High spatial resolution Raman distributed fiber sensing based on neural network dispersion compensation.

Area of Science:

• Optical Sensing
• Fiber Optics
• Machine Learning

Background:

• Raman distributed fiber sensing (RDFS) offers wide measurement range, low cost, and rapid response.
• Dispersion effects in sensing fibers degrade spatial resolution and temperature accuracy with increasing distance.
• Existing RDFS systems face limitations in long-distance, high-precision measurements due to dispersion.

Purpose of the Study:

• To propose and validate a novel RDFS system employing a 1D-DCRNN for dispersion compensation.
• To enhance the spatial resolution and temperature measurement accuracy of RDFS over long distances.
• To provide a dispersion compensation method without physical fiber modifications.

Main Methods:

• Development of a 1-dimensional dilated convolutional residual neural network (1D-DCRNN).
• Training the 1D-DCRNN using datasets affected by dispersion and theoretical models.
• Implementing the trained 1D-DCRNN to compensate for dispersion effects in RDFS.

Main Results:

• At 16.8 km, temperature error reduced from over 16.0 °C to below 1.0 °C.
• Spatial resolution improved from 3.0 m to 0.4 m after dispersion compensation.
• Effective dispersion compensation achieved without altering fiber infrastructure.

Conclusions:

• The proposed 1D-DCRNN-based RDFS effectively compensates for dispersion effects.
• This method significantly enhances accuracy and resolution for long-distance RDFS.
• Presents a new technical approach for advanced Raman distributed fiber sensing.