Resolution enhancement for interrogating fiber Bragg grating sensor network using dilated U-Net.

Area of Science:

• Optoelectronics
• Sensor Technology
• Artificial Intelligence

Background:

• Signal resolution in fiber Bragg grating (FBG) sensor networks directly impacts sensing accuracy.
• Low signal-to-noise ratios and overlapping multi-peak signals complicate resolution enhancement.
• Current interrogators have inherent resolution limits, leading to measurement uncertainty.

Purpose of the Study:

• To enhance the signal resolution of FBG sensor networks using deep learning.
• To improve the accuracy of FBG sensing systems without requiring hardware modifications.
• To overcome limitations of low-resolution interrogators and complex signal scenarios.

Main Methods:

• Implementation of a deep learning model based on the U-Net architecture.
• Training the model to process and enhance spectral signals from FBG sensor networks.
• Validation of the model's performance using root mean square error (RMSE).

Main Results:

• Achieved a 100-fold enhancement in signal resolution.
• Demonstrated an average RMSE of less than 2.25 pm.
• Successfully enabled low-resolution interrogators to function as high-resolution devices.

Conclusions:

• Deep learning, specifically U-Net, offers an effective solution for enhancing FBG sensor signal resolution.
• The proposed method significantly improves sensing accuracy and reduces uncertainty in FBG networks.
• This approach provides a cost-effective upgrade path for existing FBG sensing systems.