Fading suppression method based on redundant data within the spatial resolution and deep learning for a Φ-OTDR system

Area of Science:

• Optical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Interference fading in phase-sensitive optical time-domain reflectometry (Φ-OTDR) degrades sensing performance.
• Existing fading suppression techniques often require complex hardware modifications to the light source.

Purpose of the Study:

• To propose a novel multi-channel data synthesizing method based on deep neural networks (MDS-DNN) for interference fading suppression in Φ-OTDR.
• To improve the signal-to-noise ratio (SNR) and reduce the false alarm rate without altering the conventional Φ-OTDR setup.

Main Methods:

• Developed a deep neural network (DNN) algorithm, specifically a long short-term memory (LSTM) network.
• Utilized a multi-channel data synthesizing approach leveraging redundant information from neighboring spatial sampling points.
• Implemented an end-to-end training strategy to learn correlations between multi-channel data and the ideal sensing signal.

Main Results:

• The MDS-DNN algorithm effectively suppresses phase noise and enhances SNR at fading positions.
• Experimental results demonstrated an output SNR of 49.88 dB, a 19.65 dB improvement over input channels.
• The method reduced the false alarm rate caused by interference fading by one order of magnitude.

Conclusions:

• The proposed MDS-DNN method offers an efficient solution for mitigating interference fading in Φ-OTDR.
• This approach enhances sensing performance by improving SNR and reducing false alarms without requiring hardware modifications.
• The study highlights the potential of deep learning for advanced signal processing in optical sensing.