FBG Overlapping Spectrum Wavelength Detection Method Based on Wave U-Net and CNN-LSTM

In Fiber Bragg Grating (FBG) sensing networks, highly overlapping reflection spectra can significantly degrade the demodulation accuracy of FBG signals. To address this issue, this paper proposes a sensing system based on a region-sharing model, combining Wave UNet for FBG overlapping spectrum separation and a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model for individual spectrum demodulation and temperature prediction. Experimental results show that the predicted temperatures achieve a curve fitting accuracy ($\mathbf{R}^{\mathbf{2}}$) of 0.99953 with the actual temperatures, along with an average temperature error of $\mathbf{0. 3 9 3 4}^{\circ} \mathbf{C}$. This method enables highprecision wavelength detection even under spectral overlap conditions, offering a novel solution to enhance the temperature sensing performance of FBG networks.