Causation Analysis Of Hazardous Chemical Explosion Based On Bayesian Network

Hazardous chemicals explosion cause serious damage. It causes big economic losses and casualties. Recently, the frequent explosion accidents of hazardous chemicals arouse the attention of the whole society. Current literatures have much focused on overall statistics analysis and measures of hazardous chemical accidents. Few studies have focused on the accident causation of single type of hazardous chemical accidents. The paper collects 356 hazardous chemicals explosion accidents in 2011-2015 to analyzed the description statistics and the clustering by SPSS. We find some characteristics of explosion accident of hazardous chemicals. First, hazardous chemicals explosion are frequent and serious, especially in the production process. Second, hazardous chemicals explosion have obvious regularity in the time and place. It occurred frequently in May, June, July and October one year, on 10-12 am and 5-6 PM one day. It also more likely to occur in Guangdong, Jiangsu, Zhejiang, Shandong four provinces. Third, the causes of hazardous chemical explosion are complex and varied.The paper takes 55 cases of Hazardous chemicals explosion in production link as samples to explore the causes of the accidents and their correlation. We get a Bayesian network structure to determine the relationship between the factors that affect the accident occurrence by K2 algorithm. The results reflect three points:first, the production process of hazardous chemicals explosion is a complex system, there are obvious correlation and causality between the factors which cause the accident. Second, the final Bayesian network structure is basically in accordance with accident cross track theory and causal chain theory. Third, the enterprise security management defects play a crucial role in the whole Bayesian network.Then, the paper learn Bayesian network parameters and obtain the conditional probability of each node to receive a quantization result by maximum likelihood estimate. And then we points out two applications of this paper, one is to provide against accidents and to guide enterprises safety management in daily production. The other is to provide reference for Hazardous chemicals explosion investigation.Finally, we infer the causes of Jiangxi Haichen-Honghua Chemical Co. Ltd. explosion accident in 2012 by the joint tree algorithm. It shows that Bayesian network reasoning has an important role in accident investigation and prevention.