About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Environmentally Assisted Cracking: Theory and Practice
|
| Presentation Title |
Prediction of 5xxx Aluminum Alloy Sensitization by Computer Vision |
| Author(s) |
William Golumbfskie, Nicholas Jones, Annika Jackson, Andrew Falkowski, Taylor Sparks |
| On-Site Speaker (Planned) |
William Golumbfskie |
| Abstract Scope |
Marine Grade (5xxx) aluminum alloys are commonly used in marine applications based on their combination of general corrosion resistance and high as-welded strength. In a service environment, these alloys are prone to sensitization which can lead to stress corrosion cracking (SCC). Sensitization occurs as magnesium precipitates out of the matrix forming a deleterious beta-phase (Mg2Al3). To minimize the impact of sensitization, in-situ metallography is used to visually inspect the material microstructure in service. A current issue is the ability to accurately quantify sensitization based on visual interpretation of a microstructure.
A computer vision approach utilizing machine learning has been developed to train an existing data set to accurately predict the level of sensitization of unknown images correlated to ASTM G67 mass loss values. This study will discuss the methodology, augmentation of models used and results for microstructures closely related to the training data as well as those further afield. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Aluminum, Environmental Effects |