Bayesian network to analyze leakage risks in LNG ship loading and unloading

Analyzing the leakage risk associated with LNG（Liquefied Natural Gas） ship loading/unloading is a challenging task because of inadequate accident sample data. The BN（Bayesian Network） is introduced to address the issue. The KDE（Kernel Density Estimation） and the max-min hill-climbing algorithm is used for structure learning of BN to reduce the impact of subjective factors. Experiments show that the model can effectively mine the risk nodes leading to LNG leakage accidents, infer the accident cause chain and find the influence degree of each node through reverse reasoning and the posterior probability of the corresponding accident cause nodes. The measures to reduce the risk can be figured out with the help of the model.