Study On Key Fault Node Location And Recovery Technology In Complex Networks

With the rapid development of network technology and hardware devices,the scale of network is gradually expanding.Meanwhile,more and more large-scale failures occur in network,which leads to the losses are also more serious.A large-scale network outage caused by one or more device failures that was observed more and more common in recent years is being called cascade failure by researchers.How to restore the network performance quickly to reduce the losses when the cascade failure occurs in network,that is,the research on cascade failure recovery mechanism,it has very important application value and significance.At present,most research on cascade failure recovery mechanism focuses on the recovery after the entire network is paralyzed,and it's inefficient with low practical value.After studies on a large number of real network cascade failures,it is found that the influence scale of cascade failure is often related to the importance of the initial fault equipment that causes the cascade failure.In consideration of locating the key fault nodes in a fault nodes and restoring it as priority when the cascade failure occurs,so as to control the influence range of cascade failure.In this thesis,with the help of complex network theory,modeling the real network to combine the research of node importance ranking algorithm and cascade failure recovery mechanism.Please refer to the two points as following on the specific research work.(1).In order to locate the key fault nodes accurately in the event of cascade failure,here is to study on the node importance ranking algorithm firstly in this thesis.First of all,the influence coefficient of neighbor nodes is defined by distinguished the influence degree of different neighbor nodes.Then,the influence coefficient of neighbor nodes is combined with the KS value obtained by K-shell decomposition algorithm,which is used as the static information index of nodes.Secondly,the defined the natural connectivity as the dynamic information index of the node to measure the influence of the node on the network performance.Finally,proposed a node importance sorting algorithm based on K-shell decomposition and natural connectivity by combing static information index with dynamic information index.It's been verified through simulation experiments on five kinds of different sizes networks that,with comparison of several traditional node importance sorting algorithms and node importance sorting algorithms that only considering of unilateral information,the algorithm proposed in this thesis is with certain extent in the three aspects of discrimination,correctness and accuracy.(2).In order to restore the network performance as soon as possible when the cascade failure occurs,studies on emergency recovery mechanism in the event of cascade has been stated in this thesis.First of all,modelled the scenario of real network cascade failures more reasonably,and took into account the overload bearing capacity of nodes and the decline of efficiency after overloaded.Then,a weighted probability recovery mechanism for priority recovery of key fault nodes in the process of cascade failure is proposed by used the probability to simulate the recovery sequence of fault nodes and combined the order of node importance ranking.At last,two simulation experiments are carried out.On the one hand,found the influence of adjustable parameters in cascade failure model on recovery effect,which is significant to the construction of the real network.On the other hand,compared the weighted probability recovery mechanism with the traditional equal probability recovery mechanism,which has verified that the weighted probability recovery mechanism could control the influence range of cascade failure better and restore the network performance more quickly.