| About this Abstract |
| Meeting |
2021 TMS Annual Meeting & Exhibition
|
| Symposium
|
Fatigue in Materials: Fundamentals, Multiscale Characterizations and Computational Modeling
|
| Presentation Title |
Discovering the Structural Signature of Fatigue Crack Growth Rate Using Computer Vision and Machine Learning |
| Author(s) |
Katelyn Jones, William Musinski, Adam Pilchak, Reji John, Paul Shade, Anthony Rollett, Elizabeth Holm |
| On-Site Speaker (Planned) |
Katelyn Jones |
| Abstract Scope |
Machine Learning in materials science permits a deeper understanding of the relationship between microstructure and mechanical properties through efficient analysis of large amounts of data. Convolutional neural networks (CNNs) have been used to connect images of microstructure with processing history and properties such as fatigue life. This project uses CNNs on experimental images of fracture surfaces that have been augmented and/or segmented to predict features in crack growth such as direction, length, and rate. This study focuses on Ti-6Al-4V because of its wide usage in aerospace and medicine, abundance of data, favorable mechanical properties, and corrosion resistance. The resulting model aims to predict crack growth behavior. The application of CNNs in this instance, images used, and identified causes of crack rate transition will be presented. |
| Proceedings Inclusion? |
Planned: |