KB-2026005: Which y+ should I choose for which turbulence model?

In Simcenter STAR-CCM+, the appropriate (y+) value is determined primarily by the wall treatment associated with the selected turbulence model. 

The important thing to understand is that you should select your near-wall mesh resolution based on the wall treatment, rather than selecting an arbitrary (y+) value for each turbulence model. 


1. Recommended y+ values 

There are three primary wall treatments in STAR-CCM+.

Wall Treatment 

Recommended y+ 

Mesh Requirement 

High y+ 

30–50 

Coarse near-wall mesh; first cell should lie in the logarithmic region. 

Low y+ 

≤ 1 

Fine near-wall mesh; viscous sublayer must be resolved. 

All y+ 

≤ 1 or > 30 

Supports both fine and coarse near-wall meshes. 

Important: Although All y+ can handle intermediate values, the preferred targets are still either approximately 1 or below, or above 30. Avoid deliberately targeting the buffer layer (5 < y+ < 30). 


2. Turbulence model vs. recommended y+ 

Here is a practical reference for the commonly used STAR-CCM+ turbulence models. 

Turbulence Model 

Wall Treatment 

Recommended y+ 

Spalart–Allmaras (Standard) 

All y+ 

≤ 1 or > 30 

Standard K-Epsilon 

High y+ 

30–50 

Realizable K-Epsilon 

High y+ 

30–50 

Standard K-Epsilon Two-Layer 

All y+ 

≤ 1 or > 30 

Realizable K-Epsilon Two-Layer 

All y+ 

≤ 1 or > 30 

Standard K-Epsilon Low-Reynolds Number 

Low y+ 

≤ 1 

V2F Low-Reynolds Number 

Low y+ 

≤ 1 

Elliptic Blending 

Low y+ / All y+ 

≤ 1 or > 30 

Lag Elliptic Blending 

Low y+ / All y+ 

≤ 1 or > 30 

K-Omega SST 

All y+ (recommended) 

≤ 1 or > 30 

The exact available wall-treatment options can vary by model and STAR-CCM+ version. 


3. Which turbulence model and y+ should you choose? 

For practical CFD applications, I would use the following guidelines: 

CFD Application 

Recommended Turbulence Model 

Wall Treatment 

Target y+ 

External aerodynamics 

K-Omega SST 

All y+ 

~1 

Automotive aerodynamics 

K-Omega SST 

All y+ 

~1 

Internal pipe flow 

Realizable K-Epsilon 

High y+ 

30–50 

Centrifugal pump 

K-Omega SST 

All y+ 

~1 

Heat transfer applications 

K-Omega SST 

All y+ 

~1 

Flow separation 

K-Omega SST 

All y+ 

~1 

General industrial flow 

Realizable K-Epsilon Two-Layer 

All y+ 

~1 or >30 

Low-Reynolds-number flows 

Standard K-Epsilon Low-Re 

Low y+ 

≤ 1 

These are starting recommendations, not universal rules. The actual model selection should also consider Reynolds number, separation, pressure gradients, heat transfer and flow characteristics. 


4. Important considerations when generating the mesh 

There are a few things worth remembering: 

1. y+ is calculated at the first cell centroid - It is not calculated using the complete first-layer thickness. 

2. Check y+ after running the simulation - The actual y+ distribution depends on the local wall shear stress. Therefore, the initial mesh target may differ from the final calculated values. 

3. Don't focus only on average y+- Check the y+ distribution over the entire wall surface, particularly in regions where you expect: 

  • Flow separation. 

  • High velocity gradients. 

  • Strong pressure gradients. 

  • Significant heat transfer. 

  • Boundary-layer development. 

4. All y+ is not a replacement for mesh independence - Even when using All y+ treatment, you should perform mesh sensitivity studies to establish that your results are independent of mesh resolution. 


My recommendation 

For most engineering CFD applications, particularly external aerodynamics, rotating machinery and heat transfer, I would start with: 

  • Turbulence Model: K-Omega SST 

  • Wall Treatment: All y+ 

  • Target y+: Approximately 1 

  • Prism Layers: Sufficient layers to resolve the boundary layer. 

  • Mesh Validation: Check the final y+ distribution after convergence. 

For applications where computational resources are limited, a wall-function approach with y+ around 30–50 can be a reasonable alternative, provided the selected turbulence model and wall treatment support it. 

One important distinction: A simulation with y+ = 1 is not automatically more accurate than one with y+ = 40. The accuracy depends on whether the mesh resolution is consistent with the selected wall treatment and whether the overall mesh adequately captures the flow physics. 

Tags
KB-2026004: How to Apply Temperature Loads in OptiStruct – Understanding Temperature Subcase Options
KB-2026006: How to Estimate RAM for an MLFMM Simulation in Feko