Quantitative Analysis of the Cause-Effect Relationship of Incidents Occurred in Tank Farm Based on the Method of Logistic Regression

Abstract

Abstract A large number of incidents occurred in the petrochemical industry due to the continuous accumulation and frequent transfer of dangerous substances. For these historical incident data, a lot of efforts have been made to analyze how and why incidents occurred by use of descriptive statistics, while scarce work was done to in-depth explore the performance characteristics of the causal factors related to different types of incidents. This paper focuses on the relative importance of different causal factors for different types of incidents. A total of 1144 incidents related to tank farms of China during the period 1960–2018 were collected and classified with regard to the intuitive consequences of incidents (fire & explosion, material loss, quality variations, equipment damage and personnel harm) and whether the domino effect was involved (domino incidents and non-domino incidents). The causal factors were classified into five major categories and subdivided into fifteen subcategories. The interaction analysis of each factor with the specific consequence type was performed. The method of logistic regression was used to quantify the relative probability of different causal factors for different types of incidents and to determine which factors have a significant effect on triggering the domino incidents. It is found that human factors and organization/management factors were more common causal factors to lead to different consequences, and the same causal factor has distinct effects on the probability of occurrence for different types of incidents. The results highlight the more critical risk factors for each type of incident and the method can be applied to provide guidance on incident prevention and safety protection.