Research On Fault Location Method For Complex System Based On Community Structure

With the networking of the real world,the scale of the complex system is expanding constantly.Once the system fails,it will reduce the efficiency or lose the intended function,and even lead to catastrophic accidents.Therefore,it is very important for rapid fault diagnosis when the fault occurs.Fault diagnosis technology is the key method to find out the cause of fault,and fault location is the core of fault diagnosis,which aims to determine the location of fault according to its performance and plays an important role in the realization of efficient maintenance and normal operation.At present,more and more complex structures of complex systems lead to frequent fault propagation,which makes difficult to locate the fault source accurately.The premise of realizing fault location is to establish a model based on system characteristics,which reflects the characteristics of fault occurrence,propagation and amplification.The traditional fault location methods establish the model based on the causality logic relation of fault types and their omens and always neglect the complex network dynamics characteristics of complex systems,especially theinfluence of community structure on fault propagation,which leads to the unsatisfactory localization results.Therefore,this paper proposes a new fault location method for complex systems from the view of community structure.Firstly,in order to get the reasonable community structure,a new community detection algorithm called Mf-Net based on the topology theory of complex network is proposed to adapt to different types of networks and solve the problem of single metric and inefficient execution of traditional algorithms.The proposed algorithm adopts the strategy of multi-attribute fusion,selects three attributes from different views to measure the connection strength between nodes,and introduces the modularity to determine the weights of each attribute objectively,so as to improve the accuracy of community detection.In addition,the proposed algorithm adopts dynamic operator,immune factor and reverse learning mechanism to improve search accuracy and speed of immune network and realize rapid community detection.Secondly,the complex network from the complex system is divided into some communities by the proposed Mf-Net algorithm,and then the more reasonable community structure is obtained.According to the community structure,fault propagation mechanism and the fault propagation ability of each node in the network are analyzed.Referring to the characteristics of complex network dynamics,a dynamic and time-dependent linear threshold fault propagation model is established,and the fault rates of nodes are defined as the main reference forfault source location.Finally,combining depth first traversal and backtracking techniques,the fault area is found,and the propagation path is predicted,and the fault source nodes are located.Finally,this paper selects the experimental datasets and evaluation criterion to verify the rationality and accuracy of the proposed methods.Firstly,the experiments are carried out on three real data sets,and the experimental results are compared with GN,FN,LPA and the algorithm based on node dependence.The results show that the Mf-Net algorithm has high accuracy and good performance.And then the other experiments are carried out on two datasets of complex systems,and the results are compared with other algorithms to verify the timeliness and accuracy of the proposed location method.