| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets from Advanced Characterization Techniques
|
| Presentation Title |
AI-Driven Kikuchi Pattern Enhancement for Efficient and Robust EBSD Analysis of Highly Deformed Metals |
| Author(s) |
Ayoub Dergaoui, Siyu Tu, Noureddine Barka |
| On-Site Speaker (Planned) |
Siyu Tu |
| Abstract Scope |
Electron backscatter diffraction (EBSD) is an informative tool for studying crystalline materials but is limited by slow analysis speed and high sensitivity to surface quality. EBSD analysis of highly deformed metals often requires a small acquisition step size and suffers from low Kikuchi pattern quality, resulting in lengthy acquisition times and low indexation rates. Pattern averaging is commonly employed to improve the signal-to-noise ratio of Kikuchi bands, thereby boosting the indexation rate, albeit at the cost of increased acquisition time. In this work, we developed a generative adversarial network to enhance Kikuchi patterns for more efficient and robust indexation. This approach significantly improved EBSD acquisition speed by multiple times with the removal of pattern averaging, and increased indexation rates from 50%-70% to over 95% for highly deformed aluminum sheets. This acquire-enhance-index workflow offers a practical solution for large-area, high-quality mapping of highly deformed metals. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Aluminum |