Research On The Distributed Fiber Disturbance Measurement System Of Phase Sensitive Time Domain Reflection Based On Artificial Intelligence

Distributed optical fiber sensing technology has the advantages of fully distributed sensing,anti-electromagnetic interference,corrosion resistance and ultra-long-distance sensing,and is widely used in border security,pipeline transportation monitoring,railway bridge health monitoring,etc.Among them,the phase-sensitive optical time domain reflectometer(?-OTDR)is very sensitive to disturbances,and can simultaneously realize multi-position and multi-type disturbance simultaneous positioning and disturbance timedomain signal extraction,so the ?-OTDR system is widely used in long distances to monitor.However,in practical applications,due to the large noise in the external environment and the high sensitivity of the system,it is extremely easy to be affected by noise to cause inaccurate disturbance positioning and misjudgment of disturbance types.This article is based on the ?-OTDR system to study,the main work is to improve the ability of system to locate disturbances and to determine the type of disturbance.The thesis mainly includes the following content:(1)Introduce a variety of commonly used distributed optical fiber sensing technologies and research status,learn the working principle of the ?-OTDR system,build a ?-OTDR experimental system in the laboratory.Detailed analysis and testing of the spatial resolution,frequency response range,and frequency response resolution.(2)Conduct multiple experiments in the laboratory to test the positioning ability of the ?-OTDR system.In terms of disturbance positioning,this paper uses the spacefrequency energy spectrum positioning method to realize the disturbance positioning.Compared with the differential positioning method,this method has the advantages of anti-noise interference,accurate positioning and simultaneous measurement of the disturbance frequency.A total of 3 types of experiments were carried out to test the measurement accuracy of the space-frequency energy spectrum positioning method for the disturbance frequency,the positioning accuracy of the disturbance,and the analysis ability of different types of disturbances.The measurement accuracy of the disturbance frequency is 1Hz,and the disturbance positioning error is less than 10 m.The results are in line with expectations.(3)The one-dimensional convolutional neural network(1D-CNN)is applied to the?-OTDR system disturbance type recognition.Use the collected disturbance time domain signal to construct disturbance data set after wavelet denoising,which is divided into 5types,such as no disturbance,sinusoidal 50 Hz disturbance,sinusoidal 100 Hz disturbance,percussion disturbance,and rolling disturbance.After the model parameters are optimized,the average recognition rate of the five types of disturbances is over 98%.(4)The model can directly realize the classification of disturbances by analyzing the time-domain signals after noise reduction.Therefore,the model can be applied to the ?-OTDR system disturbance positioning,but there is a certain false alarm rate.After the optimization of the discrimination algorithm,the model realizes the positioning of pedestrian walking disturbance,and the positioning result is similar to the spacefrequency energy spectrum positioning method.