Optimized demodulation of highly overlapped fiber Bragg grating sensor networks using two-stage deep learning approach

Fiber Bragg grating (FBG) sensor arrays employ overlapped spectra in sensor channels to maximize bandwidth, often resulting in multiple local wavelength peaks that complicate accurate peak detection. Existing multiplexing methods encounter challenges due to crosstalk between adjacent sensors or high computational complexity. We propose a two-stage methodology to discern distinct wavelengths within highly overlapped FBG sensors. The method leverages a deep learning (DL) model in the initial stage to predict individual peak wavelengths. Subsequently, a peak optimization module is applied to refine these predictions by reconstructing the FBG spectrum based on the forecasts of the DL model. To validate the effectiveness of our approach, a comprehensive series of experiments was conducted. The experimental results demonstrate the superior performance of our proposed FBG demodulation scheme, achieving a remarkable 50.1% and 62.6% improvement in prediction accuracy for cases involving two and three overlapped FBG signals, respectively, in comparison to scenarios without the optimization module. The proposed method offers the potential to enhance the capacity of FBG sensors within a network, paving the way for notable advancements in signal demodulation within intricate sensor configurations.