Iot and machine learning based platform for port risk managementapplication to ship movements and wave overtopping

Resumen

Spain is the European Union country with the longest coastline (8000 km) and the closest to the axis of one of the world's major maritime routes. Its geographical location positions Spain as a strategic element in international shipping and as a logistics platform in Southern Europe. The Spanish Port System includes, as of 2022, 46 ports of general interest. The importance of ports in the Spanish economy is reflected in the state port system's activity contributing to nearly 20% of the transport sector's GDP, accounting for 1.1% of the Spanish GDP. Due to its importance, events that could disrupt the normal operations of a port and actions aimed to improve or optimize processes can have a significant economic impact. Port infrastructures are subject to different meteorological conditions, such as waves, wind, and currents, that can cause such disruptions. The work of this thesis focuses on studying these conditions, collecting data, and building tools that port operators can use to maximize port safety and the performance of regular operations due to two phenomena: moored ships' movements and wave overtopping. Regular berthing, loading, and unloading ship operations are impossible in extreme conditions, and its safety and performance are affected during mean conditions. Some regulations propose acceptable movement thresholds during operations, but these limits are general, so they do not consider every individual port's and ship's characteristics. In a port environment, wave overtopping is the event that occurs when waves meet the port breakwater, and the water passes over it. This event causes significant safety problems, disrupts normal operations, and causes personal or material damage. In this work, we developed a multipurpose IoT infrastructure (hardware and software) and used it to measure moored ship's movements and monitor overtopping events in the outer port of Punta Langosteira, A Coruña, from 2015 until 2022, recording the movements of 112 ships and identifying 3709 overtopping events. We created several datasets by merging the data collected with data for the sea state and weather conditions during the ships' movements and overtopping events. We used these datasets to create several machine learning models that, using sea state and weather forecasts, predict how a moored ship will behave and whether an overtopping event will occur. We also developed a web tool that gathers the data needed for the Machine Learning models to work and uses GIS techniques and different visualizations to facilitate the interpretation and exploitation of the models' predictions. With the aid of this tool, port operators can make decisions to minimize the risk involved in regular port operations and improve their performance.