Towards a quantitative comparison between experimental and synthetic grain structures

A method is proposed to quantitatively compare 3-D microstructural data using the Hellinger distance. Statistics from an experimentally observed 3-D IN100 microstructure are used to generate six synthetic microstructures. These are compared quantitatively to the experimental microstructure by computing the Hellinger distances between distributions of microstructural descriptors, such as the volume, the aspect ratio, and the affine moment invariant (Omega) over bar (3). The synthetic microstructures use three starting shape classes, namely ellipsoids, superellipsoids, and shapes generated by truncating a cube with an octahedron; this latter shape class is proposed here for use in microstructure generation, and we derive relevant morphological descriptors. Synthetic microstructures are generated using either random grain placement or by constraining the grain placement to the experimental grain centroids. We present a method for determining microstructure similarity by random sampling from a reference microstructure as well as quantitative shape comparisons between synthetic and experimental microstructures. (C) 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.