3D CFD Analysis of the Flow behind a Screen
No, the graphic below is not a snapshot of some species of coral. It is actually from a 3D CFD analysis of a “box” of turbulence that is gradually subsiding. Put more technically, it's an iso-surface of the vorticity in the Z direction in a periodic box of decaying homogeneous isotropic turbulence.
If you are not already familiar with simulations of isotropic turbulence, you may be wondering why anyone would bother modeling something as boring as a clump of turbulent air that's just sitting out in space somewhere. There are at least three reasons why people look at this kind of case:
Designing an Abstract 3D CFD Analysis to Compare with Experiment
- To test the spatial and temporal capabilities of their CFD solver without the interfering presence of more complex boundary conditions.
- To test the ability of a turbulence model to properly dissipate turbulent kinetic energy.
- To study the details of the evolution of turbulence and the interaction of eddies of different scales.
Given the abstract nature of the setup, it might surprise you to find out that these CFD results can be compared to experimental findings. Experimentalists are also interested in studying the evolution of turbulence, and in this case, I was duplicating the conditions of the experiments of Comte-Bellot and Corrsin.
In the experiments, a screen was installed at the beginning of a test section in a wind tunnel. Air was blown through the screen and measurements of the turbulent fluctuations were taken at several points downstream. For the simulations, the idea was to begin with the conditions measured at the first experimental measurement location downstream of the grid and then let the turbulence evolve from there.
The assumption in both the experiments and the CFD was that the turbulence would evolve essentially undisturbed as it was convected downstream. Thus, we could use CFD to predict the measurements at various downstream locations by running for a simulated time corresponding to how long it would take the experimental mean velocity to convect the flow from the first measurement location to the point of interest.
The plot below shows the resolved turbulent kinetic energy decay for three different computational grid resolutions. The experimental comparison points were obtained by integration of the measured energy spectrum over the range resolvable on the given computational mesh. As you can see, the results of the 3D CFD analysis closely match the experimental measurements at all grid resolutions.
I originally ran this case as part of the work for my PhD dissertation. The objective was to test a new turbulence model I was developing. By first testing the model in a simple, low-speed configuration, I hoped to both weed out any obvious problems and also to show a basic viability of the model before applying it to more complex cases. As this case has been investigated using many other CFD solvers, I could also compare my results with those of other codes.Large Eddy Simulation versus Conventional CFD
A final note: Most modern CFD simulations solve the Reynolds Averaged Navier-Stokes (RANS) equations (or the inviscid Euler equations), but in this case, I took an approach called Large Eddy Simulation (LES). The main difference between the RANS and LES approaches lies in their handling of turbulence.
In a RANS simulation, one attempts to model all the scales of turbulent motion. In the LES approach, the code tries to resolve whatever turbulence it can, and only the scales of motion that are too small for the computational grid are modeled. Thus, true LES simulation is always an unsteady 3-D CFD analysis (because turbulence is an unsteady 3-D phenomenon), while RANS cases can be, and frequently are, run in a 2-D steady-state mode.
When you are ready,
return to the CFD Applications page check out more examples of 3D CFD analysis.
or explore the other topics on
the Innovative CFD home page