Ember ignition enhancement — heat transfer coupling and logistic model

[ghaffari-eng]

Summary

Extended the existing EMBER_IGNITION_MODEL to support heat transfer from landed embers to solid surfaces and added an alternative logistic regression ignition model. All changes are backward compatible — defaults produce identical results to the current code.

Branch with changes: https://github.com/ghaffari-eng/fds/tree/ember-ignition
Diff: master...ghaffari-eng:fds:ember-ignition

What changed

5 source files modified: type.f90, read.f90, part.f90, wall.f90, func.f90

1. Ember-to-surface heat transfer: When a landed ember has EMBER_HEAT_TRANSFER=.TRUE. on the receiving SURF, it deposits a heat flux (radiative + convective) onto the wall cell based on ember temperature and contact area. This feeds into Q_NET_F in SOLID_HEAT_TRANSFER, so the existing 1D heat conduction solver and TMP_IGN ignition check handle the rest. Multiple embers on the same cell accumulate.
2. Logistic ignition model: EMBER_IGNITION_TYPE=1 replaces the power-based normal CDF with a logistic regression that accounts for fuel moisture content, wind speed, ember mass, and ember temperature. Coefficients are user-configurable. The original model is untouched at EMBER_IGNITION_TYPE=0 (default).
3. Heat-transfer-based ignition: EMBER_IGNITION_TYPE=2 skips the probabilistic check entirely — ignition is determined by the solid phase solver when surface temperature reaches TMP_IGN from the deposited ember heat flux.

New SURF parameters

Parameter | Default| Description
EMBER_HEAT_TRANSFER |.FALSE.| Enable ember heat flux to surface
EMBER_CONTACT_FRACTION | 0.5 | Fraction of ember area in contact
EMBER_CONTACT_HTC | 25.0 | Contact HTC (W/m2/K)
EMBER_EMISSIVITY | 0.9 | Ember emissivity
EMBER_IGNITION_TYPE | 0 | 0=current CDF, 1=logistic, 2=HT-based
EMBER_IGNITION_FMC | 10.0 | Fuel moisture content (%)
EMBER_IGNITION_COEFFS(5) | 9.98 | Logistic regression coefficients

Test results

Two verification cases in Verification/WUI/ — a 700°C ember dropped onto a wood surface:

Case EMBER_HEAT_TRANSFER Max floor temp (°C)
ember_ignition_ht .TRUE. 35.76
ember_ignition_ht_off .FALSE. (default) 31.47

Looking for feedback on the approach, particularly whether adding Q_EMBER to QDXKF in the implicit solver is the right integration point, and whether this aligns with the intended direction for ember physics.

[mcgratta]

First, thank you for taking an interest in this problem. So that we all better understand what is being proposed, can you post a simple written description of the sub-model?

[ghaffari-eng]

First, thank you for taking an interest in this problem. So that we all better understand what is being proposed, can you post a simple written description of the sub-model?

Thanks Kevin.

The basic idea is pretty straightforward: right now, when an ember lands on a surface, there's no heat transfer between them. The ember just triggers a probabilistic ignition check (in level-set mode) and that's it. What I've done is add actual thermal coupling — the ember deposits a heat flux onto the surface based on how hot it is and how much contact area it has. Radiation and convection, nothing exotic. That flux feeds into the existing 1D heat conduction solver, and if the surface gets hot enough to hit TMP_IGN, ignition kicks in through the normal pathway. No new ignition criterion needed.

For the level-set wildfire mode, I've also added a logistic regression option as an alternative to the power-based CDF. It factors in fuel moisture, wind speed, ember mass, and temperature - closer to what the experimental literature (Manzello et al.) suggests matters for ignition probability. The original model is still there as the default.

Worth noting what this doesn't cover: no smoldering-to-flaming transition, no pile radiative feedback between embers, no moisture transport in the receiving fuel. It's a first-order improvement, letting the solid phase solver do its job rather than bypassing it with a probability.

On a related note, the Road Map also mentions subgrid turbulent effects on ember transport. I've been interested in the ember modelling side of FDS for a while. Is there a preferred approach you've been considering for the SGS velocity fluctuation model, or is that open territory?

[mcgratta]

I have transferred this Discussion thread to Issue #16075