| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Aluminum Alloys: Development and Manufacturing
|
| Presentation Title |
Development of Process Control for Ultrasonic Metal Welding of Aluminum Automotive Wires Based on Machine Learning |
| Author(s) |
Andreas Gester, Guntram Wagner, Tom Kühne, Peter Gluchowski |
| On-Site Speaker (Planned) |
Andreas Gester |
| Abstract Scope |
Ultrasonic metal welding (USMW) is a technology for producing solid joints of electrical connectors, such as wires and terminals. With the ongoing substitution of copper with aluminum in conductors for weight savings, USMW sees application in automotive, aerospace and other industries. Despite its widespread use, USMW lacks sufficient process monitoring. Currently, industrial process monitoring relies on sample-based destructive testing. Monitoring all joints and avoiding false positives is unachievable with this method, leading to significant pseudo-rejects and undetected faulty welds. This work focuses on developing a process monitoring system using machine learning (ML) for analyzing USMW machine data. Utilizing ML, the system aims to classify weld quality based on this data, reducing scrap and pseudo-scrap rates. The work involves generating training data for weld faults, configuring an ultrasonic welding machine for data collection, and evaluating ML methods for accurate classification. Preliminary results indicate 99.8% accuracy, enhancing reliability and efficiency of USMW. |
| Proceedings Inclusion? |
Planned: Light Metals |
| Keywords |
Aluminum, Joining, Machine Learning |