Machine Learning-Assisted Analysis of Interferometric Optical Sensor Signals

An analysis of interferometric optical temperature sensors based on machine learning methods is presented. Here, the signals are generated by two interferometric systems which were fabricated using stacked layers at the tip of an optical fiber. These systems are theoretically modeled and experimentally tested. For the theoretical model, the summation of rays reflected between interfaces was used. Besides, the thermal properties of the layers were considered to characterize the interferometers’ response to temperature changes. In this work, the sensor that generates each signal was identified by means of the kernel linear discriminant analysis classification method. Later, the kernel ridge regression method was applied to determine the values of the physical parameter measured by each sensor. For both algorithms, the input features were extracted from the output optical spectra of the sensors, which capture changes in wavelength and amplitude. Here, it is shown that for the experimental test dataset, the achieved average accuracy of the classification, the root mean square error, the mean absolute error and the determination coefficient of the regression algorithms were 100%, 0.096oC, 0.080oC and 0.999, respectively.