Dynamic resilience modeling of maritime traffic systems: A hybrid HMM-DBN approach for analyzing LNG-Fueled vessel interactions

Abstract

The growing adoption of liquefied natural gas-fueled vessels (LNGFVs) in maritime traffic systems has heightened the need for advanced resilience modeling to mitigate operational risks and improve safety. This research introduces a novel hybrid methodology combining Hidden Markov Models (HMM) and Dynamic Bayesian Networks (DBN) to dynamically assess the resilience of LNGFV operations. Initially, a safety-control feedback structure is developed using Causal Analysis System Theory (CAST), revealing critical factors and their interrelationships that influence navigation safety and resilience. Subsequently, an HMM-based quantification model is designed to address latent-node measurement challenges within the DBN framework, enabling precise inference of complex interactions in maritime traffic systems. Real-world data from an LNGFV collision case in Northeast Australia, including ship-sensor and environmental data, are utilized to reconstruct the accident process and analyze the functional interactions between LNG and vessel operations. Simulation results demonstrate that LNGFV traffic resilience evolves through dynamic interactions among the external environment, the vessel, and LNG, exhibiting a fluctuating temporal pattern. Additionally, the proposed Triple Protection Mechanism shows significant potential in enhancing system resilience. This study provides a comprehensive modeling framework and a new perspective for improving the safety and resilience of maritime transportation, particularly for LNGFV operations. The hybrid HMM-DBN approach offers a robust tool for researchers and practitioners to address the challenges of complex maritime systems.