AI for optical sensor

Abstract

Fiber Bragg gratings (FBG) sensors are widely used to measure different parameters including temperature, pressure, electrical-field, and strain due to its unique characteristics and the ability to detect directional changes. The main objective of this project is focused on getting the separated waveforms of each FBG sensor in the overlapped condition as the Spectrum Analyzer can only get the combined waveform in a multiplexing FBG sensor network. The machine learning method Least Square approach and the deep learning method Convolutional Neural Network (CNN) is applied to train the detection model and the central Bragg wavelength of each FBG sensor can be identified from the overlapped spectrum. The result shows that it effectively improves the average testing time and root mean square (RMS) by using the CNN model.