Fault Detection Of Fiber Link Based On Improved L₁ Trend Filter Algorithm

With the rapid spread of optical networks,the maintenance and management of optical fiber link is becoming more and more arduous.In order to ensure that the network is working efficiently and safely,the realization of automatic faults detection in optical fiber links is imminent.Optical Time Domain Reflectometer(OTDR)is a commonly used optical fiber detection device,but its accurate,efficient,and automatic fault identification has always been a difficult study field to overcome.For this reason,we presents an accurate and efficient algorithm,improved L1 trend filter algorithm,to solve the automatic faults detection problem in OTDR profile.The algorithm firstly uses the sequential test to detect the frequency domain and time do-main at the same time,and identifies the starting point and the end point of the fiber.Based on the identification results of previous step,the blind area at the beginning and the noise area after end point of fiber are removed,which saves a lot of time and mem-ory for the operation of the algorithm.Then,the signal between the start point and the end point is detected using the L1 trend filter,and the faults of this segment will be initially identified.Then,the least squares method is used to fit the data between these adjacent fault points,and the parameter V is used to represent the change of intercept and slope of fitted lines which are before and after of the fault point.By comparing the parameter V with the adaptive threshold ?,whether the initially identified fault point is a true fault point or false is determined.If it is a real fault point,it will be reserved.If it is a pseudo fault point,it will be eliminated.In this paper,we detected 14 representative real-measured fiber links and located faults point accurately.Except for two pseudo faults on link 2 and one pseudo fault point on link 3 and link 6,the number of other link fault points is consistent with the actual situation.In order to increase the credibility of the experiment,this thesis com-pares the improved L1 trend filter algorithm with the ordinary L1 trend filter algorithm.The former one has higher detection accuracy than that of the latter in detection of link 8.Meanwhile,improved L1 trend filter algorithm costs less time than ordinary L1 trend filter algorithm.This thesis also explores the relationship between running time of al-gorithm and the number of data.When the number of data is gradually increasing,the running time of the improved L1 trend filter algorithm does not increase significantly.When the number of data is between 200 and 11,000,the whole process of faults de-tection can be completed in 6 seconds.The experiment results show that,with a certain extent accommodating of data noise,the improved L1 trend filter algorithm detects the fault points of fiber links efficiently and accurately.