About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Mechanical Behavior of Nuclear Reactor Materials and Components III
|
Presentation Title |
O-30: Computer Vision-assisted Oxide Thickness Determination of 304 Stainless Steel in PWR Environments |
Author(s) |
Txai TS Sibley, Rachel English, Bryan Webler, Elizabeth Holm |
On-Site Speaker (Planned) |
Txai TS Sibley |
Abstract Scope |
Environmentally assisted cracking (EAC) can be a critical factor in determining the operational safety and plant life extension of pressurized water reactors (PWRs). Oxidation kinetics, which play a key role in predicting EAC, can be determined by tracking the inner oxide thickness as a function of material exposure time. This requires the time intensive analysis and segmentation of numerous images. To expedite and standardize these measurements, automated image segmentation of the inner oxide is performed using the convolutional neural network (CNN) based U-Net architecture. Machine learning (ML) and traditional segmentation methods are compared. Removing human subjectivity in inner oxide determination and implementing high throughput segmentation and analysis enables insights into the oxidation kinetics. |
Proceedings Inclusion? |
Planned: |
Keywords |
Machine Learning, Nuclear Materials, Characterization |