| Title |
Development of Anomaly-Detection System for the Underground Cable Tunnel using Autoencoder |
| Authors |
강수경(Su-Kyung Kang) ; 박명혜(Myung-Hye Park) ; 김영현(Young-Hyun Kim) ; 김낙우(Nac-Woo Kim) ; 서인용(In-Yong Seo) |
| DOI |
https://doi.org/10.5370/KIEEP.2020.69.2.69 |
| Keywords |
Autoencoder; Stacked Autoencoder; Self-attention; Anomaly-Detection; Underground Cable Tunnel; Adversarial-Autoencoder |
| Abstract |
With the advent of the era of the 4th industrial revolution, AI technique has been applied in various fields. KEPCO developed an anomaly detection system using AI technology to detect abnormal situation in the underground cable tunnel. This anomaly-detection system consists of a robot-based data acquisition, communication and analysis module which works with stacked autoencoder neural network model. This system utilizes the data from audio sensors and determines the condition of equipment in the underground cable tunnel which is normal or abnormal. Moreover, by adding the attention-module in autoencoder neural network model we increased the recognition accuracy by 4%. The performance of this system is over 90%. Also, we investigated the performance of adversarial autoencoder (AAE) and attention based AAE model, which showed worse failure detection rate than attention added autoencoder model. |