A diffuser is a transition that widens the section to reduce velocity and gain static pressure. It appears at fan and compressor outlets, in wind tunnels, in ventilation ducts and in suction pipes.
It looks simple; because it works under an adverse pressure gradient it is one of the hardest geometries for CFD.
There is an optimum angle
If the angle is small: there is no separation, but the diffuser becomes long, friction losses accumulate and it takes up space.
If the angle is large: the boundary layer cannot withstand the adverse pressure gradient and separates. The separated flow blocks part of the section, the effective expansion ratio falls, and pressure recovery decreases.
An optimum lies between the two, within a narrow band. Diffuser design is therefore a compromise, not a maximisation.
The inlet condition matters as much as the angle
The same diffuser behaves entirely differently depending on the inlet velocity profile. A distorted, swirling flow coming straight out of a bend or a fan triggers separation much earlier. See the secondary flow note.
Catalogue diffuser efficiencies are measured with a uniform inlet; the value achieved in a real installation is almost always lower.
A modelling warning
The turbulence model decides where separation occurs, and the diffuser is the most sensitive example of this: one model sees no separation at all, another starts it early. The results can differ by a factor in pressure recovery. See the separation note.
Separated diffuser flow is also unstable: the separation can switch from one wall to the other. A steady solution either fails to converge or produces a false symmetry. See the symmetry note.
Control methods
Vanes, vortex generators or suction can be used to delay separation. All give measurable gains but are sensitive to geometric detail — which is exactly where the analysis adds value.
References
- Menter, F. R. (1994). Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal, 32(8), 1598–1605. DOI: 10.2514/3.12149
- Langtry, R. B. & Menter, F. R. (2009). Correlation-based transition modeling for unstructured parallelized computational fluid dynamics codes. AIAA Journal, 47(12), 2894–2906. DOI: 10.2514/1.42362