Full spectral interrogation of π-phase-shifted TFBG-based refractive index sensor using convolutional neural network

In this paper, a refractive index (RI) sensor based on π-phase-shifted titled fiber Bragg grating (π-PSTFBG) is proposed, using a dense connected convolutional neural network (D-CNN) for real-time interrogation. The proposed π-PSTFBG sensor has rich spectral characteristics, which could introduce deeper features of phase-shift points for all the lossy dips in spectrum. The experimental results show that the well-trained model can accurately predict the RI value of NaCl solution with the prediction accuracy of 99.67%, and it also has excellent generalization ability and good demodulation performance for fresh samples, which has broad application prospects in many fields such as biology and chemistry.