Comprehensive risk assessment in ship recycling using a Z-number Bayesian Network approach

Ship recycling involves exposure to a wide range of physical, chemical, and operational hazards, yet integrated, quantitative risk assessment frameworks for the sector remain scarce. This paper proposes a hybrid Z-number–Bayesian Network (BN) framework for occupational health and safety risk assessment in ship recycling operations. Linguistic risk assessments and confidence levels, elicited from seven professionals with direct experience in ship recycling, are converted into prior probabilities using trapezoidal fuzzy numbers and the Onisawa transformation. The Z-number's reliability component discounts each expert's contribution according to their self-reported confidence, so that low-confidence judgements carry proportionally less weight in the final probabilities. The model is validated through axiom-based consistency tests and a parameter sensitivity analysis. The analysis identified asbestos exposure, fires and explosions, and lack of respiratory protection as the highest priority hazards, while causal inference revealed an overall risk probability of 0.3583 for ship recycling; this is a rare concrete result for the sector derived from a Bayesian network framework.Key limitations, including the modest expert panel size and the causal-independence assumption underlying the Noisy-OR structure, are discussed alongside directions for future work.