Research On Performance Optimization Of Optical Fiber Vibration Sensing System Based On Co-processing With FPGA,CPU And GPU

Distributed optical fiber sensor uses communication optical fiber as sensing and transmission unit,which can realize distributed sensing and measurement of vibration,temperature,strain and other physical quantities.Among them,the distributed optical fiber vibration sensing system based on the Phase-sensitive Optical Time Domain Reflectometer(?-OTDR)can realize the distributed positioning and reduction of vibration s ignals.Because of its advantages of high sensitivity,long detection distance and fast response speed,it is widely used in many fields such as perimeter security,oil and gas pipeline,building structure health monitoring and smart grid.The coherent detection ?-OTDR system collects the coherent Rayleigh scattering signal,and obtains the intens ity and phase information of the sensing signal by us ing the orthogonal demodulation algorithm,so as to realize the rapid location and phase restoration of external disturbance events.At present,the real-time performance,signal-to-noise ratio and phase recovery degree of ?-OTDR system are required in the practical application field.Therefore,this paper takes ?-OTDR system as the research object,based on the in-depth study and analysis of the system sensing principle and signal demodulation algorithm theory,proposes a performance improvement scheme based on algorithm optimization and acceleration,introduces the idea of parallel computing to reduce the measurement time of the system,uses the image processing method to optimize the signal-to-noise ratio(SNR)of the vibration positioning curve,and uses the empirical mode decomposition algorithm to eliminate the phase low frequency drift.The research content mainly includes the following four aspects:Firstly,the vibration sensing principle and signal demodulation scheme of ?-OTDR system are studied.Based on the theory of Rayleigh backward scattering,the principle of optical phase modulation caused by external vibration of ?-OTDR system is analyzed,and then a mathematical model is built to describe the sensing principle,and the optical path structure of ?-OTDR system is built.Then,the demodulation principle of the sensing signal is analyzed,and the process of amplitude demodulation and phase demodulation of the quadrature demodulation is discussed.Finally,the key performance parameters,such as real-time performance,signal-to-noise ratio,are discussed and analyzed.Secondly,based on Field Programmable Gate Array(FPGA),Central Processing Unit(CPU)and Graphic Processing Unit(GPU)technology,the idea of parallel processing is introduced to reduce the measurement time of the system.Based on the analys is of the parallel processing requirements of the sensing system,a collaborative parallel processing scheme is designed for the data acquisition,transmission and processing of the ?-OTDR system.The data acquisition system based on FPGA is constructed to realize the control of the lower computer sensing system,parallel data acquisition and transmission,and then improve the acquisition efficiency;a data flow control system based on CPU,which realizes data transmission and processing;the data processing system based on GPU realizes the multi-thread parallel demodulation of the sensing data,thus greatly improving the real-time performance of the system.Thirdly,the amplitude demodulation process of ?-OTDR system is analyzed,and the research on measurement time optimization and SNR improvement is carried out.For measurement time optimization,this paper designs a parallel demodulation scheme based on GPU.The difference algorithm and FFT demodulation algorithm are verified by experiments.The results show that the parallel processing method is 93 times faster than the single thread processing method for the difference algorithm under the sens ing data volume of 5000*21411,and the parallel processing of the FFT reduces the data operation time by about 96 times under the sensing data volume of 8192*20000.In order to improve the SNR,this paper designs a two-dimensional digital image processing technology to achieve intens ity information noise reduction.By using 2D Gaussian filter and Sobel edge detection algorithm,the SNR of vibration location curve is 27.1 d B.Compared with the difference algorithm and FFT algorithm,the SNR of amplitude signal after image processing is increased by 11.54 d B and 14.36 d B respectively.Fourthly,the phase demodulation process of ?-OTDR system is analyzed,and the low-frequency drift noise elimination in phase recovery is studied.Starting from the factors that affect the phase recovery,the generation mechanism and drift characteristics of phase drift are analyzed theoretically.By analyzing the principle,it is proved that the phase drift is mainly composed of the laser frequency drift and the phase accumulated noise.According to its slow drift characteristics,this paper proposes an empirical mode decomposition(EMD)algorithm to eliminate phase drift noise.The experimental results show that EMD can effectively eliminate the phase drift and achieve accurate vibration reduction.Finally,the low frequency vibration signal of 0.3 Hz can be restored,and the SNR is 64.44 dB.