TY - GEN
T1 - New anomaly detection in semiconductor manufacturing process using oversampling method
AU - Song, Seunghwan
AU - Baek, Jun Geol
N1 - Funding Information:
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (NRF-2019R1A2C2005949). This work was also supported by the BK21 Plus (Big Data in Manufacturing and Logistics Systems, Korea University) and by the Samsung Electronics Co., Ltd.
Publisher Copyright:
Copyright © 2020 by SCITEPRESS - Science and Technology Publications, Lda. All rights reserved
PY - 2020
Y1 - 2020
N2 - Quality in the semiconductor manufacturing process, consisting of various production systems, leads to economic factors, which necessitates sophisticated abnormal detection. However, since the semiconductor manufacturing process has many sensors, there is a problem with the curse of dimensionality. It also has a high imbalance ratio, which creates a classification model that is skewed to multiple class, thus reducing the class classification performance of a minority class, which makes it difficult to detect anomalies. Therefore, this paper proposes AEWGAN (Autoencoder Wasserstein General Advertising Networks), a method for efficient anomaly detection in semiconductor manufacturing processes with high-dimensional imbalanced data. First, learn autoencoder with normal data. Abnormal data is oversampled using WGAN (Wasserstein General Additional Networks). Then, efficient anomaly detection within the potential is carried out through the previously learned autoencoder. Experiments on wafer data were applied to verify performance, and of the various methods, AEWGAN was found to have excellent performance in abnormal detection.
AB - Quality in the semiconductor manufacturing process, consisting of various production systems, leads to economic factors, which necessitates sophisticated abnormal detection. However, since the semiconductor manufacturing process has many sensors, there is a problem with the curse of dimensionality. It also has a high imbalance ratio, which creates a classification model that is skewed to multiple class, thus reducing the class classification performance of a minority class, which makes it difficult to detect anomalies. Therefore, this paper proposes AEWGAN (Autoencoder Wasserstein General Advertising Networks), a method for efficient anomaly detection in semiconductor manufacturing processes with high-dimensional imbalanced data. First, learn autoencoder with normal data. Abnormal data is oversampled using WGAN (Wasserstein General Additional Networks). Then, efficient anomaly detection within the potential is carried out through the previously learned autoencoder. Experiments on wafer data were applied to verify performance, and of the various methods, AEWGAN was found to have excellent performance in abnormal detection.
KW - Anomaly Detection
KW - Autoencoder
KW - Latent Space
KW - Semiconductor Manufacturing Process
KW - Wasserstein Generative Adversarial Networks
UR - http://www.scopus.com/inward/record.url?scp=85083093896&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85083093896
T3 - ICAART 2020 - Proceedings of the 12th International Conference on Agents and Artificial Intelligence
SP - 926
EP - 932
BT - ICAART 2020 - Proceedings of the 12th International Conference on Agents and Artificial Intelligence
A2 - Rocha, Ana
A2 - Steels, Luc
A2 - van den Herik, Jaap
PB - SciTePress
T2 - 12th International Conference on Agents and Artificial Intelligence, ICAART 2020
Y2 - 22 February 2020 through 24 February 2020
ER -