Probabilistic modelling for human–machine interface (HMI) failures in automation on the ship bridge using Dempster-Shafer, SPAR-H and Bayesian belief network

Increasing adaptation of automation and digitalised decision-support systems on board the ship bridge has improved operational efficiency but introduced new potential risks related to Human–Machine Interface (HMI) failures. This paper presents a hybrid expert-based probabilistic failure modelling framework integrating the Dempster-Shafer theory (D-S), the Standardized Plant Analysis Risk-Human Reliability Analysis (SPAR-H) and the Bayesian belief network (BBN) approach to perform Probabilistic Risk Analysis (PRA) for Human–Machine Interface (HMI) failures in automation on-board ship bridge. The proposed approach systematically quantifies uncertainties and predicts human–automation interaction failures in maritime navigation. Whilst the D-S evidence theory is capable of handling epistemic uncertainties in expert judgments, the SPAR-H facilitates the incorporation of failure probabilities. The BBN enables dynamic probabilistic risk assessment by capturing causal dependencies between contributing factors. The outcomes of research highlight critical risk pathways and key failures in the HMI system, providing practical insights for maritime safety regulations and operational safety. Also, the findings provide a systematic tool for assessing and mitigating HMI-related failures in ship bridge automation, contributing to the development of resilient and human-centred maritime systems.