Risk analysis using object-oriented Bayesian network: a case study of ammonia leakage of refrigeration system

The increasing complexity of refrigeration systems has introduced major concerns into industrial safety and assets. This paper aims to develop a risk analysis framework for an ammonia refrigeration system using Object-Oriented Bayesian Network (OOBN). The failure causes of ammonia leakage are identified through a historical review of past accidents over a ten-year period and the Fault Tree (FT) is then constructed. Failure probabilities are quantified using objective data sources (plant-specific accident records) for known failure rates and subjective data sources (expert judgments and fuzzy set theory) for uncertain ones. The OOBN model is employed to analyse and evaluate the leakage risk. The results revealed that valve seal failures and flange breakages are critical factors in ammonia leakage, necessitating top priority in risk management. Moreover, the developed framework provides the decision-makers a robust tool for implementing safety measures to prevent and mitigate ammonia leakage incidents effectively.