The transition to battery-powered electric vehicles (EVs) is an approach to mitigate climate change by reducing greenhouse gas emissions. However, the manufacturing process, particularly the laser beam welding for the battery contacting during the assembly phase, poses safety relevant challenges. Although cell-to-pack battery storage concepts are efficient concerning the maximum range, their production can cause cost-intensive rework or significant scrap. For the battery contacting of 4680 cells, aluminum-aluminum or aluminium-hilumin material combinations are used, respectively. In order to assess multi-spectral monitoring as a new inline quality assurance procedure, welding experiments based on equivalent geometries for both material combinations were conducted. In this study, a single mode fibre laser combined with pre-focusing optics were used. For both material configurations, references and weld seams were manufactured. Based on the references, artificially introduced failures, such as gaps or false friends, were used to simulate real failures in production. As the these failures can occur in varying combinations, several combinations of them were investigated. For the inline process monitoring, the spectral emissions were recorded through a 4D.Two sensor and spectrally resolved over 32 channels (16 of them in the visual spectrum and another 16 in the near-infrared spectrum). In addition, an approach which is able to evaluate the detectability of the individual failures and its combinations was elaborated. Based on this methodology, it is possible to identify the relevant features quickly to set up an inline process monitoring.
Keywords
- Battery Contacting
- Explainable Artificial Intelligence
- Laser Beam Welding
- Machine Learning
- Multi-Spectral Monitoring