DATA-EFFICIENT AI-BASED PROGNOSTICS AND HEALTH MANAGEMENT FOR WIRING HARNESS MANUFACTURING

Abstract

Wiring harnesses are critical components in electronic devices, providing power and connectivity. Their quality significantly impacts the product's remaining useful life due to their essential role. Ensuring high-quality manufacturing is crucial, as defects in the wiring harness can lead to poor terminal tension, causing loose connections or severe failures over time. These issues not only compromise reliability but also increase manufacturing costs. Traditional AI-based defect detection methods typically require extensive datasets containing both normal and defective examples, making them costly and impractical in manufacturing environments. This research addresses the challenge by employing a novel synthetic data generation approach powered by Artificial Intelligence (AI) and anomaly detection. Unlike traditional methods, this study uses only normal data to train the model, generating synthetic anomalies through regional scaling transformations. The model is validated using real manufacturing data, demonstrating its capability to identify machine faults effectively and prevent costly production losses. This approach provides a novel, data-efficient solution to enhance the quality and reliability of wiring harness manufacturing.