Meshing Strategy for CFD

What Is It?

The CFD mesh is the discretisation of the computational domain into small control volumes (cells) over which the governing equations are solved. The mesh determines the spatial resolution of the solution — where the flow can be resolved and where it cannot. Mesh quality — cell shape, size, orthogonality, growth rate — directly affects numerical accuracy and solver stability. Mesh strategy — what cell types to use, where to refine, how to handle boundary layers — is one of the most important engineering decisions in a CFD analysis.

Why It Matters

The mesh is the spatial representation of the flow problem. A mesh that is too coarse misses flow features — separation, wakes, shocks, recirculation. A mesh that is poorly shaped causes numerical errors and convergence problems. A mesh that is refined in the wrong places wastes computational resources without improving the result. The mesh strategy must follow the expected flow physics — refine where the gradients are, resolve the boundary layer, capture the wake. Understanding mesh types, quality metrics and refinement strategy is essential for credible CFD.

Refine where the flow gradients are, not everywhere equally. Uniform refinement wastes cells in regions where the flow is smooth. Targeted refinement — boundary layers, leading edges, trailing edges, wakes, shocks — provides the best accuracy for a given cell count.

Cell Types

Structured vs Unstructured Meshes

A structured mesh has a regular connectivity — cells are arranged in a grid with predictable neighbours. This provides excellent numerical properties — fluxes are well-defined, gradients are accurate, and the solver is efficient. However, structured meshes are difficult to generate for complex geometry — the mesh must conform to the body shape, which may require multi-block approaches with careful block decomposition. An unstructured mesh has irregular connectivity — cells can have varying numbers and arrangements of neighbours. This allows automated mesh generation for arbitrary geometry but introduces more numerical diffusion and requires more cells for the same accuracy. Most industrial CFD uses a hybrid approach — structured or prism layers near walls, unstructured in the far field.

Refinement Around Key Flow Features

The mesh must be refined in regions where the flow gradients are steep — where the velocity, pressure or turbulence changes rapidly. Refinement should be targeted at the physically important regions, not applied uniformly.

• Boundary layers — prism/inflation layers with adequate near-wall resolution (y+)
• Leading edges — high curvature, strong acceleration, suction peaks
• Trailing edges — shear layer, wake formation, base pressure
• Wakes — the region behind the body where the velocity deficit persists
• Shocks — steep pressure, density and velocity discontinuities (compressible flow)
• Recirculation regions — separated flow, reversed velocity, high turbulence
• Jets — high-velocity injection; shear layers between jet and surrounding flow

Mesh Quality Metrics

Local vs Global Refinement

Global refinement — uniformly refining the entire mesh — increases the cell count dramatically without improving the result in regions where the flow is smooth. Local refinement — refining only where the gradients are steep — provides the same accuracy in the regions of interest at a fraction of the cell count. Adaptive mesh refinement (AMR) — where the solver automatically refines based on the computed flow gradients — is the most efficient approach, though it requires solver support and careful setup. The mesh strategy should always favour targeted local refinement over uniform global refinement.

Mesh Independence

Mesh independence is the condition where further mesh refinement does not change the engineering result — the forces, pressure loss or flow rate have stabilised. A mesh independence study involves running the analysis on progressively finer meshes and checking whether the result converges. If the result changes significantly with refinement, the mesh is not adequate. Mesh independence is not about making the mesh as fine as possible — it is about finding the mesh where the result has stabilised and further refinement is unnecessary. This is a critical verification step for any CFD analysis.

MESH CONSIDERATION: A very fine surface mesh does not compensate for inadequate boundary-layer resolution. The near-wall mesh (prism layers, y+) and the wake mesh are as important as the surface mesh — sometimes more so.

Mesh Topology, Alignment & Numerical Diffusion

Cell count alone is a poor measure of CFD mesh quality. The mesh should be aligned with the dominant transport where practical, resolve geometric curvature and expected shear layers, and avoid abrupt size changes that introduce interpolation error. Long, highly skewed cells may be entirely appropriate inside an attached boundary layer but damaging in a separated shear layer or vortex core. Conversely, forcing isotropic cells everywhere can make the model unnecessarily expensive without improving the quantities of interest. Numerical diffusion is especially important when vortices, jets, wakes, interfaces or shocks must survive over long distances: a coarse or poorly aligned mesh can smear the feature even though the residuals appear converged. Mesh design should therefore start from the flow structures that must be resolved, not from a global target cell size.

Mesh Independence Must Follow the Engineering Quantity

A credible mesh study tracks the output used for the engineering decision. Global lift may converge while a local hinge pressure, heat-transfer coefficient, wall shear or recirculation length remains mesh-sensitive. At least three systematically refined meshes are useful where practical, but the refinement must affect the region and physics that govern the output. Comparing screenshots of contour smoothness is not a convergence study. Record the mesh metric, the quantity of interest and the change between levels; where possible estimate the remaining discretisation uncertainty rather than declaring the finest mesh “converged”. If the final structural load is obtained by surface integration or CFD-to-FEA mapping, include the integrated force and moment in the mesh study because local field changes can cancel globally—or vice versa.

Mesh independence is demonstrated on the decision-driving result, not on cell count or visual contour smoothness.

Key Takeaways

• The mesh determines spatial resolution — it must be refined where the flow gradients are steep
• Hexahedral cells offer the best quality; tetrahedral are easiest for complex geometry; prism layers are essential for boundary layers
• Structured meshes have better numerical properties; unstructured are more flexible
• Mesh quality — skewness, orthogonality, growth rate — directly affects accuracy and stability
• Mesh independence — the result has stabilised with refinement — is a critical verification step