Mądry Mateusz, Szczupak Bogusław, Śmigielski Mateusz, Matysiak Bartosz
Faculty of Information and Communication Technology, Wroclaw University of Science and Technology, Wybrzeże Wyspiańskiego 27, 50-370 Wrocław, Poland.
Sensors (Basel). 2024 Dec 11;24(24):7913. doi: 10.3390/s24247913.
This paper presents, for the first time to the best of our knowledge, simultaneous temperature and relative humidity (RH) measurement using a machine learning (ML) model in Rayleigh-based Optical Frequency Domain Reflectometry (OFDR). The sensor unit consists of two segments: bare and polyimide-coated fibers, each with different sensitivities to temperature. The polyimide-coated fiber is RH-sensitive, unlike the bare fiber. We propose the ML approach to avoid manual post-processing data and maintain relatively high accuracy of the sensor. The root mean square error (RMSE) values for the 3 cm length of the sensor unit were 0.36 °C and 1.73% RH for temperature and RH, respectively. Furthermore, we investigated the impact of sensor unit lengths and number of data points on RMSE values. This approach eliminates the need for manual data processing, reduces analysis time, and enables accurate, simultaneous measurement of temperature and RH in Rayleigh-based OFDR.
据我们所知，本文首次提出了在基于瑞利的光频域反射仪（OFDR）中使用机器学习（ML）模型同时测量温度和相对湿度（RH）。传感器单元由两段组成：裸光纤和涂覆聚酰亚胺的光纤，每段对温度的敏感度不同。与裸光纤不同，涂覆聚酰亚胺的光纤对RH敏感。我们提出了ML方法，以避免人工进行数据后处理，并保持传感器相对较高的精度。对于3厘米长的传感器单元，温度和RH的均方根误差（RMSE）值分别为0.36°C和1.73%RH。此外，我们研究了传感器单元长度和数据点数对RMSE值的影响。这种方法消除了人工数据处理的需要，减少了分析时间，并能够在基于瑞利的OFDR中准确、同时地测量温度和RH。
