were identified as key variables. It was also found that variables under B1, B4, B10 conditions accounted for 60% of the model explanation ratio. Final model was constructed using these variables. These findings demonstrate the potential of ML with XAI to predict grain size with fewer magnetic measurement.