Integrated fuzzy fault tree-Bayesian network modeling for rollover risk assessment of LNG road tankers

Liquefied natural gas (LNG) road tanker rollover accidents, though infrequent, often lead to catastrophic consequences. Quantitative risk assessment is significantly challenged by the scarcity of probabilistic data specific to these complex, low-frequency events. To address this data limitation and enhance assessment accuracy, this study develops an integrated fuzzy fault tree-Bayesian network (FFT-BN) methodology. Fuzzy set theory is applied, leveraging multi-source general traffic accident statistics and expert judgment, to quantify the occurrence probabilities of basic causal factors under uncertainty. A Bayesian network is then constructed from the fault tree structure to enable comprehensive probabilistic inference. Critical risk factors were rigorously identified using multiple importance measures (ROV, BIM, RRW). The analysis consistently pinpointed poor road alignment and the absence of critical traffic facilities as the two paramount contributors. Crucially, vehicle speed management emerged as the central mitigation mechanism linking these factors; controlling speed effectively counters the destabilizing effects of poor alignment and compensates for the lack of timely hazard perception. The results demonstrate that implementing targeted speed control measures on identified high-risk road sections is essential for reducing the probability of LNG tanker rollovers.