A hybrid method for holistic risk assessment of autonomous navigation control systems

With the increasing deployment of autonomous ships, identifying and prioritizing risks affecting their navigational safety has become a critical issue in maritime operations. The current literature highlights a need for integrated quantitative models that can address both technical failures and human–automation interaction risks simultaneously. This study proposes a multi-method framework that analyzes critical failure modes of the autonomous navigation control system by combining Bayesian Network modelling with expert-confidence-weighted Z-numbers. The model employs Leaky Noisy-OR and RoV (Ratio of Variation) metrics to accurately capture cause–and–effect relationships and identify key risk contributors. The findings show that steering gear and main engine failures have the strongest impact, followed by failures in communication and navigation systems. Meanwhile, context-dependent but low-probability risks are treated as latent hazards within the model. This dual-layered strategy emphasizes the need to strengthen technical reliability while monitoring situational and human–automation risks. Overall, the results are in line with existing studies and provide a novel quantitative perspective based on expert-driven reasoning for risk assessment in autonomous ship operations.