Simultaneous Demodulation of Salinity and Temperature Assisted by Deep Learning Approach Utilizing Tilted Fiber Bragg Grating and Fabry–Perot-Based Sensor
Ziqi Liu, Zhengyong Liu, Yongchang Mei, Peng Fei Hu, Zhaohui Li
Abstract
A new method of demodulation of salinity and temperature (ST) two-parameter sensor based on tilted fiber Bragg grating (TFBG) and Fabry–Perot structure using a double-branch convolutional neural network (CNN) with dense connected blocks is proposed. A total of 4344 spectral samples were collected when the sensor was immersed in sodium chloride solution with different salinities and temperatures, in an attempt to train and test the CNN model. Experimental results showed that the well-trained CNN model could realize real-time and high-precision demodulation of the ST dual-parameter sensor. The mean absolute errors (MAEs) of ST are 0.207 ‰ and 0.11 °C, respectively, and the determination coefficients are 85.51% and 99.96%, respectively. The predicted ST value of sodium chloride solution by the proposed characterization approach is consistent with the truth value, which shows great potential applications of such a sensor for high-precision double-parameter measurement.