Explosion accidents of molten aluminium in contact with water during aluminium production often occur and may cause injury and death. In this paper, a fuzzy Bayesian network (BN) was employed to probabilistically analyse the explosion accident of molten aluminium in contact with water. A fault tree-Bayesian network (FT-BN) model was established in the cause-effect analysis of the explosion accident, including three processes: electrolysis, molten aluminium transportation and aluminium casting. Fifty-three nodes were proposed in the model to represent the evolution process of the explosion accident from failure causes to consequences. Furthermore, the occurrence probabilities of basic events (BEs) were determined by expert judgement with weighted treatments based on fuzzy theory. By giving certain occurrence probabilities of each BE, the probability of an explosion accident was estimated. Subsequently, importance measures were assessed for each BE, which could reflect the impact on the occurrence of the top event (TE), and the final ranks were provided. The results indicate that using wet ladles and tools, water on the ground, breakage of the tap hole, damage to the casting mould, and leakage of circulating water are five main problems that cause explosion accidents. Safety advice was provided based on the analysis results. This study can help decision makers improve the safety management of aluminium production.

铝液生产过程中铝液与水接触爆炸事故时有发生，可能造成人员伤亡。本文采用模糊贝叶斯网络(BN)对铝液遇水爆炸事故进行概率分析。在爆炸事故的因果分析中建立了故障树-贝叶斯网络（FT-BN）模型，包括电解、铝液输送和铝铸造三个过程。模型中提出了53个节点来表示爆炸事故从失效原因到后果的演变过程。此外，基本事件（BE）的发生概率是通过专家判断和基于模糊理论的加权处理来确定的。通过给出每个 BE 的特定发生概率，估计发生爆炸事故的可能性。随后，对每个 BE 进行重要性度量评估，以反映对顶级事件 (TE) 发生的影响，并提供最终排名。结果表明，使用湿钢包和工具、地面积水、出铁口破损、铸模损坏、循环水泄漏是导致爆炸事故的五个主要问题。根据分析结果提供安全建议。本研究可以帮助决策者改进铝生产的安全管理。地面积水、出水口破裂、铸模损坏、循环水泄漏是导致爆炸事故的五个主要问题。根据分析结果提供安全建议。本研究可以帮助决策者改进铝生产的安全管理。地面积水、出水口破裂、铸模损坏、循环水泄漏是导致爆炸事故的五个主要问题。根据分析结果提供安全建议。本研究可以帮助决策者改进铝生产的安全管理。