Enhancement of Temperature Predictions in Interferometric Sensors by Using KNN Regressions

In this research study, the K-nearest neighbors regression (KNN regression) was used to estimate temperature, based on certain features of an interferometric sensor's spectrum. Typically, these systems face limitations in their measurement range due to various ambiguities induced by physical effects, such as non-linear relationships between wavelength peaks of fringes and temperature. Here it is shown that a KNN regression can be used to overcome the ambiguities and improve the measurement range of this type of sensors. As a demonstration, experimental results are presented showing that by using this technique the measurement range can be extended by a factor of 2 with a RMSE of 0.06°C.