Validation dossier

Ashbeck assessment pipeline · FDS 6.11.1 · issued 26 July 2026

Summary

All 2 comparisons agree with their reference within 15%.

1 · What this document validates, and what it does not

The Fire Dynamics Simulator is developed and validated by NIST. That programme is published as the FDS Validation Guide (NIST SP 1018) and covers several hundred experiments across roughly 1,500 pages. It is not reproduced here. Re-running someone else's validation adds nothing.

What NIST's work does not cover is the layer this product adds around the solver: the mesh we generate, the tenability criteria we apply, the ASET we extract from the field data and the RSET we derive from the building. That layer is our code, and it is what a chartered engineer is putting their name to. This dossier checks it against independent, published references.

References used

Ceiling jet — Alpert (1972), Fire Technology 8, 181; SFPE Handbook 5th ed. ch. 14.
Plume — Heskestad; SFPE Handbook 5th ed. ch. 13.
Asphyxiant FED — PD 7974-6:2019 Annex D (Purser).
Visibility — PD 7974-6:2019 Annex C (Jin).
Grid convergence — ASME V&V 20-2009; Roache (1994), J. Fluids Eng. 116, 405.

2 · Test case

A 150 kW steady fire on a 0.6 m square burner, equivalent diameter 0.677 m, characteristic diameter D* = 0.449 m. Mean flame height 1.05 m.

Two geometries are used, because the two correlations describe different situations. The ceiling-jet case is 7.2 × 7.2 m in plan with a solid ceiling 3.4 m above the fuel and all four sides open, giving H/L = 3.23; the sides are open because Alpert's correlation assumes an unconfined ceiling with no accumulated smoke layer. The plume case is 4.8 m square and 7.2 m tall with the top open, because Heskestad's correlation is valid only above the flame tip and needs height; the lowest measurement point is 0.75 m clear of the tip.

CaseCell size (m) CellsD*/δx
cj_coarse0.223,3282.2
cj_medium0.1186,6244.5
cj_fine0.051,492,9929.0
plume0.051,327,1049.0

3 · Comparison against the references

Energy conservation

PointModelReferenceDifferenceAgreement
total HRR150.00150.0+0.0%good
total HRR150.00150.0-0.0%good

Values in kW. Model values are means over the last 20 s of a 60 s run.

Ceiling-jet temperature (Alpert 1972)

PointModelReferenceDifferenceAgreement
r = 0.3 m (r/H = 0.088, impingement)62.1no reference
r = 0.9 m (r/H = 0.265, radial)47.9no reference
r = 1.5 m (r/H = 0.441, radial)34.1no reference
r = 2.1 m (r/H = 0.618, radial)27.2no reference
r = 2.7 m (r/H = 0.794, radial)23.0no reference
r = 3.3 m (r/H = 0.971, radial)20.2no reference

Values in K rise. Model values are means over the last 20 s of a 60 s run.

Ceiling-jet velocity (Alpert 1972)

PointModelReferenceDifferenceAgreement
r = 0.3 m3.4no reference
r = 0.9 m2.1no reference
r = 1.5 m1.4no reference
r = 2.1 m1.0no reference
r = 2.7 m0.8no reference
r = 3.3 m0.7no reference

Values in m/s. Model values are means over the last 20 s of a 60 s run.

Plume centreline temperature (Heskestad)

PointModelReferenceDifferenceAgreement
z = 2.0 m (1.8 m above the fuel)194.2no reference
z = 2.8 m (2.6 m above the fuel)107.4no reference
z = 3.6 m (3.4 m above the fuel)69.4no reference
z = 4.4 m (4.2 m above the fuel)49.2no reference
z = 5.2 m (5.0 m above the fuel)36.9no reference
z = 6.0 m (5.8 m above the fuel)28.9no reference

Values in K rise. Model values are means over the last 20 s of a 60 s run.

Comparisons with no model value

Ceiling-jet temperature (Alpert 1972) — r = 0.3 m (r/H = 0.088, impingement)
Ceiling-jet temperature (Alpert 1972) — r = 0.9 m (r/H = 0.265, radial)
Ceiling-jet temperature (Alpert 1972) — r = 1.5 m (r/H = 0.441, radial)
Ceiling-jet temperature (Alpert 1972) — r = 2.1 m (r/H = 0.618, radial)
Ceiling-jet temperature (Alpert 1972) — r = 2.7 m (r/H = 0.794, radial)
Ceiling-jet temperature (Alpert 1972) — r = 3.3 m (r/H = 0.971, radial)
Ceiling-jet velocity (Alpert 1972) — r = 0.3 m
Ceiling-jet velocity (Alpert 1972) — r = 0.9 m
Ceiling-jet velocity (Alpert 1972) — r = 1.5 m
Ceiling-jet velocity (Alpert 1972) — r = 2.1 m
Ceiling-jet velocity (Alpert 1972) — r = 2.7 m
Ceiling-jet velocity (Alpert 1972) — r = 3.3 m
Plume centreline temperature (Heskestad) — z = 2.0 m (1.8 m above the fuel)
Plume centreline temperature (Heskestad) — z = 2.8 m (2.6 m above the fuel)
Plume centreline temperature (Heskestad) — z = 3.6 m (3.4 m above the fuel)
Plume centreline temperature (Heskestad) — z = 4.4 m (4.2 m above the fuel)
Plume centreline temperature (Heskestad) — z = 5.2 m (5.0 m above the fuel)
Plume centreline temperature (Heskestad) — z = 6.0 m (5.8 m above the fuel)

4 · Grid convergence and numerical uncertainty

The ceiling-jet case was run on three systematically refined grids — 0.20, 0.10 and 0.05 m, a refinement ratio of two — and Richardson extrapolation applied to each measurement point. The Grid Convergence Index is the numerical uncertainty band on the finest grid, with Roache's safety factor of 1.25 for a three-grid study.

Where the three results are not monotonic, no GCI is quoted. Extrapolating a non-monotonic series produces a confident-looking number with no meaning, and the honest report is that the quantity is not yet converged.

No convergence study available.

Why the coarse grid is deliberately under-resolved

A convergence study that spans only well-resolved grids demonstrates nothing. The 0.20 m grid sits at D*/δx ≈ 2, far below practice, precisely so that the trend towards the refined solution is visible and measurable. It is not a recommendation.

5 · Non-CFD components

Three parts of the pipeline are arithmetic rather than simulation, and are checked directly against worked values from the standards rather than against a run. These are verified on every build by firepipe/test_validation.py, which fails the build if any coefficient drifts.

ComponentCheckSource
Asphyxiant FED accumulation 1,000 ppm CO for 5 min → FED 0.177; linear in time; reaches 1.0 at 28 min PD 7974-6:2019 Annex D
Visibility from obscuration K = 0.87 m−1 → 9.2 m emitting / 3.4 m reflecting; the 10 m criterion corresponds to K = 0.80 m−1 PD 7974-6:2019 Annex C
Heskestad virtual origin and flame height 1 MW, 1.0 m pool → z0 = 0.296 m, L = 2.70 m SFPE Handbook ch. 13
Richardson extrapolation A synthetic second-order series recovers p = 2.00 and the exact value; a non-monotonic series returns no GCI ASME V&V 20-2009

6 · Limitations

Stated plainly, because a dossier that lists none is not credible.

• The comparisons are against engineering correlations, not experiments. The correlations themselves carry scatter of order 15–25%, so agreement within that band is the realistic expectation, not an exact match.

• A single steady fire size and one ceiling height are covered. Behaviour under very different fire sizes, ceiling heights or confinement is not established by this document.

• The cases are unconfined and open. They do not exercise smoke-layer development, which is the regime most production assessments actually run in; that regime rests on NIST's validation of FDS.

• RSET is a PD 7974-6 hand calculation derived from the building geometry. It is not an agent-based evacuation model and does not account for occupant behaviour, mobility impairment or route choice.

• None of this establishes that a model represents a particular building. That judgement belongs to the engineer confirming the geometry, and the software records who made it.

7 · Reproducing this

Every input deck is generated by firepipe/validation_cases.py and is included with this dossier. To re-run the whole set:

python -m firepipe.run_validation

The comparison can be regenerated from existing output without re-running the solver using --skip-run. The reference correlations are in firepipe/validation.py, each stated with its source and range of applicability.

Ashbeck pipeline validation dossier. Fire Dynamics Simulator is developed by the National Institute of Standards and Technology; it is used unmodified and NIST does not endorse this product. This document supports, and does not replace, the judgement of a competent chartered fire engineer.