The Research And Implementation Of BOTDR Signal Noise Suppression And Frequency Shift Extraction Technology

The distributed optical fiber sensing system based on Brillouin Optical Time Domain Reflectometry(BOTDR)has the advantages of high spatial resolution and measurement accuracy,its system structure is also simple and easy to implement.Therefore it has developed into the most important technology in the field of optical fiber sensing.In order to improve the accuracy of the system as much as possible without affecting the signal processing speed and without affecting the spatial resolution,the specific research content of this paper is as follows:(1)This paper studies the effects of traditional algorithms such as mean filtering,median filtering,Gaussian filtering,wavelet transform denoising in BOTDR signal processing,and compares the effects of different convolution kernel sizes on filtering performance through simulation and experiment,and determines the best convolution kernel that does not affect the spatial resolution for wavelet transform denoising,the influence of the number of decomposition layers on the denoising effect is analyzed through experiments,and the best wavelet parameters are obtained.(2)The application of convolutional neural network in BOTDR signal processing is studied,and the denoising effect of neural network is verified through experiments.Finally,a comprehensive comparison of the performance of all the above algorithms in terms of calculation efficiency and filtering effect,the conclusion is that median filtering has obvious advantages in effect and calculation time,therefore it is more suitable for BOTDR signal denoising.(3)The extraction method of Brillouin frequency shift in BOTDR is studied in the dissertation.First,the dissertation introduces secondary fitting and Lorentz fitting.According to the characteristics of actual BOTDR signal,the initial value parameters of Lorentz fitting are improved,and the window treatment is introduced.Simulation and experimental results show that the improved fitting algorithm is better than the previous algorithm in terms of computing time and Brillouin frequency shift extraction accuracy.Finally,the characteristics of the two fitting algorithms are compared,and the results show that the Lorentz fitting has higher measurement accuracy under the premise of ensuring the processing efficiency.(4)According to the above research results,Qt framework of C++ language is used to write the upper computer software of BOTDR system data processing based on windows operating system and realize the visual display of processing results.The initial data processing did not use filtering,and the BFS extraction used quadratic fitting.Compared with the conventional algorithm,the improved algorithm has a maximum accuracy improvement of up to 9MHz.An average accuracy improvement of 2MHz at positions with a large signal-to-noise ratio,and an average accuracy improvement of 4MHz at positions with a low signal-to-noise ratio are obtained.