Research On Fault Location Based On Attention Mechanism In Optical Networks

With the rapid development of the fifth-generation mobile network,mobile data is showing an explosive growth trend,the number of network devices is increasing,and various new services and application scenarios are emerging,which puts higher requirements on the reliability of the network.Fault location is an important part of optical network operation and maintenance.How to mine network monitoring data information to achieve highly accurate and precise fault location of network devices,ensure service transmission quality and reduce fault recovery cost is a key issue facing network operators nowadays.In this thesis,we focus on the optical network fault location strategy based on attention mechanism,and design and implement an intelligent network control platform based on machine learning,and the main research work and innovation points are as follows:(1)According to the location of monitoring data in the network,this paper divides it into link monitoring data and node monitoring data,and models them respectively.In our work,the power data in the monitoring data will be used to locate the fault.Specifically,link power data is characterized as link power sequence,and node power data is characterized as node power matrix.(2)This thesis presents a fault location strategy based on an attention mechanism,whose input is modeled network power data and output is a set of fault probability values of network devices.The strategy consists of a sequence attention mechanism,a channel attention mechanism,a graph attention mechanism,and a fully connected neural network.Among them,the sequence attention mechanism,channel attention mechanism and graph attention mechanism are used to characterize the link data,node data and network data respectively,while the fully connected neural network analyzes the correlation among the monitoring data and completes the fault location decision based on them.The simulation results show that the introduction of the attention mechanism can filter out the key data and improve the accuracy of fault location.(3)This thesis integrates the technologies of software-defined networking and artificial intelligence,and builds an intelligent optical transmission network control platform based on the completion of the functional module development work and the existing network equipment.The platform enables flexible scheduling of network traffic and automatic location of network faults with the help of machine learning algorithms.At the same time,this thesis simulates network failure scenarios on the platform and designs experimental solutions to verify the feasibility and effectiveness of the proposed strategy.