What's the best turbulence model for your fluid dynamics simulations?

1 Reynolds-Averaged Navier-Stokes (RANS) Models

RANS models are the most widely used turbulence models in engineering applications. They are based on averaging the flow variables over time, and introducing additional terms to account for the effects of turbulence. RANS models are relatively simple, fast, and robust, but they have some limitations. They cannot capture the unsteady features of turbulence, such as vortices, fluctuations, and separation. They also depend on empirical coefficients that need to be calibrated for different flow conditions.

2 Large Eddy Simulation (LES) Models

LES models are more advanced than RANS models, as they resolve the large-scale structures of turbulence, and model only the small-scale motions. LES models can capture the unsteady and complex behavior of turbulence, and provide more accurate and detailed results. However, they also require more computational resources, such as memory, CPU time, and grid resolution. They are also more sensitive to the boundary conditions, initial conditions, and subgrid-scale models.

3 Detached Eddy Simulation (DES) Models

DES models are a hybrid approach that combines RANS and LES models. They use RANS models near the walls, where the turbulence is dominated by viscosity, and switch to LES models in the regions where the flow is separated or detached from the walls. DES models can handle flows with large separation zones, such as bluff bodies, wings, and control surfaces. They are more efficient than LES models, but still more expensive than RANS models. They also suffer from some numerical issues, such as grid dependence and model transition.

4 Direct Numerical Simulation (DNS) Models

DNS models are the most accurate and realistic turbulence models, as they solve the Navier-Stokes equations without any simplifications or approximations. They resolve all the scales of turbulence, from the largest eddies to the smallest dissipative motions. DNS models can provide detailed insight into the physics of turbulence, and serve as a benchmark for other models. However, they are also the most demanding in terms of computational resources. They require very fine grids, small time steps, and high-order schemes. They are only feasible for simple geometries and low Reynolds numbers.

5 Tips for Choosing and Using Turbulence Models

The best turbulence model for simulations is determined by many factors, such as the flow regime, geometry, Reynolds number, accuracy requirements, and computational budget. To make a good decision, start with a simple and robust model like RANS and check if it meets expectations. If not, try a more advanced model like LES or DES and compare the results. Additionally, use appropriate grid resolution and quality for the chosen model. RANS models need finer grids near the walls while LES and DES models require finer grids in high turbulence regions. DNS models need very fine grids throughout. Validate the model against experimental data or DNS results, if available, by assessing global and local quantities such as lift, drag, pressure, velocity, and turbulence intensity. Furthermore, analyze the sensitivity of your model to input parameters like boundary conditions, initial conditions, and model coefficients by trying different values to see how they affect the results. Finally, document your model selection and validation process while reporting any uncertainties or limitations of your simulations.

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?