In one line: CFD solves the equations of fluid flow numerically on a mesh. That lets engineers explore designs before building them, but the results are only as good as the mesh, the models and the validation behind them.
Core idea
Fluid flow obeys conservation of mass, momentum and energy; the momentum equations are known as the Navier–Stokes equations. They can be solved by hand only for the simplest cases. Computational fluid dynamics splits the flow region into millions of small cells, applies the conservation laws to each, and solves them iteratively. Commercial codes, such as the finite-volume solver whose theory guide the course uses, make this widely available. They also make it easy to produce colorful pictures that are wrong.
How it works
- Geometry and mesh. Model the flow region and divide it into cells. The cells must be finer where the flow changes quickly, such as near walls and in shear layers.
- Physics models. Choose what to include: compressibility, heat transfer, multiple phases, and particles. The big choice is how to handle turbulence:
- RANS models average it out. They are cheap and widely used, but they rely on assumptions.
- Large-eddy simulation resolves the large eddies. It is more faithful, but far more expensive.
- Boundary conditions. Specify the inlets, outlets, walls and operating conditions. Errors here are common, and they silently ruin the results.
- Solve and check convergence. Iterate until the residuals and key quantities stop changing.
- Verify and validate:
- Verification asks whether we solved the equations correctly. A key test is mesh independence: refine the mesh and check that the answer doesn’t change.
- Validation asks whether we solved the right equations. Compare the results with experiments or trusted data.
In practice
The course example is an inlet particle separator: the curved intake that keeps sand and dust out of a helicopter’s turbine engine. The engine air is turned sharply, and heavy particles, unable to follow the turn, are carried into a scavenge duct. A 1975 design guide for these separators was built largely on systematic tests [1]. Today the same trade-off can be explored with CFD and particle tracking: separation efficiency against pressure loss, where every bit of loss costs engine power. Testing is still needed to validate the results.
For a leader, CFD is best treated like any measurement. Ask how it was calibrated and what its uncertainty is. Use it to compare options and narrow the field, then confirm the finalists with prototypes and tests.
Questions to ask the experts
- Has the result been shown to be independent of the mesh?
- Which turbulence model was used, and why is it appropriate here?
- What experimental data has this model been validated against, and how close was it?
Connections
- The discipline: Mechanical and Aerospace Engineering. A key application: Fundamentals of Flight.
- Simulation and test together: Prototyping, Design of Experiments. Uncertainty: Statistics in Engineering.
- Capabilities exercised: Inquiry, Decision Making, Connecting Across Disciplines.
Sources
- Duffy et al., inlet particle separator design guide [1]
- Course lectures and a commercial solver’s theory guide; no public source.