Research On The Causation Of Railway Accident And Risk Based On Complex Network

In the last few years,the domestic railway system has been developed greatly.With the ever-increasing of railway mileage and occurrence of railway accident,more requirements has been put forward on railway safety.And large amounts of data can obtained easily with the universal application of new technology and equipment in the railway system,thus,it can provides a basis for mining rules of railway accidents.By using complex network theory,this paper established a railway accident causation network model and a dynamic model based on actual railway accident data.The relationships between accident causation factors and dynamic characteristics of the network are analyzed.The main research contents has been done are as follows:(1)Quantify the risk of nodes and communities.This paper built accident causation network model based on the theory of complex network.Effective communities have been found with FN algorithm.Risk of nodes and communities were quantified based on the concept of equivalent damage and risk definition.The analyzation about basic topological parameters of network proved that community detection is effective to analyze the interrelation between accident factors and lay the foundation for recognizing accident causation chains.The quantitative analyzation about risk of nodes and communities shows that strengthening the management of nodes with high-risk or easily-triggered has a great impact on system security,which lays the foundation for further research.(2)Build an accident causation relation network model,and analyze the accident causation chains.This paper built an accident causation relation network model,and recognize the accident causation chains based on the capacity-load model of cascading failure.Based on the analysis of original network,this paper choose change network to analyze the changes of community detection and causation chains recognition.The analyzation about the accident causation chains of networks proved that nodes with higher degree had similar frequency in causation chains and accidents.The comparative analyzation between networks found that the variety of individual node had changed the result obviously.It provided a view that we should monitor the network risk from a systematic perspective,apart from strengthening the supervision of important nodes.(3)Analyze the network with dynamic weights.This paper built dynamic model based on the analyzation about accident causation chains and analyzed the dynamic parameter based on community.It found that there are some nodes failing and removing from the network with risk spreading.There are also some nodes always in the same community within dynamic change.It proved that these nodes are always connected closely.It provides a new idea that strengthening the supervision of these key nodes.