Application of Bayesian network and non-additivity principle in analyzing dam break risk

Abstract

The dam break risk analysis, especially for that under the composite action of multiple risk sources, is highly important for implementing dam safety management. To study the complex system problem, this research taken dams in the Dadu river basin as the study case, and combined Bayesian network (BN) and the non-additivity principle to analyze the dam break risk under prominent dam break causes of flood and earthquake. During the research process, literature analysis, historical data, and expert knowledge have been taken firstly to identify the related factors affecting the dam break risk. Furthermore, determine variable nodes of the dam break risk, build a Bayesian network directed acyclic topology model about the dam break risk according to the causal relationship between risk factors, and construct the prior probabilities and conditional probabilities of corresponding nodes. Finally, calculate the risk probabilities under the condition of the same input variables and different variables by using based on Bayesian network reasoning. The results illustrate that the usage of BN in analyzing non-additivity is applicable, and non-additive effects exist when multiple risk factors impact a dam break simultaneously.