@article{ART003283761},
author={Tae-O Lee and KIM, TAEKOOK},
title={A Study on Anomaly Detection Methods for Defective Product Prediction},
journal={Journal of Internet of Things and Convergence},
issn={2466-0078},
year={2025},
volume={11},
number={6},
pages={26}
TY - JOUR
AU - Tae-O Lee
AU - KIM, TAEKOOK
TI - A Study on Anomaly Detection Methods for Defective Product Prediction
JO - Journal of Internet of Things and Convergence
PY - 2025
VL - 11
IS - 6
PB - The Korea Internet of Things Society
SP - 26
EP -
SN - 2466-0078
AB - This study proposes an approach for predicting defects in manufacturing environments where accurate defect labels in equipment operation data do not exist. This approach considers outliers as potential signals of defect occurrence and generates pseudo-labels. For outlier detection, we employ three algorithms with different characteristics: K-Nearest Neighbor (KNN), Isolation Forest, and Local Outlier Factor (LOF). To enhance model reliability, we compare and analyze two reinforcement strategies: (1) a majority-vote-based outlier detection method across the three models and (2) an average of the outlier scores across the three models. To address the limitations of direct performance evaluation due to the lack of labeled raw data, we validated the proposed method by applying the model to labeled data collected from other equipment with similar process characteristics. Experimental results show that, of the three approaches, the average of the outlier scores demonstrated the most stability and superior predictive performance. This study demonstrates that practical defect prediction is possible even in process data environments where defect labels are insufficient or absent, and provides an empirical and methodological foundation for building a predictive maintenance system based on the extrusion process.
KW - Internet of Things(IoT);artificial intelligence(AI);Extrusion process;Anomaly detection;;Defect prediction;Predictive maintenance;Unsupervised learning
DO -
UR -
ER -
Tae-O Lee and KIM, TAEKOOK. (2025). A Study on Anomaly Detection Methods for Defective Product Prediction. Journal of Internet of Things and Convergence, 11(6), 26.
Tae-O Lee and KIM, TAEKOOK. 2025, "A Study on Anomaly Detection Methods for Defective Product Prediction", Journal of Internet of Things and Convergence, vol.11, no.6 26.
Tae-O Lee, KIM, TAEKOOK "A Study on Anomaly Detection Methods for Defective Product Prediction" Journal of Internet of Things and Convergence 11.6 26 (2025) : 26.
Tae-O Lee, KIM, TAEKOOK. A Study on Anomaly Detection Methods for Defective Product Prediction. 2025; 11(6), 26.
Tae-O Lee and KIM, TAEKOOK. "A Study on Anomaly Detection Methods for Defective Product Prediction" Journal of Internet of Things and Convergence 11, no.6 (2025) : 26.
Tae-O Lee; KIM, TAEKOOK. A Study on Anomaly Detection Methods for Defective Product Prediction. Journal of Internet of Things and Convergence, 11(6), 26.
Tae-O Lee; KIM, TAEKOOK. A Study on Anomaly Detection Methods for Defective Product Prediction. Journal of Internet of Things and Convergence. 2025; 11(6) 26.
Tae-O Lee, KIM, TAEKOOK. A Study on Anomaly Detection Methods for Defective Product Prediction. 2025; 11(6), 26.
Tae-O Lee and KIM, TAEKOOK. "A Study on Anomaly Detection Methods for Defective Product Prediction" Journal of Internet of Things and Convergence 11, no.6 (2025) : 26.
