Abstract: In order to provide reference basis for the disaster prevention and mitigation work of rainstorm-flood disaster on mountainous highway, the Bayesian network was used to study the damage to highway structures in the disaster chain under different circumstances. The evolutional process of rainstorm-flood disasters on mountainous highway with chain-rules characteristic was divided into a rainstorm disaster chain and a highway-structure destruction chain. Based on the Bayesian complex network theory, interaction modes between highway structures were analyzed. A Bayesian network inference model was built for rainstorm-flood disaster chain on mountainous highway on the basis of historical data on disasters in Chongqing city and Hunan province. The effectiveness of the model was checked by a case study of a disaster occurred on the section of provincial highway 221 between Wuyang and Lixiqiao town in Suining county Hunan province on June 18, 2015. The result shows that Bayesian network inference under different combined evidence can be implemented by using Bayesian network toolbox with network framework construction, node parameter setting, joint tree inference engine calling and so on, and that prediction result obtained by using the conforms with the actual disaster condition.

Key words: mountainous highway, rainstorm-flood, disaster chain, Bayesian network, inference