The FeNdB-type permanent magnets form a crucial part of the components in electric vehicles (EV) and other energy conversion machines primarily due to their high energy density. The presence of rare earth (RE) metals to improve properties, such as higher Curie temperature, directly impacts costs. Therefore, the challenge lies in the search for novel magnetic phases with less RE content or inexpensive RE metals that require efficient quantitative microstructure analysis to extract information from magnetic phases and correlate it with their intrinsic magnetic parameters.

This presentation provides an end-to-end workflow for the machine learning-based tool for grain size analysis in magnet samples. The usability of the developed tool for real-world applications, such as in research and quality control labs, has been realized by incorporating the concepts of data drift and model interoperability.

Using the machine learning approaches, we have developed an automated framework for effectively characterizing the FeNdB-type permanent magnets using microstructural information from Kerr microscopy. The models developed are evaluated against the reference data from subject experts and the state-of-the-art EBSD approach to measuring the performance and degree of effectiveness. The model has proved to be efficient, robust to different magnet samples, and a time-efficient method for grain and texture analysis in magnets with accuracy in the 91-93% range for various magnet alloys and is nearly 20 times faster than existing approaches.