Approach for the Development for an Adaptive Vacuum Laser Welding Process for Hairpin Stators Using Supervised Learning

• Published in: International Electric Drives Production Conference (Online) pp. 1 - 10
• Main Authors: Kampker, Achim, Brans, Florian, Heimes, Heiner Hans, Bajah, Yazan, Dorn, Benjamin, Gerhards, Benjamin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 29.11.2023
• Subjects: adaptive; hairpin; Lasers; machine learning; Machine learning algorithms; Process control; Software algorithms; stator; supervised; Supervised learning; Vacuum systems; vacuum welding; Welding
• ISSN: 2832-6385
• DOI: 10.1109/EDPC60603.2023.10372168

Within the process chain of hairpin technology, the contacting of the hairpin ends by laser welding is the decisive quality-determining process step of the entire process chain. The reason for this is that the laser welding process is subject to a large number of interdependencies from upstream process steps which influence the process and quality. In order to counter this bottleneck in production technology this research addresses the approach for the development of adaptive welding processes taking place in a vacuum with the aim of optimizing the three target dimensions of quality, costs, and time. The intended technological development consists in the construction of a vacuum laser welding system for the hairpin contacting process that can be adaptively controlled by data utilization using machine learning algorithms. Depending on the initial welding situation (e.g., gap formation, lateral, angular and height offset), the welding process must be adapted to the situation. This will be implemented with software that is being developed and aims to determine the optimum welding parameters for the next welding situation based on the results of welds that have already been performed. To enable this adaptive welding control, a supervised learning algorithm shall be developed during the research project.