Engineering research
How to report CFD grid convergence with three meshes
Evaluate the same quantity of interest on coarse, medium, and fine meshes that are refined systematically. Establish iterative convergence for each run before attributing changes to mesh spacing.
Gather geometry, boundary conditions, turbulence model, mesh settings, and solver logs. The quantity can be drag, pressure drop, or local velocity. Keep its definition and extraction location consistent.
Mesh refinement assesses spatial discretization error. Residual reduction, mesh agreement, and accuracy of the physical model each require their own evidence.
How to do it
Freeze the operating condition and output
Record reference length, boundary conditions, discretization scheme, units, and the extraction region. For unsteady runs, keep time settings comparable and assess time-step error separately.
Use a consistent refinement strategy
Report representative spacing, cell count, and near-wall resolution. Check that key regions were refined. For complex meshes, explain how you derived an effective refinement ratio.
Inspect convergence on each mesh
Plot the quantity of interest alongside residuals. Make iterative changes small relative to the mesh differences and record the stopping criteria.
Report the trend and assumptions
Compare coarse-to-medium and medium-to-fine changes. For observed order, Richardson extrapolation, or GCI, report the formula, refinement ratio, safety factor, and asymptotic-range checks. If the sequence oscillates or the order is implausible, investigate the mesh family and solver error.
Worked example
Synthetic pressure-drop example at the same measurement points. The representative spacing is halved at each step.
| Mesh | Spacing | Pressure drop |
|---|---|---|
| Coarse | 4 mm | 118 Pa |
| Medium | 2 mm | 110 Pa |
| Fine | 1 mm | 108 Pa |
The differences are 8 Pa and 2 Pa. With monotone asymptotic behavior and a refinement ratio of 2, the example has observed order log(8/2)/log(2) = 2. Retain the iteration histories, refinement strategy, and assumptions in a real report.
Common questions
Do two close results show mesh independence?
Report the difference for those runs. A third mesh and an error estimate help characterize the trend. Scope conclusions to the quantity and operating condition.
Are four orders of residual reduction enough?
Check stability of the output and compare its variation with the mesh differences. Also record how residuals were normalized.
Verified sources
The official sources below were checked on 2026-11-13. Access conditions and update dates are rechecked before publication.
RigorMed workspace
Review my mesh refinement table
Bring spacing, cell counts, target quantities, and residual logs for three meshes. Assemble a comparison table and a methods paragraph for review.
