Aug 2026· Journal of Marine Science and Engineering· 0 citations· 21 references
Abstract
In maritime accident prevention, it is important to identify not only high-risk sea areas but also which accident-types are most likely to occur there. This study combines survey responses from 826 Korea Coast Guard practitioners with 3856 maritime accidents mapped onto an H3 grid over Korean territorial waters during 2021–2023, and proposes a practitioner-informed framework for predicting accident-type-specific risk. The survey showed limited use of quantitative, standardized accident risk criteria but high demand for AI-based prediction and area-level risk analysis. Practitioners’ perceived accident frequency differed substantially from the empirical accident distribution, whereas their prevention priorities aligned more closely with the actual pattern. Accordingly, this study treats the accident-type taxonomy not as a fixed prediction target but as a design variable of the label space for decision support. A two-stage framework first estimates accident occurrence at the H3 grid-time level and then classifies the accident-type conditional on occurrence. Comparing survey-aligned, data-aligned, union, sufficient-sample, and full administrative (7-class) framings under a common training protocol shows that accident-type organization creates trade-offs among field interpretability, coverage, class granularity, and predictive stability. The study thus reframes maritime accident prediction as an accident-type-specific decision-support problem-linking practitioner perception with empirical evidence.
Maritime accident analysis increasingly uses machine learning to support safety management, but many existing studies focus on single-output prediction, such as accident-occurrence probability, severity class, near-miss frequency, or one specific consequence. This study proposes a data-driven decision-support framework based on a Multi-Input Multi-Output Artificial Neural Network (MIMO-ANN) for the simultaneous prediction of multiple maritime accident consequences. A dataset of 582 recorded accident cases is constructed by integrating SafePASS project records with consequence, severity, and structural-damage information from the literature. The dataset includes 15 input variables covering ship characteristics, operational context, environmental conditions, accident type, and geographical zone and 15 consequence outputs covering structural damage, casualties, emergency-response indicators, total loss, and secondary consequence/escalation mechanisms. The ANN is trained using the Scaled Conjugate Gradient (SCG) algorithm and evaluated under different network configurations and data-partitioning strategies. The best-performing model uses 30 hidden neurons with a 60/20/20 split, achieving a correlation coefficient (R) equal to 0.9249 and a mean squared error (MSE) equal to 0.0240 for testing, and a R equal to 0.9278 and a MSE equal to 0.0231 for validation. Ten-fold cross-validation further confirms internal predictive stability, with mean testing R equal to 0.8803 ± 0.0827 and MSE equal to 0.0445 ± 0.0478. Permutation-based sensitivity analysis shows that accident type, zone, flag, natural light, environment, and visibility are key drivers of predicted consequences, whereas vessel-specific parameters have a secondary, context-dependent influence. The framework should be interpreted as predicting the relative likelihood, severity, or magnitude of accident consequences in recorded or scenario-defined accident cases, not the probability of accident occurrence. Future work should address dataset imbalance, include near-miss and nonserious records, incorporate richer AIS and metocean data, integrate exposure data, and validate the framework using independent accident datasets.
Mina Tadros, E. Boulougouris, Evangelos Stefanou et al.· The Scientist· 0 citations
With limited public investment, phased investment strategies (2-lane or limited 4-lane configurations) are widely adopted in Vietnam. However, this model poses safety risks due to the lack of median barriers and continuous emergency lanes. This study analyzes and quantifies factors influencing traffic accident severity using 170 accident records from 2018-2025. Integrating descriptive statistics and hierarchical logistic regression, the research identifies heavy trucks as the primary vehicles in over 71% of collisions. The model reveals that accidents on weekdays pose a severity risk 2.445 times higher than on weekends. Notably, "vehicle overturn or run-off-road" accidents - a direct result of inadequate emergency lanes - increase the probability of severe outcomes by 4.178 times. These findings provide a scientific basis for prioritizing infrastructure upgrades and optimizing traffic management on phased expressways.
X. Le, T. Chu, V. Phan· Journal of Transportation Sc...· 0 citations
Maritime administrations frequently implement short-term intensive safety actions to mitigate collision risk in mixed commercial–fishing traffic waters. However, empirical evidence on their effectiveness remains limited because maritime accidents are rare and behavior-level risk indicators are not routinely incorporated into policy evaluation. This study develops an AIS-based evaluation framework that uses monthly near-miss counts as a behavior-level proxy of navigational risk and combines this proxy with a difference-in-differences (DID) design to assess a maritime special safety action in Ningbo–Zhoushan waters. Using large-scale AIS trajectory data, near-miss events are identified based on DCPA and TCPA criteria and then aggregated to the sea area–month level. The analysis covers 9 sea areas from January to December 2023 (108 observations). In the baseline two-way fixed-effects specification, the coefficient on Treat × Post is positive but statistically insignificant (β = 0.4094, SE = 0.2581), indicating that the intervention did not produce robust evidence of a reduction in the AIS-based near-miss indicator. Event-study estimates likewise show no statistically significant persistent dynamic treatment effect within the observation window. These findings suggest that, under the proxy measure and identification strategy used in this study, the special safety action did not generate a clearly identifiable reduction in near-miss counts in treated waters. Methodologically, the study demonstrates the practical value of combining AIS-derived behavioral indicators with quasi-experimental policy evaluation. At the same time, the results should be interpreted with caution, because near-miss counts are structurally related to traffic intensity and traffic composition, and the policy period overlaps with the seasonal fishing moratorium. The proposed framework nevertheless offers a useful basis for evidence-based evaluation of non-engineering maritime safety interventions in complex mixed-traffic environments.
Tunan Xu, Yifei Mao, M. Grifoll et al.· Frontiers in Marine Science· 0 citations
Ship traffic through the Turkish Straits occurs in a highly constrained navigational environment, where inaccurate prepassage reports can significantly increase the risk of maritime accidents. Vessels are required to submit Sailing Plan declarations (SP-1 and SP-2) before entry, but these reports often fail to reflect the actual technical and operational condition of the ships. Commercial pressure, time constraints, and intentional misreporting can create discrepancies between declared and actual readiness, raising the likelihood of collisions, groundings, or loss of maneuverability in congested waters. This study develops a reliability-centered framework to assess accident risks associated with reporting, contributing to the literature on human reliability and uncertainty management in maritime traffic. Risk criteria were identified by reviewing regulatory requirements, accident records, and relevant literature, and refined through expert input from Vessel Traffic Services (VTS), pilotage, Port State Control (PSC), and ship operations. The framework combines the Fine–Kinney method with an Intuitionistic Fuzzy TODIM (an acronym in Portuguese for interactive and multicriteria decision-making). The results indicate that undisclosed propulsion deficiencies, steering problems, and nontransparent withdrawal from passage queues are the most critical accident precursors. These findings highlight the importance of reliable reporting and provide practical and transferable guidance for reducing the risks of maritime accidents and improving risk management for vessel traffic. In particular, the results support the prioritization of propulsion and steering system checks, as well as the closer scrutiny of vessels withdrawing from passage queues, offering actionable insights for VTS operators, PSC authorities and marine pilots.
A. Kuzu· Transportation Research Reco...· 0 citations
Maritime autonomous surface ships (MASSs) are reshaping the organization of navigation, ship operation, remote control and maritime supervision. However, the transition from crewed navigation to autonomy also changes the structure of safety risk. Traditional ship risk assessment approaches rely heavily on historical accident records and crew-centered operational assumptions, whereas MASS operations involve coupled risks arising from perception systems, autonomous decision-making, communication links, cybersecurity, remote control centers, environmental uncertainty and management readiness. To address the scarcity of operational accident data and the need for a structured safety evaluation method, this paper develops a Formal Safety Assessment (FSA)-based risk evaluation framework for MASS operations. A hierarchical indicator system is established from five dimensions: ship machinery, human factors, environmental factors, information technology and management. A frequency-severity risk criterion is then constructed by defining the Frequency Index (FI), Severity Index (SI) and Risk Index (RI), and by introducing the ALARP principle to classify unacceptable, tolerable and broadly acceptable risk regions. On this basis, an integrated fuzzy analytic hierarchy process is proposed to determine factor weights, transform expert judgements into membership degrees, and calculate comprehensive risk scores. A case study using 50 expert questionnaires shows that the overall risk score of MASS operation is 5.64, located in the ALARP region. Among the first-level indicators, environmental factors, information technology factors and ship machinery factors present relatively high risk levels, with scores of 6.82, 6.60 and 6.25, respectively. At the secondary-indicator level, intelligent navigation system, weather conditions, hydrometeorological conditions, communication capability, equipment and systems, navigation decision-making, and environmental perception are identified as high-intensity risk indicators. Further contribution decomposition reveals that environmental perception, routine ship management, navigation decision-making, and communication capability contribute most substantially to the overall risk profile due to their higher systemic importance. The proposed framework provides an interpretable approach for MASS safety assessment and risk-control prioritization under limited operational data availability.
This paper presents a modular decision support system that infers the primary location of the user or the reported incident and a situation-aware risk level from multi-turn Turkish disaster dialogues between a help-seeking user and an AI-supported emergency assistant. The assistant guides the user with follow-up questions about health status, number of affected people, structural damage, environmental hazards, and known nearby landmarks to complete missing information. The system manages the dialogue with a finite state machine, determines the location by linking user cues to a local GeoJSON point-of-interest database and by landmark verification, and produces explainable risk scores with Multi-Criteria Decision Analysis. The key novelty is treating landmarks as an evidence layer that verifies the current location hypothesis through proximity and clustering instead of directly replacing candidates based on a landmark signal. In a ten-scenario pilot evaluation, accuracy, end-to-end latency, and token usage are reported for three configurations.
Eren Varlıker, Yusuf Sinan Özmen, Selim Balcisoy· Signal Processing and Commun...· 0 citations