Grain size analysis in permanent magnets from Kerr microscopy images using machine learning techniques

Understanding the relationships between composition, structure, processing and properties helps in the development of improved materials for known applications as well as for new applications. Materials scientists, chemists and physicists have researched these relationships for many years, until the recent past, by characterizing the bulk properties of functional materials and describing them with theoretical models. Magnets are widly used in electric vehicles (EV), hybrid electric vehicles (HEV), data storage, power generation and transmission, sensors etc. The search for novel magnetic phases requires an efficient quantitative microstructure analysis of microstructural information like phases, grain distribution and micromagnetic structural information like domain patterns, and correlating the information with intrinsic magnetic parameters of magnet samples. The information out of micromagnetic domains helps in obtaining the optimized microstructures in magnets that have good intrinsic magnetic properties. This paper is aimed at introducing the use of a traditional machine learning (ML) model with a higher dimensional feature set and a deep learning (DL) model to classify various regions in sintered NdFeB magnets based on Kerr-microscopy images. The obtained results are compared against reference data, which is generated manually by subject experts. Additionally, the results were compared against the approach for grain analysis, which is based on the electron backscatter diffraction (EBSD) technique. Further, the challenges faced by the traditional machine learning model for classifying microstructures in Kerr micrographs are discussed. •Machine learning techniques for quantitative microstructure analysis from Kerr microscopy images•Automated grain size analysis from light microscopy images of large sample and accuracy close to the EBSD measurements•Proposed DL based grain analysis model has advantages in terms of time-efficiency and robustness on large size samples•Approach can be extended to other applications such as austenite, copper etc.