Support vector machine assisted BOTDA utilizing combined Brillouin gain and phase information for enhanced sensing accuracy

Abstract

Benefiting from both Brillouin amplitude and phase spectral responses during Brillouin scattering, a support vector machine (SVM) assisted Brillouin optical time domain analyzer (BOTDA) enabling the improvement of sensing accuracy with only a slight sacrifice of processing speed has been proposed and demonstrated. Only one SVM model, i.e. SVM-(g + p), is required to effectively combine the Brillouin gain and phase information in the training and testing phases, which avoids separate Brillouin gain spectrum (BGS) and Brillouin phase spectrum (BPS) fitting, and hence saves the processing time. Both simulation and experiments using different parameters were conducted to evaluate the improved performance of SVM-(g + p). Compared with the case of using BGS only or BPS only, SVM assisted BOTDA using combined BGS and BPS enhances the accuracy of temperature extraction by about 30% over a wide range of simulation and experiment parameters, only at a slight expense of the processing speed. Although the processing of both gain and phase information takes extra time, SVM-(g + p) assisted BOTDA still has a processing speed 80 times faster than that of using a conventional curve fitting method like Lorentzian curve fitting (LCF). The improved accuracy, together with fast processing speed, is crucial for future high-speed and accurate BOTDA sensors based on both Brillouin gain and phase detection.