Wavelength Demodulation of Overlapping Spectra in FBG Sensor Network Based on Deep Neural Network

Abstract

In order to solve the demodulation of Bragg wavelength of overlapping spectrum in FBG sensor network, a DNN wavelength detection method is proposed to deal with the wavelength detection of FBG overlapping spectra by introducing the idea of model regression. DNN is applied to learn the characteristics of spectral data and train the wavelength demodulation model. The overlapping spectra can be directly and rapidly demodulated by the well-trained DNN model. The experimental results show that the wavelength demodulation model can accurately identify the central wavelength in FBG sensor network under the condition of partial or complete spectral overlap, and 80.1 percent of the results are measured with RMS error less than 1.5pm. Compared with the existing demodulation methods based on DE and PSO algorithm, the wavelength demodulation method based on DNN achieves a certain improvement in detection accuracy and detection efficiency, which provides a new way to enhance the multiplexing capability of FBG sensor network.