About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Forming and Joining of Advanced Sheet Metal Materials
|
| Presentation Title |
PREDICTION OF SHEARED EDGE FRACTURE STRAIN OF DP980 USING U-NET BASED EXTRACTION OF EDGE GEOMETRY |
| Author(s) |
Weitian Zhou, Lu Tim Huang, Sergey Golovashchenko , Arman Hossain , Thanush Ganaparthi |
| On-Site Speaker (Planned) |
Weitian Zhou |
| Abstract Scope |
Shearing of advanced high-strength steels (AHSS), such as DP980, can produce edge conditions that compromise downstream forming operations, leading to edge cracking. This study aims at developing a machine vision system for proactive inspection and quality assurance of sheared edges in DP980 steel. To simulate a range of edge conditions, sheets were sheared under varied cutting clearances and progressive stroke counts. Edge characteristics were captured using 2D RGB imaging using a consumer DSLR camera with macro lens, and 3D scanning via high resolution laser line profilometer. Edge stretchability under the selected shearing conditions was evaluated through tensile testing with DIC(digital image correlation. Sheared edge geometries are measured with traditional machine vision algorithms and a U-Net based segmentation model. Correlations amongst shearing parameters, edge geometry, and edge stretchability were investigated using traditional statistical methods and machine learning to support data-driven quality control in high-strength steel processing. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Iron and Steel, Machine Learning, |