Near-IR Fiber-Optic Temperature Sensor

A fiber-optic temperature sensor based on the perturbations of near-IR water bands has been developed. These fiber-optic sensors are very simple and readily fabricated. Models for expressing temperature can be developed by linear regression (LR) of the absorbance at one selected wavenumber, by multilinear regression (MLR) of the absorbances at several selected wavenumbers, or by principal component regression (PCR) using entire spectra. The standard errors of prediction for temperature are 0.53 to 1.64°C for the LR model, 0.22 to 0.85°C for the MLR model, and 0.16 to 0.32°C for the PCR model over a temperature range of 5 to 85°C. Potentially, these fiber-optic sensors can be used in the remote sensing of temperature and in hostile electrical environments.