Abstract
With global warming enhancing the navigability of Arctic routes, the accurate prediction of ice resistance during navigation is of great engineering significance for the design and performance evaluation of polar ships. This study proposes a high-fidelity numerical simulation method that combines image recognition with CFD–DEM coupling and a six-degree-of-freedom (6-DOF) dynamic model to predict ship resistance in pack ice conditions. Using ice images from the CEHINAV towing tank experiments in Spain, the watershed image segmentation algorithm was applied to extract the spatial distribution and size information of ice blocks. A digital ice field was then reconstructed by surface injection of ice fragments of various sizes, thereby achieving consistency with the physical ice field. In the fluid–structure interaction simulations, a dynamic overset mesh and 6-DOF motion model were introduced to realistically reproduce the ship’s motion and the ship–ice interactions in the pack ice zone. Numerical simulations under different speeds and ice concentrations show that the average deviation from experimental data remains within 10%, thus confirming the accuracy and reliability of the proposed method. The results indicate that the bow region is the main area of ice loading and resistance concentration, with resistance increasing significantly as the ice concentration rises. The resistance curves exhibit evident nonlinear fluctuations and unloading phenomena. Further regional analysis reveals that the transverse resistance distribution along the hull gradually decreases from the midship toward both sides, while local regions exhibit transient fluctuations, a finding that highlights the complex and unsteady characteristics of ship–ice interactions.