선박 통항 분석을 통한 항로 상 위험구역 식별에 대한 연구

다국어 초록 (Multilingual Abstract)

As the maritime traffic environment evolves and the utilization of AIS(Automatic Identification System) data becomes widespread, it is becoming increasingly critical to flexibly integrate complex variables from the actual navigational environment into fairway design and management. This study aims to analyze AIS-based vessel traffic data in the Mokpo fairway to quantitatively identify the impact of interference—specifically crossing encounters with vessels using convention routes—on actual operational risk, and to propose measures to ensure fairway safety based on these findings.
The research proceeded in the following stages: First, AIS data were collected and preprocessed via outlier detection and missing value interpolation. Next, traffic density heatmaps were generated, and traffic pattern clustering based on DBSCAN was performed to analyze the actual traffic characteristics of the Mokpo fairway. Subsequently, quantitative risk was calculated for each grid by applying the concept of causation probability() from the IWRAP model, considering specific encounter situations.
The analysis results revealed that crossing encounters with vessels not adhering to designated routes were frequently identified in the vicinity of islands. These actual traffic characteristics were confirmed to be key factors aggravating collision risk, adding to the basic encounter situations (head-on and overtaking) considered during the design stage. This implies that securing additional safety margins, such as channel widening, is necessary for areas with frequent crossing encounters.
Based on these findings, this study proposes incorporating the 'degree of interference from convention routes' as a correction parameter in future fairway design. Furthermore, it provides empirical evidence for establishing an efficient, data-driven fairway management system, such as the implementation of 'Dynamic Sector Operation,' to flexibly respond to time periods and zones where risk is concentrated.