Abstract
<jats:p>The user wants just the updated abstract in plain text format, with α removed.` We present a novel 2.5-dimensional Discrete Element Method (DEM) model, termed the Thin Ice Model (TIM), for efficient simulation of ice-structure interactions in broken and unbroken sea ice fields. The model is based on Voronoi tessellation of a single-layer ice sheet, where the ratio of element area to thickness exceeds 10. By constraining the degrees of freedom from six (full 3D) to four — retaining only yaw rotation — the model ensures that all computational cells remain horizontal, enabling contact force calculations via 2D convex polygon intersection rather than costly 3D intersection algorithms. This simplification reduces the computational cost compared to full 3D DEM formulations and is naturally suited for parallel implementation, including GPU-CUDA architectures. A modified Real Multidimensional Internal Bond (RMIB) approach is employed to model intact ice sheet failure, with an original separation of normal forces into horizontal and vertical components, controlled by smoothing functions K(z). The model is validated against model-scale shallow-water ice-structure interaction experiments by Lemström and Polojärvi (2022) for weak, medium, and strong ice. The structural contact stiffness in the simulation is matched to the effective stiffness of the composite experimental panel, requiring no ice-type-specific fitting. The steady-state horizontal ice load is reproduced with errors below 1% for weak and medium ice and 18% for strong ice, the latter attributed to the simulation not having fully reached the steady-state phase. The peak load for strong ice matches the experiment within 2%. The model correctly captures the experimentally observed non-proportionality of ice load and ice strength, whereby weak ice yields a higher steady-state load than medium ice. Rubble pile geometries are reproduced for the strong ice case.</jats:p>