CFD Mesh Independence Studies: Why They Matter and How to Do Them Right

What is a mesh independence study?

In CFD, the computational domain is divided into a finite number of cells — the mesh. The governing equations of fluid flow are solved numerically at each cell. The finer the mesh, the closer the numerical solution approaches the true mathematical solution of those equations. A mesh independence study (also called a grid convergence study) is the process of demonstrating that your mesh is fine enough that further refinement would not meaningfully change your results.

Without this step, your CFD results carry an unquantified numerical error. That error can be small — or it can be 30% or more in regions of high gradient such as boundary layers, shear layers, and recirculation zones. You simply cannot know without checking.

Why mesh matters so much in CFD

Unlike FEA, where a coarse mesh typically produces a conservative (under-predicted stiffness, over-predicted deflection) result, CFD errors from insufficient mesh refinement are not consistently conservative. A coarse mesh may overpredict or underpredict drag, heat transfer coefficient, pressure drop, or recirculation zone size — and the direction of the error depends on the specific flow physics involved.

The regions most sensitive to mesh density are:

• Boundary layers: The thin region of retarded flow near a wall where viscous effects dominate. The y+ value of the first cell layer must be appropriate for the turbulence model used — typically y+ < 1 for low-Reynolds-number models and y+ between 30 and 300 for wall functions.
• Shear layers and free jets: Where two flow streams of different velocity meet, steep velocity gradients demand fine mesh to resolve mixing accurately.
• Recirculation zones: Separated flow regions behind bluff bodies, at pipe bends, or in diffusers are highly sensitive to mesh density and significantly affect pressure drop and heat transfer predictions.
• Geometric features: Small gaps, thin fins, sharp edges, and narrow channels all require local mesh refinement to capture their flow resistance correctly.

How to perform a mesh independence study

Step 1 — Define your key output quantities

Before you start refining, decide what quantities you are monitoring. For an aerodynamic study this might be drag coefficient (Cd) and lift coefficient (Cl). For a heat exchanger it might be overall heat transfer coefficient (U) and pressure drop (ΔP). For an internal combustion analysis it might be peak wall temperature and residence time. You need at least one — ideally two or three — representative quantities that are sensitive to the flow physics you care about.

Step 2 — Create three meshes at a consistent refinement ratio

Generate three meshes: coarse, medium, and fine. The refinement ratio between successive meshes should be consistent — a factor of approximately 1.5 in linear cell size in each direction is commonly used, which roughly doubles the cell count in 2D and triples it in 3D. Document the cell count and key mesh metrics (maximum skewness, minimum orthogonality, average y+) for each mesh.

Step 3 — Run all three meshes with identical solver settings

The turbulence model, boundary conditions, solver relaxation factors, and convergence criteria must be identical across all three runs. Any change in solver settings between runs introduces a variable that makes it impossible to attribute differences in results solely to mesh refinement.

Step 4 — Plot results against mesh density

Plot your key output quantities against cell count (or a mesh refinement parameter such as 1/h where h is a characteristic cell size). A well-behaved study will show the quantity converging asymptotically as mesh density increases. The difference between the medium and fine results should be significantly smaller than the difference between the coarse and medium results.

Step 5 — Apply the Grid Convergence Index (GCI)

The Grid Convergence Index (Roache, 1994) is the industry-standard method for quantifying the numerical uncertainty associated with a specific mesh. It uses the results from three mesh levels and the refinement ratio to estimate the discretisation error band on your key quantities. A GCI of less than 5% on the fine mesh is generally considered acceptable for engineering applications; less than 1–2% is required for high-accuracy work.

The GCI calculation is straightforward and should be documented in every CFD report. CHS Intl includes GCI values as standard in all simulation deliverables.

Common mistakes to avoid

Refining only globally, not locally

Global mesh refinement is computationally expensive and often unnecessary. The correct approach is targeted refinement in the regions identified as sensitive (boundary layers, shear layers, geometric features). A mesh that is globally coarse but locally refined in the right places will outperform a uniformly fine mesh of the same cell count.

Using wall functions with insufficient y+ resolution

Wall function turbulence treatments require the first cell centroid to sit within the log-law region (y+ between 30 and 300). If y+ falls below 11 with wall functions active, the wall shear stress and heat transfer predictions become unreliable. Check the y+ distribution on all walls before accepting results.

Stopping the study after two mesh levels

Two data points cannot demonstrate convergence — they can only show a difference. Three data points are the minimum needed to establish a convergence trend and calculate a meaningful GCI. Stopping at two meshes is a common shortcut that produces an unverifiable result.

Changing the turbulence model between mesh levels

Some turbulence models (k-ω SST in low-Re mode) require y+ < 1 at the wall. If your coarse mesh has y+ = 50 and your fine mesh has y+ = 0.5, you cannot compare them directly — the effective turbulence model has changed between runs. Mesh independence must be demonstrated for a specific turbulence model configuration.

What a good mesh independence study report looks like

A properly documented mesh independence study should include: cell counts for all three meshes; a table of key output quantities at each mesh level; a convergence plot; y+ statistics for all wall boundaries; GCI values for the key quantities; and a clear statement of which mesh was selected for the final production runs and why.

At CHS Intl, this documentation is included in every CFD deliverable as a matter of standard practice — not as an optional add-on. If you receive a CFD report that does not include a mesh independence study, ask for one before acting on the results.