Anomaly Detection for Telecom Aerial Optical Cables With Variational Auto Encoder Leveraging DAS Data

View at publisher
Authors: Keisuke Murakami, Chihiro Kito, Kunihiro Toge
Cited by: 2
References: 15

Abstract

Telecommunication aerial optical cables frequently experience structural failures that cause no immediate optical power loss. It is possible to discern signs of the failure if we can detect cable vibration that deviates from the normal condition due to changes in the facility type and in-situ conditions. We propose an anomaly detection method for aerial cables that uses variational auto-encoder (VAE) and distributed vibration data measured by Φ-OTDR. The proposed method offers robust anomaly detection even under a significant variety of vibration characteristics reflecting different facility conditions. Our proposal achieved 86.4 % accuracy in detecting anomalies in an evaluation of actual 22 cable failures consisting of “cable detachment from hangers”, “tree contact” and “support wire failure”; the 3 most common types of failures in the field.

Fields of Study & Subjects

Electrical Fault Detection and Protection Structural Health Monitoring Techniques Advanced Fiber Optic Sensors

Publication details

Journal / source
Journal of Lightwave Technology
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Volume
44
Issue
7
Pages
2646–2655
Publication date
12 January 2026