BROWSE
Detail
Conference Paper
A Study on the Distributed Acoustic Sensing Technology for Deep Learning-based Environmental Monitoring and Failure Classification of Underground Facilities
Cited 0 time in 
Share 
Share
- Authors
- Dae-gil Kim, Chan Il Yeo, Si Woong Park, Kyeeun Kim, Hyoung Jun Park
- Issue Date
- 2024-10
- Citation
- International Conference on Condition Monitoring and Diagnosis (CMD) 2024, pp.482-484
- Language
- English
- Type
- Conference Paper
- Abstract
- In this study, we present a system that utilizes distributed acoustic sensing (DAS) technology integrated with deep learning models to monitor and classify environmental anomalies in underground facilities. To simulate and capture relevant impact signals, we conducted experiments by burying fiber optic cables under concrete and inducing various types of vibrations using an excavator. These signals were then processed and analyzed using a combination of 1D convolutional neural networks (1D CNN) and bidirectional long short-term memory (BiLSTM) networks. The deep learning models were trained on a dataset of over 13,000 samples, with data grouped into 15-second segments to capture the temporal patterns of the anomalies. Proposed system demonstrated a high classification mean average precision of 95. 71 % and efficient real-time processing capabilities, highlighting its potential for enhancing the safety and operational security of underground facilities. The successful integration of DAS with AI underscores the system's capability to detect and respond to environmental anomalies promptly, providing a robust solution for monitoring the integrity of subterranean infrastructure.
- KSP Keywords
- Bidirectional Long Short-Term Memory, Convolution neural network(CNN), Distributed acoustic sensing, Environmental Monitoring, Learning-based, Operational Security, Real-Time processing, Robust solution, Temporal patterns, Underground Facilities, deep learning(DL)