Vision-Based Pose Estimation of Superimposed Enameled Wire Ends for Robotic Handling of Powerdense Flat Wire Stators

In the production of electric drives, achieving highly efficient performance in new use cases is essential to meet global climate targets. To maximize motor power while minimizing motor weight, the stator slots must be filled with as much copper as possible. Traditionally, round wires have been the standard; however, with increasing efficiency demands, they can no longer meet performance expectations. Form coils, also known as hairpins, offer a higher fill factor and are becoming a more common technology. In addition, continuous winding processes using flat or rectangular wires are under development, which could meet future demands for electric drives. Along with new joining solutions, this technology has the potential to reduce production costs while minimizing contact points that could otherwise increase contact resistance. A key aspect of automating these processes is the precise and reliable detection of wire positions, which is critical for achieving a fully automated process chain. In this paper, an AI-based algorithm for flat wire detection is proposed, enableling seamless integration in robot-based systems, ensuring accurate wire positioning throughout the production stages. By incorporating AI-driven recognition methods, high-quality and efficient automation of electric drive manufacturing processes can be realized, from the initial winding steps to final end-of-line testing.