Temperature Fiber Sensor Based on 1-D CNN Incorporated Time-Stretch Method for Accurate Detection
Weihao Lin, Yibin Liu, Feihong Yu, Fang Zhao, Shuaiqi Liu, Yuhui Liu, Jinna Chen, Mang I Vai, Perry Ping Shum, Liyang Shao
Abstract
In this article, we propose a novel method for temperature detection based on 1-D CNN and time-stretch method. Differing from traditional optical fiber temperature sensors that are demodulated by an optical spectrum analyzer (OSA) that needs long time to scan and a benchmark spectrum in advance, the demonstrated method can effectively solve the problem in demodulation and improve the detection speed. Besides, mapping the demodulation in the traditional spectral domain to the temporal domain through dispersion optical fiber can greatly reduce the system cost and the demodulation speed can be as high as 50 MHz. Furthermore, the designed peanut-shaped structured Mach–Zehnder interferometer (MZI) is simple to achieve, highly responsive to temperature, and easy to be mass produced at low cost. In the range of 10 °C–18 °C, the temperature sensitivity is up to 1.038 nm/°C, and the detection accuracy is up to 99.07% after using the 1-D CNN algorithm. The proposed sensor system is expected to play a potential role in important scenarios such as temperature monitoring of military weapons and space vehicle system management and control.