Cascading Failure Assessment Methods And The Applications For Complex Systems

Complex system usually has certain scale and features nonlinear,modularization,hierarchy and so on,and there exist close interactions inside the system and between subsystems,which are different from features of simple system.Along with more powerful functions,the effect of failure dependence on system reliability becomes prominent,which appears system performance degradation and even crash,and results in enormous losses,due to the failure propagation and diffusion through interactions of components or subsystems once the failure happens.With the development of technology,complex system universally appears in human production and life,and serves in a wide range from military to national economy.So it has become a critical topic about how to avoid or reduce cascading failure to further increase the reliability.Cascading failure assessment of complex system is an important component of the topic.Cascading failure assessment includes several aspects such as mechanism analysis of cascading failure,construction of cascading failure model,determination of index of cascading failure.The work has contribution to the research of cascading failure assessment methods and the applications to evaluate the influence of cascading failure on the reliability of complex system.Firstly,cascading failure mechanism on macroscale is analyzed based on brittleness theory.Performance model and brittleness link form of a complex system composed of a controller and several subsystems are given.Based on them,a brittleness evolution model is established,and the dissipativeness and instability are analyzed.After that,Lyapunov exponent spectrum and Kolmogorov entropy are presented as cascading failure index and a criterion of cascading failure is given.Then the failure mechanism is illustrated combining with the analysis and simulation of self-organization process among fixed point,periodic state and chaos.Secondly,identification of cascading failure path and brittleness source is shown.A brittleness Bayesian network model is proposed to realize it.The graph model is constructed by mapping fault tree into Bayesian network and a kind of classification of brittleness rate,brittleness degree and corresponding failure probability is given.Because brittleness information is usually deficient and uncertain,the probabilities of root nodes in different brittleness states such as two states or multi states are obtained based on random sampling method and fuzzy theory,and then causal reasoning and diagnosis inference are exhibited by using junction tree algorithm.After that,the effect of component failure on system failure is analyzed and the method to identify cascading failure path and brittleness source is further given.The proposed method is applied to evaluate cascading failure of ship automatic fire alarm system.Thirdly,assessment of cascading failure due to load redistribution is carried out.The brittleness complete graph of the considered system is shown,in which the subsystems are set as nodes,and the definition of load is also given.After that,an improved cascading failure load-capacity model based on the complete graph is proposed.The effect of initial load of brittleness source,initial load distribution and capacity on cascading failure is illustrated,and cascading failure probability is calculated.The method is applied to evaluate the cascading failure of ship automatic fire alarm system,which is probably caused by load redistribution.Finally,the effect of cascading failure on the reliability of complex repairable system is evaluated.A generalized Markov model of the load redistribution repairable 2/3(G)system is established based on a presented failure rate model.The steady availability is obtained after the analysis of well-posedness and stability of the instantaneous solution by using theory of C0 semi-group.In addition,the state transition diagrams of load-sharing repairable system with ideal capacity,maximum capacity or limit capacity are shown,and generalized Markov models of the system composed of 5 subsystems with different capacities are established as examples.Then,cascading failure effect on reliability of the two kinds of system are analyzed combining with simulations.