A Study of Safety Culture Assessment Framework for Process Industries and Its Application to a Bayesian Belief Network Analysis
MetadataShow full item record
Investigations of major catastrophes in process industries have revealed that deficiency of good safety culture is one of the underlying causes of such disasters. Not only has safety culture been recognized as a root cause, but also it is increasingly accepted as an influential factor in a risk analysis and considered as a legal requirement. Most of current quantitative risk analyses (QRA) rely on technical factors but more and more effort is being made for the incorporation of human and organizational factors (HOFs). Especially, safety culture largely represents an organizational attitude towards safety. Thus, how to measure safety culture in more effective manners and how to utilize such assessment data in a QRA are chosen as major objectives of this research. For the measurement of safety culture, this study suggests an approach that assesses values and assumptions by looking through artifacts, e.g., management level and employee’s behavior. Such approach employs following two methods: a matrix structure composed of safety culture dimensions, and grading schemes that provide different levels of safety practices. Using such an approach and suggested methods, a safety culture assessment questionnaire is developed as a results. For the incorporation of such safety culture data into a risk analysis, this study employs a risk model based on Hybrid Causal Logic (HCL) and a Bayesian Belief Network (BBN) to represent cause and effect relationships among variables. Mock-up safety culture data is generated for this analysis. Findings from investigation of Universal Form Clamp incident (2006) are used to establish a case scenario upon which a fault tree and an event tree are constructed. To make a transition from qualitative knowledge about safety culture to quantitative probability data, some of the safety culture dimensions are selected as Risk Influencing Factors (RIFs), while Safety Culture Influencing Factors (SCIFs) are developed and introduced in this work. Using the established BBN, prior generic probability data are updated with newly obtained evidences such as mock-up safety culture assessment data. In addition, several analyses, e.g., predictive and diagnostic reasoning are conducted to determine how a change in safety culture affects the probabilities of safety-related events and also to identify which safety culture aspects need improvement.
Son, Changwon (2016). A Study of Safety Culture Assessment Framework for Process Industries and Its Application to a Bayesian Belief Network Analysis. Master's thesis, Texas A & M University. Available electronically from