Big Data Analytics Methods for Collision and Grounding Risk Analysis in Real Conditions: Framework, Evaluation, and Applications

Collisions and groundings are the most frequent maritime accidents. They often lead to damages resulting in ship flooding and subsequent capsizing, loss of human life or oil spills. Collision and grounding risks can be evaluated qualitatively by expert judgment or quantitatively through the analysis of maritime traffic data. Yet, data-driven studies reflecting real operational conditions using big data remain limited. This research proposes big data analytics methods for the evaluation of collision and grounding risk in real operational conditions. The ship is modelled as a rigid body dynamic system subject to wave-induced motions, traffic flows, and environmental conditions. The notion developed introduces a data-driven framework that aims to support proactive risk management practice, intelligent monitoring and the implementation of associated risk control options in accordance with existing and emerging regulatory requirements. The methods make use of big data streams from the Automatic Identification System (AIS) as well as nowcast hydrometeorological (e.g., wind, wave, current, etc.) and bathymetry data. Results are validated by studying potential collisions and groundings of cruise ships operating in the Gulf of Finland (GoF) during ice-free periods. It is demonstrated that the ideas introduced can assist with (1) the identification of critical collision and grounding scenarios that are not currently accounted for by existing accident databases, (2) the idealization of the fleet at risk in real conditions, (3) the definition of novel collision and grounding risk criteria for ad hoc use within the context of emerging performance-based standards, and (4) the prediction of time-varying ship motion trajectories for recognizing risky situations in advance and proactive risk mitigation. It is confirmed that the probabilistic implementation of big data streams is useful for the proactive evaluation of collision and grounding risks in real operational conditions and Machine Learning (ML) methods can increase a priory our insight into the influence of ship motion trajectories on operational risk. It is therefore concluded that over the medium to long term, the proposed approaches could help develop maritime risk management tools and intelligent decision support systems for ongoing ships in operation. Such methods can help improve ship safety standards and operational practices.