Bayesian estimation of ammonia leak frequency for risk assessment of ammonia-fueled vessels

• Limited data available on ammonia leakage risks in maritime fuel applications. • Bayesian model developed to estimate leak frequencies for ammonia-fueled vessels. • Model updated using LNG vessel data and Japanese onshore ammonia leak records. • Leak frequencies for ammonia-fueled vessels estimated by component and leak size. • Findings support improved risk assessments for ammonia-fueled vessels. Risk assessments for ammonia-fueled vessels are crucial due to ammonia's unique characteristics and leakage risks. Limited data on component leak frequency (LF) for risk assessment in the maritime sector exist due to the scarcity of ammonia-related accidents. The LFs derived from hydrocarbon fuels have been used as a substitute for ammonia risk assessment, but accurate LF estimation is essential for reliable risk assessments in the maritime sector. We aimed to estimate the component LF of ammonia-fueled vessels, considering ammonia's characteristics and leakage accidents for a more precise risk assessment. We developed a Bayesian model using prior distributions based on generic component LFs and LNG-fueled vessel LFs to represent likelihood. The model was then updated with ammonia leak accident data from Japanese onshore facilities reported under the High Pressure Gas Safety Act in Japan. We estimated the LFs of each component and leak size for ammonia-fueled vessels. The sensitivity analysis confirmed that the estimated LFs for ammonia-fueled vessels were lower than those for conventionally fueled vessels, consistent with previous studies. Our findings suggest that the estimated component LFs are applicable for risk assessments of ammonia-fueled vessels. This study validated the estimated LFs using accident data from Japanese onshore facilities. Future research should focus on collecting more data as ammonia-fueled vessels become operational and refining the model with new insights.