Domain-knowledge-informed functional outlier detection for line quality control systems

Detecting defective products at quality inspection stations is crucial. Consequently, modern production systems collect diverse sensor data during inspections to monitor the condition of products. However, a significant challenge in the pursuit of zero -defect manufacturing emerges with the presence of latent defects. These defects are not discoverable during the quality inspection phase and become apparent in the early stages of customer use. As a result, detecting such defects solely based on collected data becomes almost impossible. In this study, we introduce a novel functional outlier detection method that leverages domain knowledge to identify defective products, especially those with latent defects. The proposed method presents a systematic framework for integrating domain knowledge into the recently developed functional outlier detection method known as sequential transformations (Dai et al., 2020). To validate our proposed method's effectiveness, we evaluated its performance using simulated data and real sensor data from refrigerator inspection lanes.