A Practical Risk-Based Model for Early Warning of Seafarer Errors Using Integrated Bayesian Network and SPAR-H

Area of Science:

• Maritime Safety
• Human Factors Engineering
• Risk Analysis

Background:

• Unsafe crew acts (UCAs) are primary causes of maritime accidents, necessitating predictive models for prevention.
• Existing human risk analysis (HRA) models often have a gap between research and industry application.
• Maritime safety research has historically focused on hazard identification and accident analysis, not proactive UCA prediction.

Purpose of the Study:

• To develop a novel prediction model for seafarers' unsafe acts (UCAs) in maritime navigation.
• To bridge the gap between academic HRA research and practical industry needs.
• To provide an early warning system for maritime accidents by predicting UCAs.

Main Methods:

• Integration of the Standardized Plant Analysis Risk-Human Reliability Analysis (SPAR-H) method with Bayesian networks (BNs).
• Identification of performance-shaping factors (PSFs) influencing seafarers' unsafe acts.
• Utilizing BN's inference capabilities for probabilistic risk assessment of UCAs and PSFs, even with limited data.

Main Results:

• Demonstrated the practicability of the BN-SPAR-H model through case studies.
• Successfully quantified the probabilistic risk associated with unsafe crew acts.
• The model effectively evaluates seafarer performance and provides actionable early warnings.

Conclusions:

• The BN-SPAR-H model offers a practical tool for quantitatively predicting UCAs in maritime operations.
• This approach facilitates early warning systems, enabling decision-makers to prevent human errors and maritime accidents.
• The study serves as a foundation for applying HRA research advancements to real-world maritime safety practices.