Robust Vector BOTDA Signal Processing with Probabilistic Machine Learning

Area of Science:

• Optoelectronics
• Machine Learning
• Fiber Optic Sensing

Background:

• Vector Brillouin optical time-domain analysis (VBOTDA) is crucial for fiber optic sensing.
• Accurate estimation of Brillouin frequency shift (BFS) is essential for VBOTDA performance.
• Current methods for BFS estimation may lack precision and uncertainty assessment.

Purpose of the Study:

• To introduce a novel probabilistic machine learning (PML) framework for BFS estimation in VBOTDA.
• To assess the predictive uncertainty associated with the PML-based BFS estimation.
• To compare the performance of the PML framework against conventional methods.

Main Methods:

• Development of a probabilistic machine learning (PML) framework.
• Application of the PML framework to estimate BFS from Brillouin gain and phase spectra.
• Comparison with a conventional curve fitting method.
• Validation using two BOTDA systems (10 km and 25 km fiber).

Main Results:

• The PML framework accurately predicts BFS along the sensing fiber.
• The framework provides reliable assessment of BFS predictive uncertainty.
• Demonstrated effectiveness in both 10 km and 25 km VBOTDA systems.
• Potential for reduced data processing time compared to conventional methods.

Conclusions:

• The proposed PML framework offers a robust approach for BFS estimation in VBOTDA.
• This method enhances VBOTDA system performance by providing accurate BFS prediction and uncertainty quantification.
• PML presents a promising direction for advancing fiber optic sensing technologies.