A Case Study of Anomaly Detection in Tinplate Lids: Supervised vs Unsupervised Approaches

This research presents a comparative study of deep learning-based anomaly detection methods, both supervised and unsupervised, applied to industrial systems for detecting product defects in manufacturing. The study implements the OPC-UA protocol for data communication and algorithm execution using finite state machines, demonstrating its practical application in a tinplate lid system. Integration with OPC-UA ensures realtime data access, interoperability, and scalability across various industrial environments. The experimental results, evaluated using metrics such as Average Precision, Mean AUROC, Mean Pixel AUROC, and Execution Time (CPU and GPU), reveal the strengths and limitations of each approach, providing valuable insights for addressing modern challenges in industrial anomaly detection.