Most of the effort in a CFD study goes into the mesh and the model; most of the errors are made in the boundary conditions. The reason is simple: a bad mesh is visible, a wrong boundary condition is not. The solution converges, a colourful picture comes out, and the number is wrong [1] [2].
The question to ask
Do not ask what is known in the system; ask what is imposed.
- If a pump or fan drives the flow and you are looking for the operating point, do not impose the flow rate. If you do, you have defined the machine's curve yourself and the analysis merely confirms your assumption.
- If the system is driven by a pressure difference between two reservoirs, specify pressure at inlet and outlet; the flow rate is the result.
- If you are modelling a line whose flow rate has been measured, flow rate at the inlet and pressure at the outlet is the correct choice.
Specifying pressure at both ends means leaving the flow rate to the solver. Sometimes that is exactly what you want; sometimes it is why the solution never settles.
Where the boundary sits matters more than its type
The outlet should be placed where the flow has recovered. An outlet placed just downstream of a separation region suppresses reverse flow, corrupts the upstream solution, and that corruption propagates into the model.
A practical measure: move the outlet at least ten hydraulic diameters downstream of the last geometric disturbance. A cheap extension volume beats an expensive mistake.
Turbulence quantities
Turbulence intensity and length scale are requested at the inlet and usually left at the default. In problems where separation is decisive, these values change the result visibly. If they are unknown, run two extreme values; if the difference changes your decision, that is an uncertainty item and belongs in the report.
Boundaries that get confused
Symmetry states that no flux crosses that surface. It is not a wall and produces no friction. Confusing the two corrupts the result silently.
In external flow, a far-field boundary placed too close makes the body see its own blockage. Keeping the blockage ratio (body area / cross-section area) below 5% is a good starting point.
References
- Ferziger, J. H., Perić, M. & Street, R. L. (2020). Computational Methods for Fluid Dynamics, 4th ed. Springer. DOI: 10.1007/978-3-319-99693-6
- Menter, F. R. (1994). Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal, 32(8), 1598–1605. DOI: 10.2514/3.12149