Laser brazing as well as laser welding of aluminum are well known laser applications, and the processes are fully understood in terms of robustness for a series production. Nevertheless, pores and other surface defects are common failures that reduce the quality of the seam.


Recent studies show how the AI-based analysis can be used for brazing and for aluminum welding. Here, the fillet joints of a battery box are of greatest interest due to high quality requirements. Particularly in the case of the battery boxes, it is not only a question of being able to detect smallest defects such as pores and spatter, but also other defect classes such as one-sided wetting or misalignment of wire to beam. Therefore, multiple defect classes need to be distinguished automatically.


The talk will present latest results and further improvements on AI-based analysis for laser brazing and welding. Results of a series production for roof brazing as well as for trunk lid brazing will be presented. The transfer of this technology to aluminum welding applications will be reported. The robust and stable classification of multiple defect classes is demonstrated.


To further improve the AI analysis and to combine the analysis of image data and data from other sensors, Scansonic is working to gether with the Fraunhofer-Institute for Laser Technology ILT in a public funded project – DiReAL.


As an outlook a clear and simple visualization of the processed parts and the corresponding defects found by the AI will be addressed.