Holistic Quality Monitoring Based on Machine Learning Methods How Data-driven Approaches Could Revolutionize Process Reliability in the Contact Processing Industry

Contact and wire processing is characterized by a high component variety, short cycle times, and increasing requirements regarding quality, documentation, and traceability. To fulfil these requirements, this paper presents a holistic approach based on machine learning for quality monitoring. The approach is based on an automatic optical inspection with 360-degree views of stripped and contacted wires. In addition, quality monitoring is realized based on the intelligent analysis of crimp force curves. The evaluation of image data and time series enables failure classification and anomaly detection at the crimping machine without sacrificing cycle time. For the visualization and worker acceptance of the quality parameters and predictions of the deep learning models, methods for explainability are integrated. © 2023 Walter de Gruyter GmbH, Berlin/Boston, Germany.