Automatic Multimodal Machine Learning for Quality Prediction of Hairpin Welds

Baader M, Raffin T, Henrich V, Spacke J, Franke J, Risch F (2026)

Publication Language: English

Publication Type: Conference contribution

Publication year: 2026

Conference Proceedings Title: 2025 2nd International Conference on Production Technologies and Systems for E-Mobility (EPTS)

Event location: Karlsruhe

ISBN: 979-8-3315-8845-8

DOI: 10.1109/EPTS67931.2025.11547763

Abstract

The growing demand for electric vehicles is driving advances in electric motor manufacturing, particularly with the introduction of Hairpin stators. These stators offer deterministic manufacturing processes and improved performance, though also present manufacturing challenges due to complex interdependencies between the production steps. Among these, the laser welding process, which is critical to stator functionality and quality, requires robust in-line quality monitoring systems. Despite ongoing research, no automated solutions exist. This study explores the use of automatic Machine Learning (AutoML) libraries to assess Hairpin welding quality through multimodal learning, combining image and time series data. A case study evaluates the effectiveness of these methods using sensor data from a vision system, a broadband process light sensor, as well as an optical coherence tomography system, focusing on the prediction of the cross-sectional area of the welds, a key quality indicator. The results show that manual pre-processing, transfer learning, and AutoML ensemble methods improved prediction accuracy, with infrared process light data performing best. The study demonstrates the potential of AutoML to predict the cross-sectional area of Hairpin welds, paving the way for more efficient and scalable solutions in electric motor manufacturing.

Authors with CRIS profile

How to cite

APA:

Baader, M., Raffin, T., Henrich, V., Spacke, J., Franke, J., & Risch, F. (2025). Automatic Multimodal Machine Learning for Quality Prediction of Hairpin Welds. In 2025 2nd International Conference on Production Technologies and Systems for E-Mobility (EPTS). Karlsruhe, DE.

MLA:

Baader, Marcel, et al. "Automatic Multimodal Machine Learning for Quality Prediction of Hairpin Welds." Proceedings of the 2nd International Conference on Production Technologies and Systems for E-Mobility (EPTS) 2025, Karlsruhe 2025.

BibTeX: Download