Integrated berthing risk assessment framework using a combination of machine learning with measured berthing velocity

Port terminals face increasing operational demands that often result in elevated berthing velocities and potential structural damage to berthing facilities. Traditional berthing design criteria primarily rely on static velocity thresholds without considering dynamic interactions between ship characteristics, environmental conditions, and structural response of marine infrastructure. This study presents a comprehensive berthing risk assessment framework that integrates measured berthing velocity with machine hybrid machine learning approaches to enhance port terminal safety and infrastructure protection. Field measurements were conducted over five years (2017–2022) at a major tanker terminal using laser-based Docking Aid Systems, capturing over 900 berthing events across three jetties with varying vessel capacities. The proposed methodology employs K-means clustering to establish risk levels for various influence factors, including weather conditions, ship particulars, and human factors such as pilot experience. Multiple machine learning algorithms were combined to determine optimal weight factors for each risk category, enabling comprehensive risk quantification. The developed model achieved a detection rate of approximately 79 % for high-risk situations, successfully identifying 41 out of 52 cases that exceeded safety criteria. The ensemble approach, combining multiple algorithms with performance-weighted coefficients, demonstrated superior accuracy compared to individual models. This proposed methodology enables quantitative risk evaluation before berthing operations, providing early warning capabilities and objective decision-making support for port operations.