About this Abstract |
Meeting |
2021 AWS Professional Program
|
Symposium
|
2021 AWS Professional Program
|
Presentation Title |
Weld Penetration Model Based on Coaxial Image and Photodiode Signal During Laser Welding of Al and Cu Overlap joints |
Author(s) |
Kidong Lee, Sanghoon Kang, Minjung Kang, Sung Yi, Cheolhee Kim |
On-Site Speaker (Planned) |
Cheolhee Kim |
Abstract Scope |
Al/Cu dissimilar metal joining is required for the tap-to-tap and tap-to-busbar joining in automotive secondary battery manufacturing. Sound Welds between sheets are crucial to achieve joint strength and electrical conductance. In this study, laser welding was implemented on the overlap joints of thin sheets. Weld pool images and light irradiation were collected using a coaxial camera and a photodiode, respectively, to establish weld penetration depth prediction models. One model (uni-sensor model) was developed using only the weld pool image, and the other model (multi-sensor model) used using both the weld pool image and the photodiode signal. Convolution neural networks (CNNs) were adopted in the models. Both prediction models had an r-squared value of 0.98; however, slightly lower error was observed in the multi-sensor model. |
Proceedings Inclusion? |
Definite: Other |