Scholarlink

<p>With electric vehicle (EV) batteries becoming increasingly compact, the copper sheets used for current collectors and electrode tabs are proportionally reduced in thickness. Laser welding is widely employed for joining such thin copper sheets; however, maintaining consistent weld quality at high production speeds remains challenging. This study introduces a real-time approach for predicting weld quality in thin copper sheets during laser welding, leveraging high-speed thermal monitoring and artificial intelligence (AI). Experiments were conducted on 0.1- and 0.3-mm-thick copper sheets at a welding speed of 200 mm/s. Peak temperature and cooling rate data were obtained from thermal images acquired at 1000 Hz and were used as input features for a one-dimensional convolutional neural network model to predict tensile shear load. The model achieved a mean absolute error of 8.85 N and a mean percentage error of 4.56%. A classification model was developed to differentiate between pass and fail cases based on the predicted tensile shear load. Results confirm the potential of AI-driven thermal monitoring for reliable, real-time weld quality assessment in high-speed EV battery manufacturing.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Tensile shear load prediction in laser welding of thin copper sheets using real-time thermal monitoring and one-dimensional convolutional neural network

  • 期刊论文

  • 发表日期：

    2026年4月14日

查看全文
  • Hyunjong Yoo,
  • Heeshin Kang,
  • Ikgeun Jeon

摘要

With electric vehicle (EV) batteries becoming increasingly compact, the copper sheets used for current collectors and electrode tabs are proportionally reduced in thickness. Laser welding is widely employed for joining such thin copper sheets; however, maintaining consistent weld quality at high production speeds remains challenging. This study introduces a real-time approach for predicting weld quality in thin copper sheets during laser welding, leveraging high-speed thermal monitoring and artificial intelligence (AI). Experiments were conducted on 0.1- and 0.3-mm-thick copper sheets at a welding speed of 200 mm/s. Peak temperature and cooling rate data were obtained from thermal images acquired at 1000 Hz and were used as input features for a one-dimensional convolutional neural network model to predict tensile shear load. The model achieved a mean absolute error of 8.85 N and a mean percentage error of 4.56%. A classification model was developed to differentiate between pass and fail cases based on the predicted tensile shear load. Results confirm the potential of AI-driven thermal monitoring for reliable, real-time weld quality assessment in high-speed EV battery manufacturing.