Intelligent Failure Localization And Maintenance Of Network Based On Reliability Theory

With the development of network technology,new types of networks continue to emerge.In order to ensure the network operating with high reliability,the network maintenance and failure localization are particularly important.The main contributions of this thesis include:First,we propose one importance indicator based on reliability theory to measure the importance of each link to the reliability of the whole network system.Due to the calculation of network reliability in importance indicator is NP-hard,we design the corresponding Monte Carlo algorithm which can handle the problem of importance indicator calculation of networks of different sizes quickly and efficiently.Further,we adopt one advanced machine learning method named BP neural network to deal with the problem of network failure localization.Based on the dataset created by Weibull distribution,we train the corresponding BP neural network model to predict the failure probability of each link.By sorting the failure probabilities,we choose the link with the largest failure probability as the failed link.At last,by combining the failed link located with corresponding importance indicator,we can determine whether the failed link is worth repairing based on network reliability theory.Simulations indicate the high failure localization accuracy of the proposed approach.Comparative experiments based on graph-based correlation algorithm show that BP neural network can effectively reduce the use of monitoring equipment in network failure localization,thereby reducing related hardware cost and the interference of information transmission,but still guarantee the network operating with high reliability.As the execution of the approach is not restricted to specific network technologies,it can be commonly applied to different types of networks.