Optical Vernier effect-based demodulation system for fiber-optic sensors: an AWG and a neural network-based perspective

Accurate demodulation of fiber-optic sensors is crucial for real-world engineering applications in monitoring and control. This paper presents a method that integrates neural networks with arrayed waveguide gratings (AWGs) for the demodulation of fiber-optic sensors based on the Vernier effect and a novel, to our knowledge, Fabry-Pérot (FP) strain sensor structure. Conventional demodulation techniques exhibit limited generalization capabilities, whereas neural networks can establish complex nonlinear mappings. AWG-based wavelength-division multiplexing (WDM) enables the encoding of spectral shifts induced by external environmental variations into distinct transmitted light intensities in each channel. A neural network is employed to model the intricate nonlinear relationship between transmitted light intensity and the actual wavelength, thereby achieving absolute wavelength interrogation. This approach does not rely on extensive datasets or artificially generated pseudo-data. Experimental results demonstrate that the proposed demodulation method offers low energy consumption, high efficiency, and enhanced sensor sensitivity.