CFAST Validation (NUREG 1934 Appendix B)
CFAST Validation (NUREG 1934 Appendix B)
This case extends the Appendix A MCR validation to a scenario with fire growth and spread: the U.S. NRC NUREG-1934 (EPRI 1023259), Appendix B — Cabinet Fire in a Switchgear Room. Where Appendix A validates deck replication of a fully prescribed fire, Appendix B additionally validates FRI3D's fire-spread modeling: the NRC model prescribes the secondary cable-tray fire as a hand-built HRR table derived offline with the FLASH-CAT method (NUREG/CR-7010), while FRI3D computes that secondary fire online — tray ignition from the CFAST-predicted thermal exposure, and tray HRR from the FLASH-CAT methodology — from nothing but the plant model geometry and the cable library. The published NRC CFAST model is run unmodified as the baseline, the FRI3D-generated deck for the same room is run with the same CFAST 7.7.3 build, and the two are compared over the full 3600 s simulation.
Reference case
A 464 kW electrical cabinet fire (PE/PVC fuel, C₂H₃.₅Cl₀.₅, heat of combustion 20.9 MJ/kg, radiative fraction 0.49, CO yield 0.147, soot yield 0.136) in a 26.5 × 18.5 × 6.1 m switchgear room, igniting a stack of three horizontal cable trays above the cabinet. Ventilation is three supply and three return fans (0.472 m³/s each, pressure cutoffs 200–300 Pa), and the room is otherwise sealed — the reference model defines a door-undercut vent but holds it closed for the whole run, so the compartment pressurizes to ≈ +250 Pa. Simulation duration 3600 s.
Input deck comparison
Every input group of the FRI3D-generated deck was compared field-by-field against the NRC model: ambient conditions, compartment geometry, leakage, mechanical ventilation, fire chemistry and yields, and the ignition-source HRR curve. All prescribed inputs match, including fan flow and the sealed-compartment configuration (both decks report a leakage area of 0.0 m²).
The remaining differences are not inputs. They are what FRI3D computes where the NRC model was given a hand-derived answer, which is the substance of this case:
- Secondary fire representation. NRC prescribes the tray fire as a fixed table at a fixed 3.8 m elevation, as a second fire object alongside the cabinet. FRI3D computes each tray's ignition time and HRR contribution and folds them into the ignition source's fire, whose effective elevation rises with the burning front (from the cabinet at 2.4 m toward the upper trays at ≈ 4.7 m as they take over). The FRI3D tray fuel loading uses the generic FLASH-CAT library defaults, not values tuned to this scenario.
- Target representation. FRI3D targets are thin gas-exposure probes because cable failure (THIEF) is evaluated outside CFAST; NRC embeds THIEF plates in-deck. The two measure different things, so tray surface temperatures are indicative rather than like-for-like.
Results
Matched-time comparison over the 3600 s run (FRI3D-generated deck vs NRC baseline):
| t (s) | HRR (kW) | Upper layer temp (°C) | Layer height (m) | Pressure (Pa) | Upper layer O₂ (%) |
|---|---|---|---|---|---|
| 300 | 89 / 81 | 23.2 / 23.1 | 5.45 / 5.45 | 229 / 229 | 20.4 / 20.4 |
| 600 | 617 / 441 | 38.9 / 33.3 | 4.83 / 4.55 | 244 / 243 | 19.8 / 20.1 |
| 900 | 1228 / 1055 | 69.8 / 67.5 | 4.38 / 3.86 | 246 / 248 | 18.2 / 19.1 |
| 1200 | 1673 / 1599 | 100.7 / 100.9 | 4.28 / 3.51 | 245 / 243 | 15.9 / 17.5 |
| 1800 | 873 / 755 | 84.7 / 76.9 | 4.35 / 3.09 | 229 / 99 | 15.1 / 16.8 |
| 3600 | 713 / 678 | 73.2 / 68.4 | 4.32 / 3.01 | 234 / 233 | 17.3 / 17.5 |
Peak / extreme values over the full run:
| Quantity | FRI3D | NRC | Difference |
|---|---|---|---|
| Peak upper layer temperature | 104.2 °C | 103.6 °C | +0.6 % |
| Peak HRR (computed vs prescribed) | 1673 kW | 1599 kW | +4.6 % |
| Peak compartment pressure | 246.8 Pa | 248.9 Pa | −0.8 % |
| Minimum upper layer O₂ | 14.5 % | 16.5 % | −2.0 pt |
| Minimum layer height | 4.27 m | 3.00 m | +1.27 m |
| Tray target surface temperature peaks | 861 / 990 / 1004 °C | 808 / 999 / 1000 °C | ≤ 6.6 % |
FRI3D generated CFAST deck vs NRC CFAST baseline — switchgear cabinet fire with FLASH-CAT tray spread
The heat release rate is the strongest statement in this case: FRI3D's peak of 1673 kW and post-burnout tail of 713 kW are computed by the FLASH-CAT engine from generic library cable data, against NRC's offline-derived 1599 kW peak and 678 kW tail — the +4.6 % difference is the methodology reproducing the reference, not a copied input. The mid-transient differences (e.g. 617 vs 441 kW at 600 s) reflect tray ignition timing: FRI3D ignites trays from the CFAST-predicted exposure, while NRC prescribed a fixed 480 s ignition. Peak upper layer temperature and peak pressure follow the baseline to within 1 %.
The single-point pressure difference at 1800 s is a narrow transient, not a systematic offset. As the fire decays the compartment falls back through the 200 Pa fan cutoff, and the supply and return fans recover asymmetrically; the NRC case dips to ≈ 40 Pa near 1730 s and recovers within ≈ 300 s. FRI3D's computed HRR is higher through that window (873 vs 755 kW at 1800 s), so it stays above the cutoff and never dips. Outside 1650–1900 s the two pressures agree within a few Pa for the entire run.
Known limitation — layer height
The hot-gas layer settles ≈ 1.3 m higher in FRI3D than in the NRC baseline, and upper layer O₂ runs ≈ 2 points lower. Both follow from a single cause, which we have isolated by sensitivity testing rather than inference.
CFAST entrains air per fire plume. NRC runs the cabinet and the tray fire as two fire objects, so two plumes draw from the lower layer independently. FRI3D folds the FLASH-CAT tray contribution into the ignition source's fire, producing one plume of the same total heat release — and one plume entrains substantially less than two. Re-running the FRI3D deck with its own computed HRR split into two plumes at the reference elevations reproduces the baseline almost exactly:
| FRI3D deck variant | Minimum layer height | Minimum upper layer O₂ |
|---|---|---|
| As generated (one plume) | 4.27 m | 14.5 % |
| Fire elevation pinned to 3.8 m | 3.85 m | 15.1 % |
| Same total HRR split into two plumes | 2.96 m | 16.4 % |
| NRC baseline | 3.00 m | 16.5 % |
So the difference is a representation choice in deck generation, not an error in the computed fire: total heat release, peak temperature and pressure are unaffected, and the same computed HRR reproduces the reference layer height to within 0.04 m once it is emitted as two plumes. Multi-plume emission for FLASH-CAT secondary fires is tracked for a future release. Until then, layer height and upper layer oxygen are conservative for temperature exposure but non-conservative for layer descent in scenarios with significant secondary cable-tray involvement, and should be treated accordingly where layer position drives the result (detector and sprinkler activation, habitability, target immersion).
Conclusion
The FRI3D-generated CFAST deck reproduces the NRC NUREG-1934 Appendix B switchgear baseline within +0.6 % on peak upper layer temperature, −0.8 % on peak pressure, and +4.6 % on peak HRR — with the secondary cable-tray fire computed by FRI3D's FLASH-CAT implementation from generic library data rather than prescribed. The layer height difference is documented above, quantified, and traced to single-plume emission of the computed secondary fire.
Reference
- NUREG-1934 / EPRI 1023259, Nuclear Power Plant Fire Modeling Analysis Guidelines, Appendix B: Cabinet Fire in Switchgear Room.
- NUREG/CR-7010 Vol. 1, Cable Heat Release, Ignition, and Spread in Tray Installations During Fire (CHRISTIFIRE) — FLASH-CAT method.
- NUREG-1824 / EPRI 1011999, Verification and Validation of Selected Fire Models for Nuclear Power Plant Applications (CFAST volume).