About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Fatigue in Materials: Fundamentals, Multiscale Characterizations and Computational Modeling
|
| Presentation Title |
In Situ Failure Analysis of Ni-718 Using Machine Learning to Identify Failure States |
| Author(s) |
Jesse Yochens, Thomas Miller, Dino Celli, Brian Wisner |
| On-Site Speaker (Planned) |
Jesse Yochens |
| Abstract Scope |
Machines in failure typically have a noticeable difference in function and standard operation than those in normal operating status. Laser-vibrometry and piezo electric sensors and other non-destructive evaluation techniques detect sources that describe different operational states of machines or different states of materials during testing. However, correlation and detection of these failure sources have been a challenge in situ, due to different testing environments, surface constraints and computational limitations. Generally, the use of machine learning has been with purpose-built computers but with today’s commercial processors, easy to use asynchronous programming and clustering techniques allow for small-scale machine learning based systems to be used in restrictive testing environments. This work introduces a way to combine data intake from an AE sensor, hit detection, feature extraction, and machine learning in real-time to predict damage in mechanical systems to allow for a more accurate analysis of material characteristics and to increase machine life. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Mechanical Properties |