Search for: All records

We present here a survey of important machine learning (ML) methodologies and associated applications that have led to recent major advances in the acquisition and interpretation of materials microstructures. These advances have enabled, for example, the automated processing of micrographs to identify defects, such as grain and interphase boundaries, and the subsequent classification of structures and compilation of grain statistics for use in quantitative microstructural characterization. To summarize the role of ML in this domain, we first outline different approaches to image segmentation and feature extraction and then describe various microscopies and associated distinctive issues relating to image partitioning and image object identification. For these microscopies, we review illustrative studies in which ML has played a crucial role in parsing the resulting micrographs. Next, we highlight the contributions of ML in microstructural interrogation and modeling by considering various strategies for image interpretation that exploit, for example, neural networks (NN) in tasks such as classification and clustering. As NNs have also furthered the goals of microstructural optimization and inverse design, we also examine how these tools have been employed to create materials having desirable properties. Finally, we illustrate the use of correlative statistical methods to assess and predict the occurrence of microstructural anomalies (e.g., abnormal grain growth) and report on progress in the creation and curation of microstructural repositories that facilitate data sharing. The aim of this Overview is to make the case that ML tools have become, in a relatively short time, indispensible aids for microstructural image processing and interpretation. 

Metallic thin films offer a platform to experimentally study the dynamics of microstructural evolution, but the required transmission electron microscopy (TEM)-based imaging generates complex images that are challenging to segment and quantify. This work provides a comparative analysis of a new YOLOv8 model and an established U-Net model for bright-field TEM images of polycrystals, employing a framework leveraging physical observables to evaluate performance against two hand-traced benchmark datasets. This methodology obviates the comparison of large, diversely structured, and manually labeled datasets that are required to assess performance on a per-image/per-pixel basis. It is found that the YOLOv8 model, adapted for real-time instance segmentation, has up to 43× faster inferencing (NVIDIA GeForce RTX 4090) compared to U-Net and reconstructs hand-traced grain size distributions (GSDs) with excellent fidelity, finding mean diameter within 3% for grains near an optimal magnification; for grains that deviate from the optimal pixel-diameter, the size of small- (large)-diameter grains is systematically over- (under)-estimated. This is partially mitigated by including scale-aware augmentations during training. Moreover, when the bias is corrected post-inference by a rigid shift in distribution, the YOLOv8 model reproduces ground truth GSDs with exceptional fidelity, with statistical tests indicating <5% probability that the distributions are distinct. Based on ground truth data, calibration curves pertaining to this shift can be constructed for a given model. This issue is not present in the U-Net model’s results, indicating that for quantitative measurements where the true size of objects is of interest, special procedures must be implemented for YOLO-based models. 

Abstract We develop a thin-film microstructural model that represents structural markers (i.e., triple junctions in the two-dimensional projections of the structure of films with columnar grains) in terms of a stochastic, marked point process and the microstructure itself in terms of a grain-boundary network. The advantage of this representation is that it is conveniently applicable to the characterization of microstructures obtained from crystal orientation mapping, leading to a picture of an ensemble of interacting triple junctions, while providing results that inform grain-growth models with experimental data. More specifically, calculated quantities such as pair, partial pair and mark correlation functions, along with the microstructural mutual information (entropy), highlight effective triple junction interactions that dictate microstructural evolution. To validate this approach, we characterize microstructures from Al thin films via crystal orientation mapping and formulate an approach, akin to classical density functional theory, to describe grain growth that embodies triple-junction interactions.