2019/06/13 by Elizabeth J. Kautz, Wufei Ma, Kautz, Elizabeth +11
Engineering · Materials Science · #Advanced Materials Characterization Techniques #Applied Physics (physics.app-ph) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG) #Machine Learning in Materials Science #Nuclear Materials and Properties
paper · pdf · doi:10.48550/arxiv.1906.05496
openalex publication_date 2019/06/13 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Micrograph quantification is an essential component of several materials\nscience studies. Machine learning methods, in particular convolutional neural\nnetworks, have previously demonstrated performance in image recognition tasks\nacross several disciplines (e.g. materials science, medical imaging, facial\nrecognition). Here, we apply these well-established methods to develop an\napproach to microstructure quantification for kinetic modeling of a\ndiscontinuous precipitation reaction in a case study on the uranium-molybdenum\nsystem. Prediction of material processing history based on image data\n(classification), calculation of area fraction of phases present in the\nmicrographs (segmentation), and kinetic modeling from segmentation results were\nperformed. Results indicate that convolutional neural networks represent\nmicrostructure image data well, and segmentation using the k-means clustering\nalgorithm yields results that agree well with manually annotated images.\nClassification accuracies of original and segmented images are both 94 % for a\n5-class classification problem. Kinetic modeling results agree well with\npreviously reported data using manual thresholding. The image quantification\nand kinetic modeling approach developed and presented here aims to reduce\nresearcher bias introduced into the characterization process, and allows for\nleveraging information in limited image data sets.\n