Fire Dynamics Simulator - Revision history

Emma: Add 3 watermarked SVG illustrations from the Advanced visual plan

2026-11-02T16:11:56Z

Add 3 watermarked SVG illustrations from the Advanced visual plan

← Older revision		Revision as of 17:11, 2 November 2026
Line 20:		Line 20:

	FDS uses a staggered arrangement of variables. Scalar quantities such as density, temperature and pressure are associated with cell centres, while velocity components are located at cell faces. This arrangement supports the coupling of pressure and velocity. It also gives a specific numerical meaning to the locations of the calculated variables.<ref name="F-6">McGrattan et al. (2026), pp. 6–7.</ref>		FDS uses a staggered arrangement of variables. Scalar quantities such as density, temperature and pressure are associated with cell centres, while velocity components are located at cell faces. This arrangement supports the coupling of pressure and velocity. It also gives a specific numerical meaning to the locations of the calculated variables.<ref name="F-6">McGrattan et al. (2026), pp. 6–7.</ref>
			[[File:fse-fire-dynamics-simulator-v1.svg\|thumb\|upright=1.5\|alt=A cubic cell shows scalar variables at its centre and velocity components on its faces.\|FDS places scalar quantities at cell centres and velocity components at the corresponding faces. Based on McGrattan et al. (2026), pp. 6–7.]]

	[[Mesh resolution\|Mesh resolution]] determines the scale at which spatial variation can be represented. In the supplied LES formulation, the filter width is associated with local cell size. Refinement changes the division between resolved motions and modelled smaller-scale effects. It therefore affects more than the visual smoothness of the output.<ref name="F-6" />		[[Mesh resolution\|Mesh resolution]] determines the scale at which spatial variation can be represented. In the supplied LES formulation, the filter width is associated with local cell size. Refinement changes the division between resolved motions and modelled smaller-scale effects. It therefore affects more than the visual smoothness of the output.<ref name="F-6" />
Line 64:		Line 65:

	The refinement therefore increases the number of cells by a factor of eight. It may also require smaller time steps. The guide's Courant–Friedrichs–Lewy (CFL) constraint relates time-step size to advection across a cell. For an illustrative one-directional flow with no velocity divergence, it reduces to the expression below.<ref name="F-47">McGrattan et al. (2026), p. 47, equations 4.98–4.100.</ref>		The refinement therefore increases the number of cells by a factor of eight. It may also require smaller time steps. The guide's Courant–Friedrichs–Lewy (CFL) constraint relates time-step size to advection across a cell. For an illustrative one-directional flow with no velocity divergence, it reduces to the expression below.<ref name="F-47">McGrattan et al. (2026), p. 47, equations 4.98–4.100.</ref>
			[[File:fse-fire-dynamics-simulator-v2.svg\|thumb\|upright=1.5\|alt=A grid-refinement example compares cell counts, time steps and a simplified computational-work measure.\|In this simplified comparison, halving cell width multiplies cell count by eight and the cell-update measure by sixteen. Based on McGrattan et al. (2026), pp. 6 and 47–49.]]

	<div class="fse-equation">CFL = UΔt/Δx</div>		<div class="fse-equation">CFL = UΔt/Δx</div>
Line 92:		Line 94:

	Smokeview converts simulation data into graphical representations. This can make plume development, smoke spread and changing temperatures easier to examine. It does not replace scrutiny of the underlying numerical values or assumptions. A convincing animation remains a visualisation of the selected model and inputs.<ref name="VR-11.3" /><ref name="VR-11.5" />		Smokeview converts simulation data into graphical representations. This can make plume development, smoke spread and changing temperatures easier to examine. It does not replace scrutiny of the underlying numerical values or assumptions. A convincing animation remains a visualisation of the selected model and inputs.<ref name="VR-11.3" /><ref name="VR-11.5" />
			[[File:fse-fire-dynamics-simulator-v3.svg\|thumb\|upright=1.5\|alt=A workflow separates fire-model inputs, numerical calculations and Smokeview visualisation.\|Smokeview presents calculated data; interpretation still depends on the scenario, submodels and supporting evidence. Based on Van Rossum (2027), §§11.3–11.5, and McGrattan et al. (2026), pp. 5–6.]]

	[[CFAST\|Consolidated Fire and Smoke Transport (CFAST)]] offers a different representation based on a small number of uniform gas zones in each compartment. FDS resolves more spatial variation through its grid. The additional detail is useful when local flow and geometry matter to the assessment. It also requires suitable input information and numerical resolution to support that extra detail.<ref name="F-1" /><ref name="F-2" /><ref name="VR-9.5">Van Rossum (2027), §9.5.</ref>		[[CFAST\|Consolidated Fire and Smoke Transport (CFAST)]] offers a different representation based on a small number of uniform gas zones in each compartment. FDS resolves more spatial variation through its grid. The additional detail is useful when local flow and geometry matter to the assessment. It also requires suitable input information and numerical resolution to support that extra detail.<ref name="F-1" /><ref name="F-2" /><ref name="VR-9.5">Van Rossum (2027), §9.5.</ref>

Peter: New FSE wiki article: Fire Dynamics Simulator (written by GPT-6)

2026-11-01T23:05:59Z

New FSE wiki article: Fire Dynamics Simulator (written by GPT-6)

New page

'''Fire Dynamics Simulator''' ('''FDS''') is a numerical model designed to simulate fire-driven flow and smoke movement. Van Rossum identifies it as a specialist application of [[Computational fluid dynamics|computational fluid dynamics]] (CFD), with Smokeview providing graphical presentation of its results.<ref name="VR-11.5">Van Rossum (2027), §11.5.</ref> The supplied technical reference describes a low-Mach-number formulation that solves transport relationships for gas flow, energy and species, with additional models for combustion, radiation and other fire processes.<ref name="F-5">McGrattan et al. (2026), pp. 5–6.</ref><ref name="F-12">McGrattan et al. (2026), p. 12.</ref>

Its default turbulent-flow treatment is [[Large eddy simulation|large eddy simulation]] (LES). Larger motions are resolved on a numerical grid, while smaller-scale effects require modelling.<ref name="F-2">McGrattan et al. (2026), pp. 2–3 and 6.</ref> This article describes the supplied sixth-edition technical reference, revision 6.11.1 dated 10 July 2026. Its assumptions and capabilities should be interpreted as those of that identified version, with application-specific evidence required before drawing engineering conclusions.<ref name="F-title">McGrattan et al. (2026), title page.</ref><ref name="VR-11.4">Van Rossum (2027), §11.4.</ref>

== Low-speed flow and the governing formulation ==

Fire-driven motion commonly involves thermal expansion, buoyancy and mixing at speeds well below the speed of sound. The FDS formulation simplifies the general fluid equations for that class of flow. The simplification allows attention to the transport processes relevant to many building fires. It does not retain every phenomenon represented by a fully compressible formulation.<ref name="F-2" /><ref name="F-10">McGrattan et al. (2026), pp. 10–11.</ref>

The guide distinguishes a background thermodynamic pressure from a smaller pressure perturbation driving motion. The background pressure enters the equation of state. The pressure and velocity solution then satisfies the expansion implied by heat release and transport. This separates the low-speed flow problem from unnecessary acoustic detail.<ref name="F-10" />

The introductory discussion associates the low-Mach approximation with velocities below Mach 0.3. It separately states that the validation discussed there extends to approximately Mach 0.1. Those are different statements about mathematical scope and supporting evidence. The broader approximation range should not be presented as demonstrated accuracy across all such flows.<ref name="F-2" />

Mass, species, momentum and energy are connected within the numerical formulation. The solver obtains mixture density from transported species densities and enforces the equation of state. Energy conservation is expressed through the velocity-divergence treatment used in advancing the flow. A successful calculation therefore depends on coupled transport, rather than independent temperature and smoke animations.<ref name="F-10" />

The default numerical method uses explicit, second-order approximations on a structured grid. The guide also lists default turbulence, combustion and radiation treatments. Optional alternatives are available for some processes, with additional computational cost. The user must justify whether the extra detail improves the prediction required for the particular application.<ref name="F-5" />

A calculation's physical meaning follows those choices. For example, using a low-speed model does not establish a capability to predict every pressure-wave phenomenon. Similarly, resolving a gas flow does not ensure that the fuel-decomposition model is appropriate. The selected scenario must remain within the demonstrated scope of the relevant equations and submodels.<ref name="F-2" /><ref name="F-12" /><ref name="VR-11.4" />

== Grid representation and resolved motion ==

FDS uses a staggered arrangement of variables. Scalar quantities such as density, temperature and pressure are associated with cell centres, while velocity components are located at cell faces. This arrangement supports the coupling of pressure and velocity. It also gives a specific numerical meaning to the locations of the calculated variables.<ref name="F-6">McGrattan et al. (2026), pp. 6–7.</ref>

[[Mesh resolution|Mesh resolution]] determines the scale at which spatial variation can be represented. In the supplied LES formulation, the filter width is associated with local cell size. Refinement changes the division between resolved motions and modelled smaller-scale effects. It therefore affects more than the visual smoothness of the output.<ref name="F-6" />

Rectilinear grids provide a convenient representation of many building features. In the basic obstruction treatment, rectangular solids align with the underlying grid. A feature smaller than the selected cells may consequently be altered or omitted in that representation. The modelled geometry must be checked against the flow paths important to the assessment.<ref name="F-6" /><ref name="VR-11.3">Van Rossum (2027), §11.3.</ref>

The supplied 2026 guide also documents a cut-cell method for more complex geometry. This qualifies the simplified rectangular-only description in Van Rossum's introductory account. The relevant capability depends on the implementation and options being used. A general statement that every FDS geometry must be reduced to rectangular obstructions would be too broad for this version.<ref name="F-6" /><ref name="VR-11.3" />

Multiple meshes can be processed in parallel. That can extend the calculation that available computing resources can support, but does not remove resolution requirements. Van Rossum recommends documenting the chosen grid, the sensitivity checks and remaining uncertainty. The purpose is to relate spatial detail to the actual prediction being relied upon.<ref name="F-6" /><ref name="VR-11.3" />

The resolved motion still represents filtered fields. A practical grid does not resolve the thin chemical reaction zones of a real flame. The guide consequently relies on submodels for mixing and combustion below the resolved scale. More cells can improve some aspects of a result without making all the underlying fire physics exact.<ref name="F-2" /><ref name="F-12" />

== Fire specification, chemistry and smoke ==

A [[Design fire|design fire]] establishes the fire scenario being assessed. Van Rossum emphasises that a CFD model does not automatically identify the real fire history for the building. Geometry, material properties and the assumed fire development must be defined during model preparation. The solver calculates the consequences of that represented scenario.<ref name="VR-11.3" /><ref name="VR-11.4" />

FDS includes models for thermal decomposition and gaseous combustion, but their use introduces further material and modelling assumptions. Their availability is different from having sufficient data to predict any furnishing or storage arrangement. The guide's introduction identifies real-material characterisation as a major difficulty. A detailed solid-phase treatment cannot replace missing evidence about the material being represented.<ref name="F-1">McGrattan et al. (2026), p. 1.</ref><ref name="F-12" />

For many applications, the combustion treatment is based on mixing-limited reaction between fuel and oxidiser. Reactants convert to products according to a characteristic mixing time. The model also provides options addressing combustion in oxygen-limited spaces. Their appropriate use depends on the conditions and level of prediction required.<ref name="F-12" />

The chemistry representation can combine several gas constituents into lumped species. For example, background air and combustion products each contain multiple primitive constituents. Transporting those mixtures reduces the number of independent equations needed for the selected reaction system. The model then recovers individual species fractions through the specified composition relationships.<ref name="F-9">McGrattan et al. (2026), pp. 9–10.</ref>

[[Soot yield|Soot yield]] is a material input with consequences for smoke and radiation. The guide's simple-chemistry treatment defines yield in relation to the mass of fuel reacted. It states that carbon monoxide and soot yields are zero by default in that treatment unless specified. A plausible smoke prediction therefore requires deliberate attention to these inputs.<ref name="F-53">McGrattan et al. (2026), p. 53.</ref>

A calculated species field must be interpreted through the reaction and transport assumptions that generated it. A correct velocity field does not independently validate a selected soot yield. Nor does an assigned yield predict how that yield changes in every ventilation condition. The model configuration and supporting fuel data remain part of the explanation of the result.<ref name="F-9" /><ref name="F-53" /><ref name="VR-11.4" />

== Radiation, surfaces and ventilation ==

The default radiation treatment solves a transport equation for a grey, absorbing gas. Absorption depends on composition and temperature, with soot playing an important role in many fires. The guide also describes optional spectral treatment. The selected representation balances computational practicality with the information available about the fire gases.<ref name="F-12" />

Radiation near flames presents a particular difficulty because grid-average temperatures can be lower than the temperatures within unresolved reacting regions. The source term therefore requires special treatment. This is a physical consequence of limited resolution, rather than merely a plotting issue. Near-fire exposure should be interpreted using the documented radiation assumptions.<ref name="F-12" />

Surface and particle processes extend the model beyond gas transport. The guide describes thermal decomposition, water and fuel droplets, and detector or sprinkler-related features. Each adds source terms or boundary conditions to the calculation. The result remains dependent on the specific model used for the represented interaction.<ref name="F-12" />

For example, the supplied water-suppression treatment includes an empirical reduction in local burning related to water present on a surface. Its coefficient depends on material and configuration. This is not a universal prediction of extinguishment for every sprinkler and fuel arrangement. The relevant empirical assumptions must be justified for the intended application.<ref name="F-103">McGrattan et al. (2026), p. 103.</ref>

Ventilation can be represented by prescribed boundary flows or by the documented heating, ventilation and air conditioning (HVAC) network solver. Fixed inlet and outlet conditions do not automatically couple the behaviour of all branches. The network representation instead links nodes and ducts through conservation equations. The distinction affects how ventilation-system interactions are represented during the fire.<ref name="F-111">McGrattan et al. (2026), p. 111.</ref>

The ventilation network can carry heat and combustion products between spaces and supply air to a fire. It can also represent parts of a smoke-control arrangement. The numerical representation must correspond to the intended operating scenario. Merely assigning an extract boundary does not establish the performance of the complete physical system.<ref name="F-111" />

== Worked grid and time-step comparison ==

Consider a selected rectangular domain measuring 20 m by 10 m by 4 m. Uniform cubic cells of 0.20 m give 100, 50 and 20 cells along the three directions. Multiplication gives 100,000 cells. Reducing the spacing to 0.10 m gives 200 by 100 by 40 cells, or 800,000 cells.<ref name="F-6" />

The refinement therefore increases the number of cells by a factor of eight. It may also require smaller time steps. The guide's Courant–Friedrichs–Lewy (CFL) constraint relates time-step size to advection across a cell. For an illustrative one-directional flow with no velocity divergence, it reduces to the expression below.<ref name="F-47">McGrattan et al. (2026), p. 47, equations 4.98–4.100.</ref>

<div class="fse-equation">CFL = UΔt/Δx</div>

{| class="wikitable"
! Symbol !! Meaning in this one-directional illustration !! SI unit
|-
| CFL || Courant number || dimensionless
|-
| U || Selected advection speed || m/s
|-
| Δt || Time-step duration || s
|-
| Δx || Cell width along the flow || m
|}

Select a velocity of 2 m/s and a Courant number of one. The illustrative time step is 0.10 seconds for 0.20 m cells and 0.05 seconds for 0.10 m cells. The finer grid therefore needs twice as many updates for the same simulated duration. These values isolate the advection relationship and are not prescribed FDS settings for a real fire.<ref name="F-47" />

For a selected 600-second duration, the corresponding counts are 6,000 and 12,000 time steps. Multiplying cell count by step count gives 600 million and 9.6 billion cell updates respectively. This simplified measure increases by a factor of sixteen. It illustrates why refining a three-dimensional transient calculation can increase work faster than the cell count alone.<ref name="F-6" /><ref name="F-47" />

This is an arithmetic comparison, not a measured runtime prediction. Actual computational work also depends on solver processes, model options and parallel execution. The full FDS stability treatment includes multiple velocity components and expansion, with other constraints where relevant. Its guide describes automatic adjustment rather than one universal fixed time step.<ref name="F-5" /><ref name="F-6" /><ref name="F-49">McGrattan et al. (2026), pp. 47–49.</ref>

The example also does not identify an acceptable mesh for a particular fire. That requires examination of the physical scales, geometry and outputs of interest. The selected 0.20 m and 0.10 m values only make the refinement arithmetic visible. Suitability must come from model documentation, sensitivity checks and relevant evidence.<ref name="VR-11.3" /><ref name="VR-11.4" />

== Outputs, visualisation and model choice ==

FDS provides a spatially varying representation of the simulated gas environment. Van Rossum describes CFD outputs such as gas temperature, velocity, smoke concentration and heat exposure. These quantities differ in meaning and evidential requirements. A useful interpretation identifies the location, time and physical quantity relevant to the engineering question.<ref name="VR-11.2">Van Rossum (2027), §11.2.</ref>

Smokeview converts simulation data into graphical representations. This can make plume development, smoke spread and changing temperatures easier to examine. It does not replace scrutiny of the underlying numerical values or assumptions. A convincing animation remains a visualisation of the selected model and inputs.<ref name="VR-11.3" /><ref name="VR-11.5" />

[[CFAST|Consolidated Fire and Smoke Transport (CFAST)]] offers a different representation based on a small number of uniform gas zones in each compartment. FDS resolves more spatial variation through its grid. The additional detail is useful when local flow and geometry matter to the assessment. It also requires suitable input information and numerical resolution to support that extra detail.<ref name="F-1" /><ref name="F-2" /><ref name="VR-9.5">Van Rossum (2027), §9.5.</ref>

The choice should therefore follow the required prediction. A compartment-average question and a local plume-deflection question do not demand the same representation. The presence of a powerful CFD tool does not itself justify every possible application. The model must represent the mechanisms governing the decision with evidence appropriate to the requested output.<ref name="F-1" /><ref name="F-2" /><ref name="VR-11.4" />

== Verification, evidence and reporting ==

The supplied technical reference documents mathematical methods, while Van Rossum emphasises comparison with suitable experimental evidence for quantitative interpretation. A model's documented capability does not eliminate uncertainty in an individual scenario. Inputs, resolution and submodel applicability must be examined together. The conclusion should remain tied to the conditions actually assessed.<ref name="F-5" /><ref name="VR-11.4" />

Grid sensitivity is one component of that examination. Results should be checked for changes that matter to the required output when resolution is altered. Time-step behaviour also deserves attention because it interacts with the resolved scales. A finer mesh with unsuitable time resolution does not automatically preserve the additional transient detail.<ref name="VR-11.3" /><ref name="F-47" />

Software identification is necessary because the available methods and defaults belong to a particular implementation. The revision used, selected physical options and important input assumptions should accompany the result. The difference between the textbook's geometry summary and the supplied technical guide demonstrates why version-specific documentation matters. Historical descriptions should not be treated as unchanging software limits.<ref name="F-title" /><ref name="F-6" /><ref name="VR-11.3" />

The final interpretation should distinguish calculated fields from the real event they approximate. Combustion, turbulence, radiation and suppression each contain assumptions that can influence the result. The output is most useful when those assumptions and the relevant evidence are visible. Numerical detail alone does not establish the accuracy of the engineering conclusion.<ref name="F-1" /><ref name="F-5" /><ref name="VR-11.4" />

== Notes ==
<references />

== References ==
* McGrattan, K., Hostikka, S., Floyd, J., McDermott, R., Vanella, M., Mueller, E., & Paul, C. (2026). ''Fire Dynamics Simulator technical reference guide. Volume 1: Mathematical model'' (NIST Special Publication 1018-1, 6th ed., revision 6.11.1, 10 July 2026). National Institute of Standards and Technology.
* Van Rossum, E. (2027). ''Introduction to Fire Safety Engineering''. Fire Safety Academy.

[[Category:Fire Modelling]]