Use the coarsest global mesh that still allows targeted local resolution where the physics demands it. Global refinement is expensive; local controls should follow curvature, gaps, near-wall gradients, free shear, interfaces and expected solution features.
Surface mesh: preserve the geometry that controls the flow
The surface mesh establishes the geometric fidelity available to the volume mesh. Use curvature and proximity refinement where shape or small gaps matter. Over-refining large flat areas rarely improves the simulation, while under-resolving a leading edge, narrow gap or small jet can fundamentally change the result.
Inspect the surface mesh before generating millions of volume cells. Confirm that edges, holes, contacts and local curvature look physically correct.
Choose the volume mesher from the geometry and application
Trimmed, polyhedral and directed-style workflows each have different strengths. The correct choice depends on geometry complexity, dominant flow direction, memory/runtime targets and whether structured alignment is valuable. Do not choose a mesher only because it was used in a previous unrelated tutorial.
Complex general geometry
Robust unstructured/polyhedral or trimmed approaches are often practical for large industrial models.
Strong directional structure
Directed or swept strategies can be useful when topology supports a controlled inlet-to-outlet mesh.
Near-wall resolution must match the wall model
Choose wall treatment first, then target y+, first-layer distance, prism count, growth and total thickness. A good core mesh cannot compensate for an internally inconsistent near-wall strategy.
Use the complete near-wall workflow
Wall-treatment advisor → y+ calculator → prism-layer calculator.
Local controls should follow expected gradients
- Curvature: leading edges, fillets, blade profiles and sharp geometry changes.
- Proximity: gaps, seals, tip clearances and thin channels.
- Wakes / jets: refine downstream volumes, not only source surfaces.
- Rotating interfaces: maintain compatible scales and sufficient resolution on both sides.
- Free surfaces: refine around the interface and combine with a timestep strategy.
- Heat transfer: resolve fluid and solid thermal gradients, not only velocity gradients.
Local refinement is useful only when it covers the full region through which the important feature moves or develops.
Mesh quality is local and solver-specific
Global histograms are useful screening tools, but the worst cells matter most when they occur in high-gradient regions or near interfaces. Inspect solver-specific quality metrics, prism collapse, abrupt size transitions, non-orthogonality/skewness-type indicators and tiny isolated cells.
A few pathological cells can force a smaller timestep or destabilize a case even when the overall mesh statistics look excellent.
Build a feature-based checklist
Select walls, curvature, gaps, wakes, rotating regions, VOF, thermal gradients and motion.
Mesh independence means output independence
Do not define a mesh study only by total cell count. Refine the mesh features that control the reported result and compare engineering outputs such as pressure drop, force, heat transfer, mass flow or temperature. If the quantity changes materially, the solution is still mesh-sensitive.
- Choose a baseline mesh.
Document base size, local controls and prism strategy.
- Refine the dominant error source.
Near-wall cells, wake, gap or interface—not necessarily the whole domain.
- Compare the same engineering monitors.
Use normalized differences when possible.
- Check cost versus change.
The useful mesh is the one whose added resolution no longer changes the decision enough to justify the expense.