Simcenter STAR-CCM+ What turbulence model should I use for my simulation?

Turbulence Model Classes in Simcenter STAR-CCM+

Simcenter STAR-CCM+ offers three major classes of turbulence models:

1. Eddy-Viscosity Models: These models assume a linear relationship between the strain rate and the Reynolds-stress tensor (Boussinesq assumption). The turbulent eddy-viscosity is defined using different approaches:
   • Spalart-Allmaras: Solves a transport equation for the turbulent eddy-viscosity directly. Suitable for fully turbulent, attached flows with small pressure gradients (e.g., flow over wings at moderate angles of attack).
   • k-epsilon: Defines turbulent eddy-viscosity as a function of turbulent kinetic energy (k) and turbulent dissipation rate (epsilon). It is well-suited for industrial applications with multiple recirculation zones, heat transfer, or impingement.
   • k-omega: Similar to k-epsilon but uses specific dissipation rate (omega) instead of epsilon. Performs well for separated flows, low-Reynolds number flows, or when near-wall physics resolution is important.
2. Reynolds Stress Transport Models: Solves transport equations for the stress-tensor components directly. More accurate for strongly curved flows (e.g., cyclone separators) or flows with strong rotation (e.g., turbomachines). Less robust than eddy-viscosity models.
3. Transient Models: Directly describes part or all of the turbulent-eddy motion. Includes Detached-Eddy Simulation (DES) and Large-Eddy-Simulation (LES).
   • LES: Computes only the largest scales of turbulent motion, modelling the smallest scales using eddy-viscosity assumption. More robust but computationally expensive (100-fold increase). Suitable for low-Reynolds number internal flows.
   • DES: Hybrid method combining eddy-viscosity models (Spalart-Allmaras or k-omega) with LES. Suitable when flow is poorly predicted in steady conditions and large-scale turbulent structure prediction is important (e.g., aeroacoustics). Requires significant flow separation.

Factors to Consider When Selecting a Turbulence Model

1. Flow Characteristics: Assess the flow regime (laminar, transitional, turbulent), Reynolds number, pressure gradients, separation, and recirculation zones.
2. Turbulence Intensity: Determine the expected level of turbulence in the flow. High-intensity flows may require more advanced models like Reynolds stress transport or LES.
3. Computational Resources: Consider the available computational power and time constraints. Transient models (LES, DES) are more computationally demanding than eddy-viscosity models.
4. Validation: Always validate the selected turbulence model against experimental data or benchmarks to ensure accurate results.

Recommendations

• For fully turbulent, attached flows with small pressure gradients, use the Spalart-Allmaras model.
• For industrial applications with recirculation, heat transfer, or impingement, use k-epsilon models.
• For separated flows, low-Reynolds number flows, or when near-wall physics resolution is critical, use k-omega models.
• For strongly curved flows or flows with a strong rotation, consider Reynolds stress transport models.
• For low-Reynolds number internal flows or when large-scale turbulent structure prediction is essential, use LES or DES, respectively.

Remember, LES and DES require finer spatial and temporal resolutions than eddy-viscosity and Reynolds stress transport models. Ensure the grid quality is sufficient and choose appropriate time steps to resolve finer structures.

Limitations and Potential Pitfalls

1. Eddy-Viscosity Models:
   • Assume isotropic turbulence, which may not be accurate for flows with strong anisotropy or complex strain fields.
   • May not capture the correct physics in flows with strong separation, curvature, or rotation.
   • Tend to overpredict turbulent kinetic energy and underpredict turbulent dissipation rate in regions with adverse pressure gradients.
2. Reynolds Stress Transport Models:
   • More computationally expensive than eddy-viscosity models due to the additional transport equations solved.
   • May encounter convergence difficulties or numerical instabilities, especially for complex geometries or high-Reynolds number flows.
   • Require careful selection of model constants and boundary conditions to ensure accurate results.
3. Transient Models (LES, DES):
   • Extremely computationally expensive, especially for high-Reynolds number flows or complex geometries.
   • Require very fine spatial and temporal resolutions to capture turbulent structures accurately.
   • Sensitive to grid quality and numerical scheme selection.
   • May not accurately predict near-wall behavior without proper wall modeling or resolution.

Being aware of these limitations is essential when selecting a turbulence model for your CFD analysis. If the chosen model's assumptions are violated or the flow conditions are not suitable, the simulation results may be inaccurate or misleading. Always validate your results against experimental data or benchmarks, and be prepared to iterate on your turbulence model selection if necessary.

Simulating Physics > Modeling Turbulence > Selecting a Turbulence Modeling Approach