MACHINE LEARNING CLASSIFICATION OF HYDROGEN-INDUCED FRACTURE MECHANISMS USING SEM IMAGES

The large-scale development of hydrogen infrastructure is essential for the transition to low-carbon energy systems, yet it remains constrained by material integrity challenges, such as hydrogen-induced fracture. This study explores the application of Machine Learning techniques to classify brittle and ductile mechanisms in Scanning Electron Microscopy (SEM) micrographs, aiming to provide a computationally efficient alternative for fractographic analysis. A comprehensive methodology was created, incorporating a hybrid feature extraction pipeline based on Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP) descriptors and the evaluation of multiple supervised models. Among these, the Support Vector Machine (SVM) achieved an accuracy of 99.45%, comparable to that reported for ResNet-50, with a substantially lower training time. These findings demonstrate the viability of hand-extracted feature-based models in providing accurate and scalable diagnostic tools, contributing to risk assessment in hydrogen storage and transport. Ultimately, the proposed approach supports the safe and cost-effective monitoring of materials in transitioning to sustainable energy systems.