Research On Failures Of Complex System Based On Complex Networks

Many professional terms including "internet of things", "big data", "cloud computing", "internet plus", presented everywhere, which show today's society is undergoing profound changed, indicating that complex systems of real-life social networks, transportation networks, communication networks linked more closely. Complex systems can do things of intelligence and adaptive behaviors replying on local information, the integrity of the structure and function are related to human life. For example, public transports used by thousands of residents and the normal power gird by hundreds of millions of people. Therefore, the study of complex systems failures has obtained of great concern of academia and industry. It has become an urgent problem to be solved about how to better understand the mechanisms of complex systems failure occurs and how to better prevent failures.Complex networks are abstracted of complex systems, which are convenient to research and analysis by a common perspective, because of ignoring differences in real-world. Traditional evaluations about static failures of complex systems consider giant component, which lacks fragmentation-depth study of the overall network. At the same time, the assessment of many dynamic failures considers the macro perspective with single nodes or macro perspective with the entire network, which are short of meso-perspective with community structure of node.The paper builds failures models for connected subgraph and community structures of complex networks to explore how to better assess the failures of complex systems on the study of the static failures and dynamic failures of existing complex systems, considering the number of connected subgraph and community structure. Considering the other outside connected subgraphs subjected to the maximal connected subgraph still communicating, we build a static failures model for the number of connected subgraphs to assess failure network fragmentation by the number of connected subgraphs. The dynamic failures model for community structure is to be built for the disadvantages of conventional linear models and nonlinear models, which combines community structure to compute the capacity of node. The paper detailed compares and analyses different various factors on the failures results through by simulations of different sets of data. Compared to traditional failures studies, the models of connected subgraph and community structure are better conform to the real failures conditions, the relevant conclusions of which can be better assess and prevent failures of complex systems.