Quantitative risk analysis on leakage failure of submarine oil and gas pipelines using Bayesian network

Abstract Submarine pipeline is the major transportation way of subsea oil and gas production. Due to the internal and external factor, the failure probability of submarine pipeline is increasing, which could lead to the spill accidents of oil and gas. Efficient risk analysis is vital for preventing and mitigating such potential accident. This paper presents a risk-based accident model to conduct quantitative risk analysis (QRA) for leakage failure of submarine pipeline. Firstly, we employ bow-tie method to model the causal relationship between pipeline leakage and potential accident scenarios. Subsequently, in order to overcome the difficulties of bow-tie in modeling uncertainties and conditional dependency, a Bayesian network model for pipeline leakage is developed through mapping from the former bow-tie. Meanwhile, an object-oriented Bayesian network that has a smaller and more clarified structure is also constructed by modularizing the primary Bayesian network. Eventually, the probability updating is implemented in risk analysis using Bayesian network when a new evidence or observation occurs, and an experience learning from accident precursor data is also conducted through Bayesian approach. The proposed accident model based on Bayesian network can provide a more case-specific and realistic analysis consequence compared to bow-tie method, since it could consider the common cause failures and conditional dependency in accident evolution process of pipeline leakage.