Towards an Inline Quality Monitoring for Crimping Processes Utilizing Machine Learning Techniques

With the increasing demand for electric vehicles, automobile manufacturers and suppliers need to adjust their value chains to meet future requirements. In wire harness assembly, which is part of the production of vehicle electrical systems, the joining of wires and connecting elements, usually realized by crimping, is one of the most complex and quality-critical processes. For quality assessment, the crimp height and pull-out force are measured as primary quality criteria. In the context of the increased safety requirements of autonomous driving, monitoring the crimp connections should be automated, holistic, and in real-time for each wire-crimp connection during production. However, the high complexity of internal and external influences as well as the variety of crimp connections complicates the operation of conventional process monitoring systems based on hard-coded criteria. In this context, data-driven approaches using the methods of artificial intelligence are moving into focus. The emerging development of information technology opens up new perspectives for process monitoring systems with inherent intelligence. To provide a proof of concept for an intelligent process monitoring, in this paper experiments are carried out on a crimping station, provoking different types of error conditions. The resulting dataset is then analyzed using deep learning. Finally, a comparison of the approaches shown is carried out, followed by an outlook on future research activities.