Abstract: For the purpose of quantifying the leakage risk of water-submerged natural gas pipelines under the condition of missing data, a probabilistic risk analysis method based on BN and FST was worked out. First, causation analysis was carried out for leakage failure by using FTA. Then, a corresponding BN model was obtained through mapping. In view of the absence of basic failure probabilities into account, probabilities were elicited from expert knowledge to substitute the missing statistical failure probabilities. To handle the uncertainty caused by the ambiguity and subjectivity of expert knowledge, FST and multi-expert AHP were combined to elicit fuzzy probabilities, as the practical prior probabilities to be inputted into the BN model for quantitative analysis. The proposed method was applied to a water-submerged dual natural gas pipeline for leakage risk analysis. The result shows that the method can be used to characterize and quantify the leakage risk in the absence of data, and that the dynamic risk assessment and the identification of critical failure factors can be realized by utilizing the forward prediction and probability update abilities of BN.

Key words: water submerged natural gas pipelines, risk analysis, fault tree analysis(FTA), Bayesian network (BN), missing data, fuzzy set theory