A probabilistic model of ship performance in ice based on full-scale data

Navigation in ice covered waters is a challenging and hazardous activity, threatening the ship herself and the environment. Proper evaluation of the attainable ship speed in ice prior to setting off the journey is essential to mitigate the risk of besetting in ice which may cause further damages. The majority of the existing models evaluating ship speed in ice are physics-driven requiring detailed, not always accessible, input data, thus featuring high uncertainty. In this paper we propose an event-oriented model, reflecting the ice features under which an event of interest occurs (a given speed interval or a significant speed drop). To this end we utilize a large set of full-scale data and Bayesian learning algorithm to organize the data into a probabilistic model linking the ice conditions with the ship speed. Conditional probabilities are derived from the detailed dataset obtained in the course of full-scale tests of S.A. Agulhas II (ice class PC 5) carried out in the Baltic Sea in 2012. A set of parameters related to ice conditions; rudder and engine settings are recorded with the use of on board instruments. Ice thickness is measured with the use of electromagnetic device. Various types of ice navigation are performed i.e. steaming straight through the level and ridged ice as well as in ice channel. The model yields the probability of a certain speed under given ice conditions, and has application potential in risk assessment of besetting in ice or route finding problem. This approach could be developed further e.g. by incorporating other types of maneuvers.