Icing phenomena pose significant risks to aviation, infrastructure, and energy systems. Accurate prediction of ice formation and accumulation is critical for safety and operational efficiency. Computational fluid dynamics (CFD) models rely on meshes—discrete representations of physical space—to simulate these complex processes. The fidelity of the mesh directly influences the model’s ability to capture fine-scale icing details. This article explores why high-resolution meshes are essential for accurately capturing icing phenomena, the physics involved, and the trade-offs between computational cost and predictive power.

What Are High-Resolution Meshes?

In numerical simulations, a mesh (or grid) divides a continuous domain into small elements—cells or nodes—where governing equations are solved. High-resolution meshes feature a dense grid with small element sizes, enabling the resolution of small-scale physical features. For icing simulations, such meshes are necessary to represent thin ice layers, sharp gradients in temperature and humidity, and turbulent airflow near surfaces. The resolution determines how well the model captures phenomena such as boundary layer separation, droplet impingement, and ice accretion rates.

High-resolution meshes are characterized by:

  • Fine spatial discretization: Element sizes on the order of micrometers to millimeters in critical regions.
  • Anisotropic adaptation: Stretching or refining cells along flow gradients, such as near wing leading edges or along ice horns.
  • Multiscale capability: Resolving both large-scale flow patterns and microscale nucleation processes.

Why Resolution Matters for Icing Physics

Icing involves a cascade of physical processes: supercooled droplet transport, impact and splashing, film formation, freezing, and ice growth. These processes occur across a wide range of length and time scales. A coarse mesh smears out critical gradients, leading to inaccurate predictions of ice shape, location, and mass accumulation. High-resolution meshes directly address three key challenges:

Capturing Thin Ice Layers and Roughness

During the early stages of icing, water films and rime ice layers can be only tens of micrometers thick. Standard meshes with element sizes around 1 mm cannot resolve such thin features, resulting in false ice-free areas or underpredicted accretion. High-resolution meshes enable the model to capture the initial roughness that triggers transition from smooth film to glaze ice, fundamentally altering the downstream flow.

Resolving Droplet Impingement and Runback

Supercooled droplets follow trajectories governed by aerodynamic forces. Their impact location and subsequent runback depend on local airspeed, pressure, and surface wettability. Fine meshes near the surface (y+ < 1) are needed to accurately compute wall shear and heat transfer, which drive droplet evaporation and freezing. Without such resolution, impingement limits are shifted, and runback patterns are smeared, leading to erroneous ice shapes.

Modeling Turbulent Flow and Heat Transfer

Icing is intimately coupled with turbulence. Iced surfaces generate complex separation bubbles and reattachment zones that influence local heat transfer and mass flux. High-resolution meshes—especially those utilizing wall-resolved large eddy simulation (LES) or delayed detached eddy simulation (DDES)—capture these unsteady effects. Studies have shown that coarse Reynolds-averaged Navier-Stokes (RANS) meshes systematically underpredict ice growth on aircraft wings by 20–40% due to poor resolution of the thermal boundary layer.

Applications Where High-Resolution Meshes Are Critical

Aircraft Icing Certification

Certification authorities such as FAA and EASA require manufacturers to demonstrate safe flight in icing conditions. CFD with high-resolution meshes is increasingly used to supplement wind tunnel testing, especially for complex geometries like wing slats, engine inlets, and tail surfaces. NASA’s Glenn Research Center has developed high-fidelity icing codes (LEWICE) that rely on dense meshes to predict ice accretion on airfoils. These tools help identify critical ice shapes that could cause control loss or engine damage.

Wind Turbine Ice Accumulation

Wind turbines in cold climates suffer from ice accretion on blades, reducing aerodynamic efficiency and causing safety hazards from ice shedding. High-resolution meshes are essential for simulating the rotational effects and the complex interplay between centrifugal force, droplet trajectory, and boundary layer transition. The National Renewable Energy Laboratory (NREL) uses adaptive mesh refinement (AMR) in its OpenFAST framework to accurately model blade icing (NREL Icing Research).

Power Line and Infrastructure Icing

Ice accumulation on transmission lines can lead to catastrophic failures due to galloping or overloading. Predicting ice loads requires resolving small-diameter cables and the local wind field. High-resolution meshes with element sizes smaller than the cable diameter (e.g., 5 mm for a 30 mm cable) are necessary to capture the stagnation region where ice accretion is highest. The Ice Engineering Research Lab at Université du Québec à Chicoutimi demonstrated that coarse meshes overpredict ice mass by up to 60% because they fail to model the shedding of water droplets from the cable’s leeward side.

Challenges of Using High-Resolution Meshes

Computational Cost

Refining a mesh by a factor of two in each dimension increases the cell count by a factor of eight (3D). For a typical icing simulation on a 3-meter wing section, a coarse mesh might contain 1–2 million cells, while a high-resolution mesh may exceed 50–100 million cells. This leads to longer solution times—hours to days on high-performance computing clusters—and larger storage requirements. Researchers must balance accuracy with available resources.

Mesh Generation and Quality

Creating a high-quality high-resolution mesh for complex icing geometries (e.g., iced airfoils with scalloped formations) is nontrivial. Poor element quality—high skewness, negative volumes—can cause divergence or introduce numerical diffusion. Automated meshing tools often struggle with the jagged surfaces of glaze ice. Many research groups resort to manual refinement or overset grids, which increases engineering time.

Numerical Diffusion and Stability

Higher resolution reduces numerical diffusion (smearing of gradients) but requires smaller time steps to maintain stability (CFL condition). For multiphase flows involving droplets and phase change, the explicit time step can become prohibitively small. Implicit solvers alleviate the time step restriction but may introduce iterative errors. Adaptive time stepping is often used in conjunction with mesh refinement to manage cost.

Overcoming Limitations: Adaptive Mesh Refinement

Adaptive mesh refinement (AMR) dynamically increases resolution only where needed—near icing fronts, droplet impact regions, or high-gradient zones. AMR reduces the total cell count by 10–100× compared to a uniformly refined mesh, making high-resolution simulations feasible. The NASA Langley FUN3D solver and the open-source SU2 code both support AMR for icing simulations. Key strategies include:

  • Feature-based refinement: Cells are refined based on gradient sensors (e.g., density, velocity) or ice thickness.
  • Curvature-based refinement: Regions of high surface curvature (ice horns, ridges) receive finer cells.
  • Error-estimator refinement: Local truncation error metrics drive adaptation, ensuring the solution accuracy is uniform.

AMR has been successfully applied to simulate ice accretion on swept wings and rotating blades, showing excellent agreement with experiments while using 1/20th the cells of a uniform mesh.

Future Directions in High-Resolution Icing Modeling

GPU-Accelerated Solvers

Modern GPUs offer massive parallelism that can accelerate high-resolution mesh computations. Solvers like OpenFOAM and commercial codes (ANSYS Fluent) now include GPU-native solvers for pressure-based flows. Early results show speedups of 5–10× for icing simulations, making daily turnaround possible for certification tasks.

Machine Learning–Enhanced Meshing

Recent work uses neural networks to predict optimal mesh resolution from geometry parameters and flow conditions. A trained model can generate a mesh that concentrates cells near expected icing hot spots, reducing the need for iterative mesh refinement. This approach is in early development but promises to cut meshing time from days to minutes.

Multiphysics Coupling

Future icing models will couple high-resolution aerodynamics with detailed microphysics (droplet size distributions, non-equilibrium freezing, ice crystal slip). This requires meshes that resolve both the aerodynamic boundary layer and the subgrid-scale droplet motion. Hybrid Eulerian-Lagrangian frameworks, combined with AMR, are being developed to handle this coupling efficiently.

Summary of Benefits

  • Enhanced safety: Accurate ice shape prediction prevents dangerous flight conditions and infrastructure failures.
  • Better understanding of climate impacts: High-resolution models improve ice accumulation forecasts in weather and climate models.
  • Reduced physical testing costs: Reliable CFD reduces reliance on expensive wind tunnel and flight tests.
  • Optimized design: Engineers can design ice-phobic surfaces, heating systems, and de-icing schedules with greater confidence.

In conclusion, high-resolution meshes are indispensable for accurately capturing icing phenomena. They enable models to resolve thin ice layers, droplet dynamics, and turbulent heat transfer that drive ice growth. While computational and meshing challenges remain, advances in adaptive refinement, GPU computing, and machine learning are making high-resolution simulations practical for industrial and research applications. As icing risks grow with expanding aviation and renewable energy operations, investing in high-fidelity meshing techniques will be critical for ensuring safety and performance.

For further reading, refer to the NASA technical report on high-resolution icing simulations and the Cold Regions Science and Technology review of mesh sensitivity in ice accretion modeling.