Is your turbulence model inconsistent with the near-wall mesh?

Problem: the turbulence model looks reasonable, but wall behavior does not

Typical symptoms include pressure drop that changes strongly with mesh, heat-transfer coefficients that are too high or too low, separation moving unexpectedly, skin friction spikes, or large zones where y+ falls between intended low-y+ and wall-function ranges.

The model name alone does not tell you whether the near-wall solution is resolved appropriately.

Two fundamentally different near-wall strategies

Some solver formulations offer all-y+ or blended treatments, but “blended” does not remove the need for a coherent mesh. A domain with some critical walls at y+≈1 and others wandering through intermediate values can still create inconsistent accuracy.

Common mismatch patterns

• Targeting y+≈1 but using too few prism layers, so the first cell is fine while the rest of the boundary layer transitions too abruptly.
• Using wall functions while local y+ drops into the buffer region around stagnation, separation or low-speed zones.
• Assuming a single first-cell height produces a single y+ value over an entire complex surface.
• Changing turbulence models without redesigning the near-wall mesh or checking the solver's wall-treatment behavior.
• Using coarse surface resolution so strong curvature and separation points are under-resolved even though first-cell height is small.
• Ignoring thermal near-wall requirements when heat transfer, not only momentum, is the quantity of interest.

How to diagnose wall-treatment inconsistency

1. Plot solved y+ over every engineering-critical wall.

   Use distributions or histograms, not a single average. Identify where the mesh falls outside the intended regime.

2. Inspect prism-layer cross-sections.

   Check actual layer count, growth and total thickness through attached and separated boundary-layer regions.

3. Compare wall shear or heat flux spatially with y+.

   Artificial changes aligned with mesh-regime changes are a warning sign.

4. Check sensitivity with a coherent alternative wall strategy.

   For example, compare a properly designed low-y+ mesh with a properly designed wall-function mesh rather than two partially inconsistent meshes.

5. Separate turbulence-model sensitivity from mesh sensitivity.

   Do not change model and mesh simultaneously if the goal is diagnosis.

Fix order

1. Decide the wall-treatment regime from the engineering quantity.

   Wall heat transfer and separation often demand more near-wall fidelity than a bulk-flow estimate.

2. Set the target y+ and first-cell height accordingly.

   Use an estimate as a starting point, then verify solved y+.

3. Design the complete prism stack.

   Layer count, growth and total thickness should cover the relevant boundary-layer region smoothly.

4. Repair local prism collapse or coarse surface regions.

   A nominal global setting does not help if critical walls lose layers.

5. Then compare turbulence models if model-form uncertainty remains.

   Model comparisons are meaningful only when each case has an appropriate near-wall treatment.

Common mistakes

• Calling y+≈30 or y+≈1 “correct” without stating the wall treatment.
• Using average y+ to hide large regions in the wrong regime.
• Believing SST k-ω automatically requires one exact y+ value regardless of solver formulation.
• Refining first-cell height while leaving only a few layers and a poor outer transition.
• Ignoring y+ changes between operating points when velocity, density or viscosity changes.