Understanding the Coercivity of Ga-Containing Nd-Fe-B Sintered Magnets from Feature Extraction and Selection of X-Ray Diffraction Patterns via Dimension Reduction and Sparse Modeling

Coercivity enhancement of Ga-containing Nd-Fe-B as-sintered magnets driven by the microstructural changes occurring during post-sinter annealing was studied via feature extraction and selection based on several data science methods. These data science methods reveal the correlation between the coercivity and the intensity and shape change of the specific structures in the X-ray diffraction pattern. Principal component analysis, a linear model with least absolute shrinkage and selection operator regression, and nonlinear random forest regression modeling of coercivity based on X-ray diffraction patterns objectively identified the microstructural changes correlating with coercivity enhancement. Further, quantitative evaluation of the correlation magnitude of microstructural changes to the coercivity of the magnet revealed a preferred direction of the correlation magnitude even in an identical phase. The validity of our findings was qualitatively confirmed based on the results of previous studies, indicating that sparse modeling accelerates materials research and development via automated decision-making.