Association analysis of accident factors in petrochemical storage tank farms

Abstract

In order to identify and clarify the association between the factors leading to accidents in a petrochemical tank area, this study analyzes investigation reports of 212 petrochemical tank farm accidents and combines this with the “association rule” mining and science related to complex networks. The main risk factors are determined and a risk factor data set is constructed; 75 association rules are extracted from the factor data set based on the Apriori algorithm. Then the obtained association rules are used to construct an accident factors network of the petrochemical storage tank area, and the topology characteristics of the network are further analyzed to reveal the importance of factors. Factors with large node degree, betweenness, and clustering coefficients are obtained, such as “violation of operating regulations”, “high concentration of flammable gas in the air”, “lack of experience and professional skills”, etc. These factors play an important role in the formation and development of accidents. The results also show that the accident cause network of the petrochemical storage tank area has a small average shortest path length and a large cluster coefficient, indicating a relatively close connection between the accident factors. The contributions of this study is not only extracting the hidden relationships among contributory factors to tank farm accidents using association analysis, but also revealing which factors are more important for the tank farm safety through the complex network.