Brillouin Scattering Spectrum Feature Extraction Of BOTDA Sensing System

Since the research of distributed fiber optic sensing system in the 20 th century,this system has become an indispensable means to detect the health condition of large civil buildings,such as submarine cables,oil pipelines,tunnels and railways,etc.Among them,the Brillouin optical time domain analysis(BOTDA)sensing system has become a research hotspot for scholars at home and abroad due to its advantages such as long detection distance,high accuracy in identifying environmental influence factors,strong signals and high spatial resolution.This paper focuses on the extraction of single-peak and multi-peak stimulated Brillouin scattering(SBS)spectral features,and the simultaneous measurement of temperature and strain based on the dual frequency shift characteristics of the SBS spectrum.For traditional single-peak Brillouin scattering spectrum fitting methods exist the problems of depending on initial value and easily trapping in local minima.First of all,according to the sensing optical fiber in the occurrence of stimulated Brillouin scattering effect,build a Pseudo-Voigt linear SBS single-peak spectrum theory model,and then put forward single-peak SBS spectrum fitting method based on differential evolution algorithm(DE)optimized general regression neural network(GRNN).Through simulation experiments,GRNN and DE-GRNN algorithm are used to fit single-peak SBS spectrum under different signal-to-noise ratio and different line width respectively,and the fitting results verified that the method can achieve high precision of SBS spectrum feature extraction.In order to solve the multi-peak SBS spectrum feature extraction due to factors such as fiber doping,special fibers and simultaneous measurement of temperature and strain,as well as the tedious segmentation process of traditional multi-peak fitting method and some singlepeak fitting methods are not applicable the problem of multi-peak SBS spectrum fitting,firstly,the theory model of multi-peak SBS spectrum is constructed and the simulation of three-peak SBS spectrum is realized without loss of generality.Then a multi-peak SBS spectrum fitting method based on wavelet packet de-noising and genetic algorithm(GA)optimization of GRNN was proposed.The simulation results show that this method can effectively solve the problem of multi-peak SBS spectrum feature extraction.At the same time,simulation experiments comparing with GA-optimized Back-Propagation(BP)algorithm and DE-GRNN algorithm verified the superiority of this method in a low signal-to-noise ratio environment.Finally,this paper studied the simultaneous measurement of temperature and strain based on dual frequency shift characteristics of the SBS spectrum in the BOTDA sensing system,and used the actual measured SBS spectrum data to simulate the proposed fitting algorithm and verify the results.First,the design of the experimental system and the application principle of the dual-frequency shift measurement method are given,and then the calibration of system temperature and strain coefficient is realized by fitting the measured data.Finally,the wavelet packet de-noising with GA-GRNN algorithm was used to fit the bimodal SBS spectrum discrete data obtained from the experiment of simultaneous temperature and strain measurement.The fitting performance of the algorithm in the actual detection was analyzed,and the feasibility of the method was verified.