FDS cell size: D*/δx and the mesh sensitivity study

fdsrun editorial team · 6 November 2026 · 6 min read

How to choose the FDS cell size: characteristic fire diameter D*, the D*/δx index, cubic cells and a grid convergence check.

"What cell size should I use?" is the most common question from new FDS users. The NIST documentation answers honestly: it depends on what you are trying to achieve. The recommended route is to start with a relatively coarse mesh and refine it until the results stop changing appreciably — a mesh sensitivity study (FDS User Guide, section 6.3.6). Below are the tools that make it a sensible exercise.

Characteristic fire diameter D*

For buoyant plumes, resolution is measured by the non-dimensional ratio D*/δx, where δx is the nominal cell size and D* is the characteristic fire diameter:

D* = ( Q / (ρ∞ · cp · T∞ · √g) )^(2/5)

Q is the heat release rate, ρ∞ ≈ 1.2 kg/m³, cp ≈ 1.0 kJ/(kg·K), T∞ ≈ 293 K, g = 9.81 m/s². D*/δx can be read as the number of cells spanning the characteristic diameter of the fire — the more, the better resolved the fire dynamics. If the heat release rate changes over time, so does the resolution.

Is there a "safe" D*/δx?

Engineering practice often quotes a range of 4–16 (used, among others, in the US NRC validation report NUREG-1824). Keep the NIST caveat in mind, though: in the FDS validation cases D*/δx spans two orders of magnitude and was chosen from grid studies of the specific scenarios, so taking any of these values as a universal minimum is inappropriate. A 10 cm cell may be adequate for smoke spread from a sizeable fire in a building, but not for a small smouldering source.

Cell shape and mesh interfaces

• Near-cubic cells — the length, width and height of a cell should be similar. Strongly flattened cells (aspect ratio above roughly 2:1) reduce accuracy.
• At mesh interfaces abutting cells should have the same cross-section or an integer ratio (for example 1:2). Resolution jumps larger than 2:1 and non-integer ratios are a common source of trouble (User Guide, section 6.3.4).
• J and K cell counts are best factored into small primes (2, 3, 5) so the FFT pressure solver runs efficiently. The gain is usually small, because the pressure solver uses less than 10% of the CPU time.

The cost of resolution

Halving the cell size gives about 8 times more cells and a roughly 2 times smaller time step — together about 16 times longer runs (details in how long an FDS simulation takes). A good model therefore has fine cells only where they matter: around the fire source and in the zones you assess (for example escape routes), with coarser meshes elsewhere.

How to run a mesh sensitivity study

1. Prepare three variants: coarse, baseline and fine (for example 20, 10 and 5 cm in the fire zone).
2. Choose the quantities your conclusions rest on: temperatures and visibility at devices, smoke layer height, flows through openings, HRR.
3. Compare the curves. If going from baseline to fine does not change the results in a way that matters for the assessment, the baseline resolution is sufficient.

The variants are independent, so the fastest way is to run them in parallel — in fdsrun each variant starts on its own server, and before the start you see the forecast time and cost of each.

Sources: FDS User Guide (NIST SP 1019), sections 6.3.1, 6.3.4, 6.3.6; FDS Validation Guide (NIST SP 1018), chapters 3 and 6.1; US NRC NUREG-1824. FDS and Smokeview are software of the National Institute of Standards and Technology; NIST is not affiliated with fdsrun.