Project overview
The backward-facing step is one of the most studied flows in CFD. It looks simple, a channel with a sudden drop in the floor, but the step forces the flow to separate, recirculate and then reattach to the wall further downstream. Reproducing that correctly is a real test of a turbulence model, which is why the step is used as a benchmark.
The same separation and reattachment happens behind cars, over aircraft flaps, inside diffusers and in combustor and cooling passages, so a model that handles the step well can be trusted on real designs. In this project we built a backward-facing step in Ansys Fluent, with the step height of Driver's 1985 experiment, and compared the wall static pressure on both walls with the published measurements.
Objectives
Domain, mesh and solver
The domain was drawn as a two-dimensional channel with the step set into the floor, and the mesh was refined along the walls, where the separated shear layer and the reattachment point decide the wall pressure:
| Setting | Value | Notes |
|---|---|---|
| Step height | 1.27 cm | Taken from the Driver benchmark |
| Domain | 8 × 12 step heights | Tall and long enough for the flow to reattach |
| Mesh | 8,032 elements | 8,440 nodes; 250 divisions along each wall |
| Turbulence model | Standard k-ω | Good for near-wall and separated flow |
| Inlet | 20 m/s | Air in the streamwise direction |
| Outlet and walls | Pressure outlet, no-slip | Including the step face |
| Numerics | Second-order upwind | For momentum and the turbulence equations |
Second-order upwind is worth the extra effort: a first-order scheme smears out the sharp gradients in the shear layer that decide where the flow reattaches.
Results: velocity and pressure
The velocity contour at the top of this page shows the whole story of the step. The flow arrives at its full 20 m/s and separates cleanly at the step edge. Behind the step sits a pocket of slow, recirculating fluid, while the fast flow above carries on, then bends down and reattaches to the lower wall.

The pressure field fits the velocity picture. Pressure is lowest in the recirculation region behind the step, then recovers as the flow reattaches and slows. This pressure recovery matters in practice: poor recovery in a diffuser or duct means lost efficiency.
Wall pressure compared with the experiment
The wall pressure was calculated along the step-side wall and the opposite wall, at twelve stations each, and plotted against distance to compare with Driver's measurements:

To test robustness, the inlet velocity and step height were each changed by ±10%. The depth of the pressure dip behind the step changed with them, while the overall shape of the curves held, so the model responds steadily rather than jumping around.

Why validated separation CFD matters
Separation, recirculation and reattachment appear wherever a flow meets a sudden change of shape. These flows are hard to predict by intuition, and checking a model against known data, as with this benchmark, is what gives the confidence to use it on a design where no experimental data exists yet. Similar separation drives the drag of the Ahmed body in our slant angle CFD study.
Frequently asked questions
Why is the backward-facing step a CFD benchmark?
Because its simple geometry produces separation, recirculation and reattachment, which are hard for turbulence models to get right, and it has well-documented experimental data such as Driver's 1985 measurements to compare against.
What does the flow do behind the step?
It separates at the step edge, leaving a pocket of slow, recirculating fluid behind the step, while the faster flow above bends down and reattaches to the lower wall further downstream, where the pressure recovers.
How was the CFD model set up?
As a 2D channel with a 1.27 cm step, eight step heights tall and twelve long, meshed with about 8,000 elements refined at the walls, and solved in Ansys Fluent with the standard k-ω model, a 20 m/s inlet, a pressure outlet and second-order upwind numerics.
How sensitive were the results?
The inlet velocity and step height were each changed by ±10%. The depth of the pressure dip behind the step changed with them, while the overall shape of the wall pressure curves stayed the same.






















