Research On Signal To Noise Ratio Enhancement And Brillouin Frequency Shift Extraction Method In BOTDA System

Brillouin Optical Time Domain Analysis(BOTDA),as a distributed optical fiber sensing technology,can measure the temperature and strain of each position of the sensing optical fiber.It has been widely used in the safety monitoring of rail transit,bridge construction,oil and gas pipelines and other fields.How to improve the performance index of BOTDA system has always been a hot research topic at home and abroad.First,this thesis studies the enhancement of signal-to-noise ratio(SNR)in BOTDA system by image processing algorithm,in view of the problems of high complexity and long time consuming of the non-local means algorithm,an algorithm combining non-local means algorithm with discrete wavelet transform is proposed and compared with the non-local means and the discrete wavelet transform algorithms in Brillouin gain spectrum,SNR,spatial resolution,measurement precision and measurement time.The experimental results show that the image processing algorithm has a certain increase in the SNR of the BOTDA system.Even when the average number of data acquisition cards is less,the image processing algorithm can also restore the shape and the basic features of the Brillouin gain spectrum,and the spatial resolution of the system has no loss.Among them,the algorithm combining the non-local means algorithm and the discrete wavelet transform improves the SNR of the system by about 10 dB.While removing the noise in brillouin gain spectrum effectively,the uncertainty of the frequency is shortened from 5MHz to 2.6MHz.Compared with the running time of the non-local means algorithm,the running time of the algorithm is reduced by approximately 80%,which is beneficial to the real-time improvement of the BOTDA system.Then,this thesis studies the extraction method of Brillouin frequency shift in BOTDA system,the artificial neural network algorithm has the problems of slow learning convergence and unstable precision,so the genetic algorithm is used to optimize the initial weights and thresholds of the artificial neural network.The algorithm compares the results obtained by Lorenz curve fitting algorithm,cross-correlation method and artificial neural network algorithm.The experimental results show that the neural network algorithm optimized by genetic algorithm can control the error value of temperature within 1.5 C.The result is closer to the true temperature value,and the accuracy of the measurement is more accurate.When increasing the frequency scanning step to reduce the measurement time,the algorithm can effectively avoid the sacrifice of measurement accuracy and is more conducive to real-time monitoring.Finally,based on the data processing algorithms in BOTDA system,a set of data processing software was written in the LabVIEW and MATLAB development environment to provide a software platform for data processing,allowing the users to select different algorithms for processing according to requirements and actual conditions.And the Lorenz fitting algorithm module adopts the parallel processing method,which saves the running time of the algorithm.