Abstract
Monitoring weld pool geometry without the appropriate auxiliary light source remains challenging due to the interference from the intense arc light. In this work, a new software framework was developed to measure the key features related to welding pool three-dimensional (3D) geometry based on the two-dimensional (2D) passive vision images. It was found that the interference of the arc light on the weld pool image can be effectively controlled by adjusting the camera exposure time based on the decision made from machine learning classifier. Weld pool width, trailing length, and surface height (SH) were calculated in real time, and the result agreed with the measurement of the weld bead geometry. The method presented here established the foundation for real-time penetration monitoring and control.
| Original language | English |
|---|---|
| Pages (from-to) | 379s-386s |
| Journal | Welding Journal |
| Volume | 98 |
| Issue number | 12 |
| DOIs | |
| State | Published - Dec 2019 |
Funding
This research was sponsored by the U.S. Department of Energy, Office of Nuclear Energy, for Nuclear Energy Enabling Technologies Crosscutting Technology Development Effort, under a prime contract with Oak Ridge National Laboratory (ORNL). ORNL is managed by UT-Battelle LLC for the U.S. Department of Energy under Contract DE-AC05-00OR22725.
Keywords
- 3D Geometry
- Adaptive Passive Vision
- Penetration
- Weld Pool