Toward quantitative fractography using convolutional neural networks

• Published in: Engineering fracture mechanics 2020-05, Vol.231, p.106992, Article 106992
• Main Authors: Tsopanidis, Stylianos, Moreno, Raúl Herrero, Osovski, Shmuel
• Format: Article
• Language: English
• Subjects: Algorithms; Aluminum oxide; Artificial neural networks; Dimpling; Fractography; Fracture surfaces; Image analysis; Image segmentation; Intergranular fracture; Machine learning; Morphology; Neural networks; Qualitative analysis; Quantitative analysis; Reliability analysis; Striations; Transgranular fracture
• ISSN: 0013-7944; 1873-7315
• DOI: 10.1016/j.engfracmech.2020.106992

•Typological segmentation of brittle fracture surfaces using pretrained CNN.•Area fraction of intergranular and transgranular fracture is measured quantitatively.•The network, trained on MgAl2O4 can be used for other ceramics without further training. The science of fractography revolves around the correlation between topographic characteristics of the fracture surface and the mechanisms and external conditions leading to their creation. While being a topic of investigation for centuries, it has remained mostly qualitative to date. A quantitative analysis of fracture surfaces is of prime interest for both the scientific community and the industrial sector, bearing the potential for improved understanding on the mechanisms controlling the fracture process and at the same time assessing the reliability of computational models currently being used for material design. With new advances in the field of image analysis, and specifically with machine learning tools becoming more accessible and reliable, it is now feasible to automate the process of extracting meaningful information from fracture surface images. Here, we propose a method of identifying and quantifying the relative appearance of intergranular and transgranular fracture events from scanning electron microscope images. The newly proposed method is based on a convolutional neural network algorithm for semantic segmentation. The proposed method is extensively tested and evaluated against two ceramic material systems (Al2O3,MgAl2O4) and shows high prediction accuracy, despite being trained on only one material system (MgAl2O4). While here attention is focused on brittle fracture characteristics, the method can be easily extended to account for other fracture morphologies, such as dimples, fatigue striations, etc.