In-process Monitoring of Penetration Depth and Contact Width During Laser Lap Welding Using Fused OCT Depth and Coaxial High Speed Image

Electric mobility has emerged as a paradigm shift towards cleaner, more efficient, and sustainable transportation systems. Laser lap welding is being adopted in battery manufacturing due to its unique advantages including high throughput, noncontact access, excellent precision, and small heat-affected zone. An important challenge of laser lap welding is the limited capability for in-process monitoring of penetration depth and contact width of weldment, which determines electrical resistance and joint strength. This requires three-dimensional live information of the molten pool/keyhole. The existing techniques are limited in surface imaging or indirect measurement of penetration depth, hence the precision is limited due to the intrinsic starving-signal in all dimensions. To address this issue, an optical system was designed to couple optical coherence tomography depth imaging with a coaxial high-speed camera to capture the three-dimensional information synchronously. A multimodal deep learning network was created to fuse the two-source data for predicting the depth and contact width. Experiments were conducted with different laser power and welding speeds, and the cross-sections of the weldment were cut and micro-imaged as the ground truth. Compared to the individual sensing systems, the proposed fused system exhibits improved precision on both the depth and contact width.