Various methods have been proposed to assess the navigational collision risk of maritime transportation in open sea and restricted waterways. However, there is a gap in the literature concerning an effective method that studies the same risk in port approach manoeuvrings. In this study, we fill this gap with a method stemming from two proposals: First, we present three new parameters (distance, area and speed) to improve the navigational collision risk models applied in port basins. Second, we propose a novel methodology based on machine learning and fuzzy inference to assess the risk of collision in port approach. To support our discussion, we conduct ship-handling simulation experiments with 20 expert pilots and compile a large dataset. Furthermore, we test our methodology in a port approach manoeuvring scenario. Overall, our simulation results show that the proposed method is adequate for weighing both the severity of the port approach manoeuvrings and the relative importance of the involved parameters.

Additional Metadata
Keywords Collision risk evaluation, Data analysis, Machine learning, Maritime safety, Navigation risk parameters, Port approach manoeuvring
Persistent URL dx.doi.org/10.1016/j.oceaneng.2019.106558, hdl.handle.net/1765/120703
Journal Ocean Engineering
Citation
Ozturk, U. (Ulku), Birbil, S.I, & Cicek, K. (Kadir). (2019). Evaluating navigational risk of port approach manoeuvrings with expert assessments and machine learning. Ocean Engineering, 192. doi:10.1016/j.oceaneng.2019.106558