Machine learning assisted BOFDA for simultaneous temperature and strain sensing in a standard optical fiber

Area of Science:

• Optoelectronics and Photonics
• Machine Learning Applications
• Fiber Optic Sensing Technology

Background:

• Distributed fiber optic sensing is crucial for structural health monitoring and environmental sensing.
• Temperature and strain cross-sensitivity in optical fibers is a significant challenge in accurate measurements.
• Brillouin optical frequency domain analysis (BOFDA) offers high spatial resolution but requires advanced signal processing.

Purpose of the Study:

• To demonstrate simultaneous distributed temperature and strain sensing in standard telecom optical fiber for the first time.
• To address the temperature-strain cross-sensitivity problem using a machine learning-assisted BOFDA system.
• To evaluate the performance of different machine learning algorithms for discriminating temperature and strain effects.

Main Methods:

• Development of a high signal-to-noise ratio BOFDA system.
• Application of machine learning algorithms, including Gaussian Process Regression (GPR), to analyze Brillouin frequency shifts.
• Experimental setup involving a 450-m SMF-28 fiber segment within a climate chamber and stretcher.
• Leave-one-out cross-validation for unbiased performance estimation.

Main Results:

• Successfully achieved simultaneous distributed temperature and strain sensing with a spatial resolution of 6 m.
• Gaussian Process Regression (GPR) demonstrated the best performance, with errors of 2 °C for temperature and 45 µɛ for strain.
• The system resolved four highly defined peaks in the Brillouin spectrum, enabling feature extraction for machine learning.
• Measurement time of 16 minutes was sufficient to achieve high signal-to-noise ratio and resolve spectral peaks.

Conclusions:

• The developed machine learning-assisted BOFDA system effectively solves the temperature-strain cross-sensitivity problem in optical fiber sensing.
• GPR provides a robust method for accurate discrimination of temperature and strain, paving the way for enhanced distributed sensing applications.
• This approach offers a promising solution for precise, simultaneous monitoring of temperature and strain in various industrial and scientific fields.