Separation of distorted overlapping spectra in FBG sensor networks using self-supervised contrastive learning.

Area of Science:

• Optical Engineering
• Materials Science
• Artificial Intelligence

Background:

• Fiber Bragg Grating (FBG) sensor networks face challenges with spectral overlap and distortion, particularly in applications like residual strain monitoring in composite structures.
• These spectral issues limit the multiplexing capability and sensing accuracy of FBG networks, hindering their effectiveness in complex environments.

Purpose of the Study:

• To propose and validate a self-supervised learning framework, the contrastive spectrum separation model (CSSM), for effectively separating distorted and overlapping spectra in FBG sensor networks.
• To enhance the accuracy and multiplexing capabilities of FBG sensor networks in scenarios with non-uniform physical fields.

Main Methods:

• Developed a contrastive spectrum separation model (CSSM) utilizing a dual-encoder architecture with parallel convolutional neural networks.
• Employed a self-supervised learning approach for direct feature extraction from distorted spectra, minimizing the need for extensive labeled training data.
• Validated the model through simulations and experimental measurements under varying spectral overlap and noise conditions, including strain gradients.

Main Results:

• CSSM demonstrated superior robustness, achieving a 54.5% Signal-to-Noise Ratio (SNR) improvement at 15 dB noise levels.
• Simulation results showed high accuracy in wavelength detection (1.6388 pm) and spectral similarity (0.9074) for separated spectra.
• Experimental validation with strain gradients up to -650 µε/mm reduced strain measurement error from ±37.2 µε to approximately 1.35 µε.

Conclusions:

• The CSSM framework effectively separates distorted overlapping spectra in FBG sensor networks, overcoming limitations of current methods.
• The model significantly enhances the accuracy and multiplexing capacity of FBG sensors, proving practical effectiveness in real-world structural health monitoring.
• This advancement enables more reliable monitoring of composite structures and other complex systems using FBG sensor technology.