High Performance AWG-Based FBG Interrogator Using Cascaded Neural Network

The arrayed waveguide grating (AWG) is an optical device that can be used to interrogate optical signals reflected from fiber Bragg gratings (FBGs). However, achieving high-precision interrogation over a large dynamic range is hard for conventionally manufactured AWGs due to limitations in their spectral characteristics. In this paper, the structure of AWG is optimized to increase the bandwidth by incorporating a multimode interference (MMI) coupler to increase the dynamic range of interrogation. A cascaded neural network consisting of a Transformer, convolutional neural network (CNN), and backpropagation neural network (BPNN) is proposed to significantly improve interrogation precision. The experimental results demonstrate that our proposed interrogation system can continuously demodulate in the C-band, improving interrogation precision from an average of 37 pm to better than 1 pm. This research offers a solution for achieving high-performance wavelength interrogation.