Research On Failure Early Warning And Location In The High-speed Railway Optical Transport Network

With the development of high-speed railway,railway network continues to expand and Optical Transport Network act as the important part of ground infrastructure of high-speed railway,which carries more and more important services to ensure the safe operation of trains.Once failure happens in OTN,a large number of services will be lost.Therefore,it is necessary to study the technologies of failure early warning and location to ensure safe operation of high-speed railway.According to the form of failure,it can be divided into soft failure and hard failure,and different technologies are proposed to deal with different types of failure.The main work of this paper is as follows:(1)Combined with Binary Particle Swarm Optimization(BPSO)algorithm and Support Vector Machine(SVM),the BPSO-SVM algorithm for network performance monitoring is proposed in this paper.The creativity of BPSO-SVM algorithm is as follows.Firstly,BPSO algorithm is used to select features adaptively without setting threshold which can reduce computational complexity of BPSO-SVM algorithm.Secondly,BPSO-SVM algorithm can monitor different kinds of performance parameters at the same time without prior knowledge such as signal modulation formats.Results of simulation show that compared with SVM algorithm,BPSO-SVM algorithm improves the identification accuracy of modulation format by 5.0%,and reduces the prediction error of OSNR and CD coefficients by 8.2% and 14.6% respectively.(2)Aiming at the problems of BPSO algorithm,the Binary Particle Swarm Optimization combined with Feature-based Variance Information(FVI-BPSO)alogrithm is proposed.The creativity of FVI-BPSO algorithm is as follows.Firstly,the variance information of features is encoded into initialization and the process of position information update to improve the convergence speed of BPSO algorithm.Secondly,the updating method of inertia weight is improved by the optimization progress information,which balances the global and local search ability of BPSO algorithm.Then FVI-BPSO-SVM algorithm is proposed by combining FVI-BPSO algorithm with SVM algorithm for network performance monitoring.Results of simulation show that compared with SVM algorithm,FVI-BPSO-SVM algorithm improves the identification accuracy of modulation format by 5.0%,and reduces the prediction error of OSNR and CD coefficients by 7.9% and 20.8% respectively.(3)A method based on the deterioration trend of Asynchronous Amplitude Histograms(AAHs)is proposed to rank the risk level of performance parameters which cause soft failure.By combining the method and proposed performance monitoring method above,the failure early warning mechanism of OTN in high-speed railway is established to complete the monitoring of performance parameters and proactively prevent the soft failure caused by the degradation of performance parameters.The simulation results show that BPSO-SVM algorithm and FVI-BPSO-SVM algorithm are 7.5% and 8.0% higher than SVM algorithm in the prediction accuracy of risk warning level respectively,which means BPSO-SVM algorithm and FVI-BPSO-SVM algorithm have better effect of early warning.(4)A method of monitoring tree construction based on Improved Monitor Location Searching(IMLS)algorithm is proposed to locate single link failure in OTN of high-speed railway.Compared with Monitor Location Searching(MLS)algorithm,the improvement direction of IMLS algorithm is as follows.Firstly,when selecting nodes for monitoring signal forwarding,if there is no node with node degree 2,the node with the largest node degree is preferred for monitoring signal forwarding.Secondly,when the node with the largest node degree is selected for forwarding,if the node degree is greater than 2,then two outgoing links with the largest end node degree of the node are selected for monitoring signal forwarding.The simulation results show that IMLS algorithm is better than MLS algorithm both when p =0.15 and p =0.18.Compared with MLS algorithm,IMLS algorithm reduces(27.7%,33.2%),(15.9%,20.0%),(27.8%,32.7%),(16.2%,19.4%)in terms of number of monitors,costs of monitor,number of monitors in each link and costs of monitor in each link respectively when p =(0.15,0.18)respectively.The monitoring tree constructed by IMLS algorithm is closer to the theoretical optimal monitoring tree.