Machine Learning Estimation of the Phase at the Fading Points of an OFDR-Based Distributed Sensor.

Area of Science:

• Optoelectronics
• Signal Processing
• Machine Learning

Background:

• Distributed acoustic sensing (DAS) is crucial for monitoring applications.
• Optical frequency domain reflectometry (OFDR) is a common technique for DAS.
• Phase estimation in OFDR-based DAS can be challenging, especially at signal fading points.

Purpose of the Study:

• To develop a machine learning-based approach for robust phase estimation in OFDR-based DAS.
• To enhance the accuracy and reliability of phase estimation, particularly under fading conditions.
• To compare the performance of the proposed method against traditional homodyne detection.

Main Methods:

• A neural network was trained using simulated optical signals based on Rayleigh scattering patterns.
• The trained neural network was validated using numerically generated scattering profiles.
• The method was experimentally tested on real-world measurements from a perturbed fiber optic cable.

Main Results:

• The machine learning approach demonstrated enhanced robustness at signal fading points.
• Numerical simulations showed comparable or improved accuracy against the standard homodyne detection method.
• Real experimental measurements indicated a detection improvement of at least 5.1 dB over the standard approach.

Conclusions:

• Machine learning offers a promising solution for improving phase estimation in OFDR-based DAS.
• The proposed neural network method provides significant performance gains in real-world sensing scenarios.
• This advancement has implications for more reliable and sensitive fiber optic sensing systems.