Complex system development risk evolution analysis based on Bayesian learning.

System structures and developing process of large-scale complex weapon are characterized with networks. Research on risk evolution mechanism is useful to control risk and reduce complexity. Based on dynamic data samples presented by system process modeling and simulations, risk evolution networks was refined by Bayesian learning to recognize correlations among nodes with different risk levels. By this way, subjectivity is reduced than risk networks built only by experience. The risk networks from Bayesian learning was further used to implement Bayesian inference and calculate risk posterior probability distributions of risk network nodes under the conditions of system total risk at high level, and then the critical nodes and propogation chains of risk evolution were identified. Finally, comparing with static characteristics based on complex network characteristic indicators, the differences between dynamic and static characteristics of risk networks were under discussion. It is found that the critical nodes and chains of risk evolution are jointly determined by network structure characteristics and dynamic features of risk evolution. [ABSTRACT FROM AUTHOR]