NIST Technical Note 1889v1

CFAST – Consolidated Fire

and Smoke Transport
(Version 7)
Volume 1: Technical Reference Guide
Richard D. Peacock
Kevin B. McGrattan
Glenn P. Forney
Paul A. Reneke

http://dx.doi.org/10.6028/NIST.TN.1889v1
 NIST Technical Note 1889v1

CFAST – Consolidated Fire

And Smoke Transport
(Version 7)
Volume 1: Technical Reference Guide
Richard D. Peacock
Kevin B. McGrattan
Glenn P. Forney
Paul A. Reneke

http://dx.doi.org/10.6028/NIST.TN.1889v1

May 4, 2025
Revision: CFAST-7.7.5-0-g6feca9299

E N T OF C O M
TM M
AR

ER
D EP

CE
ICA
UN
IT

ER

D
E

ST AM
ATES OF

U.S. Department of Commerce

Howard Lutnick, Secretary

National Institute of Standards and Technology

Craig Burkhardt, Acting NIST Director and Acting Under Secretary of Commerce for Standards and
Technology
CFAST Developers

The Consolidated Model of Fire and Smoke Transport (CFAST) and Smokeview are the products
of a collaborative effort led by the National Institute of Standards and Technology (NIST). Its
developers and contributors are listed below.

Principal Developers of CFAST

Richard Peacock, NIST

Glenn Forney, NIST
Paul Reneke, NIST
Kevin McGrattan, NIST
Walter Jones

Principal Developer of Smokeview

Glenn Forney, NIST

Contributors

Kristopher Overholt, Continuum Analytics, Austin, Texas, USA

Jason Floyd, Fire Safety Research Institute, UL Research Institutes, Columbia, Maryland, USA

iii
iv
About the Developers

Richard Peacock is a chemical engineering in the Fire Research Division of NIST. He received a
bachelor of science degree from the Clark School of Engineering of the University of Maryland
in 1973. He joined the NIST staff in 1974 (then the National Bureau of Standards) and has
worked on real-scale testing and the development and validation of fire models, most notably
CFAST.

Glenn Forney is a computer scientist in the Fire Research Division of NIST. He received a bach-
elor of science degree in mathematics from Salisbury State College and a master of science
and a doctorate in mathematics from Clemson University. He joined NIST in 1986 (then the
National Bureau of Standards) and has since worked on developing tools that provide a better
understanding of fire phenomena, most notably Smokeview, an advanced scientific software
tool for visualizing Fire Dynamics Simulation data.

Paul Reneke is a computer scientist in the Fire Research Division of NIST. He received a bachelor
of science degree in mathematical sciences from Clemson Univerity and a master of science
degree in applied mathematics from The Johns Hopkins University. He joined NIST in 1990.
He has worked on the development of user interfaces, graphics and improved numerics in fire
models, notably CFAST. His research interests include sensitivity analysis and validation of
fire models.

Kevin McGrattan is a mathematician in the Fire Research Division of NIST. He received a bach-
elor of science degree from the School of Engineering and Applied Science of Columbia Uni-
versity in 1987 and a doctorate at the Courant Institute of New York University in 1991. He
joined the NIST staff in 1992 and has since worked on the development and validation of fire
models, most notably the Fire Dynamics Simulator.

Walter Jones was a physicist at NIST (now retired). He received a bachelor of arts degree in
physics from Oberlin College and a doctorate in physics from the University of Maryland. He
was the original developer of the CFAST model. In addition to the development of fire models,
he has worked on smart fire alarms and smoke control for naval vessels.

v
vi
Preface

This document provides the theoretical basis for the Consolidated Fire And Smoke Transport
(CFAST) model, following the general framework set forth in the “Standard Guide for Evaluating
the Predictive Capability of Deterministic Fire Models,” ASTM E 1355 [1]. Instructions for using
CFAST are contained in a separate user’s guide, and model assessment information is contained in
a separate verification and validation guide.

vii
viii
Disclaimer

The US Department of Commerce makes no warranty, expressed or implied, to users of CFAST,

and accepts no responsibility for its use. Users of CFAST assume sole responsibility under Federal
law for determining the appropriateness of its use in any particular application; for any conclusions
drawn from the results of its use; and for any actions taken or not taken as a result of analysis
performed using these tools.
Users are warned that CFAST is intended for use only by those competent in the fields of
fluid dynamics, thermodynamics, heat transfer, combustion, and fire science, and is intended only
to supplement the informed judgment of the qualified user. The software package is a computer
model that may or may not have predictive capability when applied to a specific set of factual
circumstances. Lack of accurate predictions by the model could lead to erroneous conclusions
with regard to fire safety. All results should be evaluated by an informed user.
Throughout this document, the mention of computer hardware or commercial software does not
constitute endorsement by the National Institute of Standards and Technology, nor does it indicate
that the products are necessarily those best suited for the intended purpose.

National Institute of Standards and Technology Technical Note 1889v1

Natl. Inst. Stand. Technol. Tech. Note 1889v1, 55 pages (May 4, 2025)
CODEN: NTNOEF

http://dx.doi.org/10.6028/NIST.TN.1889v1

ix
x
Acknowledgments

CFAST was originally developed by Walter Jones, formerly of NIST.

Continuing support for CFAST is via internal funding at NIST. In addition, support is provided
by other agencies of the U.S. Federal Government, most notably the Nuclear Regulatory Commis-
sion (NRC) and the Department of Energy (DOE). The NRC Office of Research has funded key
validation experiments, the preparation of the CFAST manuals, and the continuing development
of sub-models that are of importance in the area of nuclear power plant safety. Special thanks to
Mark Salley and David Stroup for their support. Support to refine the software development and
quality assurance process for CFAST has been provided by the DOE. The assistance of Subir Sen
and Debra Sparkman is gratefully acknowledged.
Doug Carpenter, Combustion Sciences and Engineering, has contributed numerous corrections,
clarifications, and updates to the guides and the model through his detailed review of the model
and documentation. Allan Coutts, Washington Safety Management Solutions, provided insight
into the application of fire models to nuclear safety applications and detailed review of the CFAST
document updates for DOE.
Colleen Wade of the Building Research Association of New Zealand (BRANZ), Fred Mowrer
of the California Polytechnic State University, and David Sheppard of the U.S. Bureau of Alcohol,
Tobacco and Firearms (ATF) provided useful comments on the development of CFAST version 7.

xi
xii
Contents

CFAST Developers iii

About the Developers v

Preface vii

Disclaimer ix

Acknowledgments xi

1 Overview 1
1.1 Model Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Model Version . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Model Developers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Relevant Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 Governing Equations and Assumptions . . . . . . . . . . . . . . . . . . . . . . . 2
1.6 Input Data Required to Run the Model . . . . . . . . . . . . . . . . . . . . . . . 4
1.7 Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.8 Model Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.9 Organization of this Document . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 The Basic Transport Equations 7

2.1 Model Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Solution Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Species Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 The Fire Plume 13

3.1 Combustion Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Heat Release Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Plume Entrainment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3.1 Unobstructed Plumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3.2 Wall and Corner Plumes . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.3 Weak Plumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4 Plume Temperature and Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.5 Ceiling Jet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

4 Ventilation 21

xiii
 4.1 Wall Vents (Doors, Windows, Leakage) . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.1 Calculating Flow Through Wall Vents . . . . . . . . . . . . . . . . . . . . 21
4.2 Ceiling and Floor Vents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.3 Forced Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 Leakage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5 Heat Transfer 27
5.1 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.1.1 Solving the Radiation Equations . . . . . . . . . . . . . . . . . . . . . . . 29
5.1.2 Configuration Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.1.3 Solid Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1.4 Transmittance and Absorptance . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1.5 Radiation Heat Flux to a Target . . . . . . . . . . . . . . . . . . . . . . . . 32
5.2 Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.1 Walls and Floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.2 Ceiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.3 Heat Conduction within Solid Walls or Targets . . . . . . . . . . . . . . . . . . . 34
5.4 Coupling the Gas and Solid Phase Calculations . . . . . . . . . . . . . . . . . . . 35

6 Fire Protection Calculations 37

6.1 Sprinkler and Heat Detector Activation . . . . . . . . . . . . . . . . . . . . . . . 37
6.2 Visibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.3 Smoke Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.4 Fire Suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.5 Tenability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

References 41

Appendices 44

A Nomenclature 45

B Modeling Equation Derivations 49

B.1 Pressure Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
B.2 Upper Layer Volume Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
B.3 Temperature Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

C Solving the Modeling Equations 53

xiv
List of Figures

2.1 Schematic of control volumes in a two-layer zone model. . . . . . . . . . . . . . 8

4.1 Flow patterns for horizontal flow through a vertical vent. . . . . . . . . . . . . . . 22

4.2 Relative fraction of upper layer gases extracted via a ceiling vent . . . . . . . . . 24

5.1 Schematic diagram of a compartment with indices for computing configuration

factors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2 Solid angle formed by the vectors v1 , v2 , and v3 . . . . . . . . . . . . . . . . . . . 31

xv
xvi
Chapter 1

Overview

This chapter provides a general description of the Consolidated Fire And Smoke Transport (CFAST)
model following the general guidance put forth in ASTM E1355 [1].

1.1 Model Type

CFAST is a two-zone fire model that predicts the thermal environment caused by a fire within a
compartmented structure. Each compartment is divided into an upper and lower gas layer (zone in
the term zone fire model refers to the layers being modeled). The fire drives combustion products
from the lower to the upper layer via the plume. The temperature within each layer is uniform,
and its evolution in time is described by a set of ordinary differential equations derived from the
fundamental laws of mass and energy conservation. The transport of smoke and heat from zone
to zone is dictated by empirical correlations. Because the governing equations are relatively sim-
ple, CFAST simulations typically require a few tens of seconds of CPU time on typical personal
computers.

1.2 Model Version

The first public release of CFAST was version 1.0 in June, 1990. This version was restructured
from FAST [2] to incorporate the “lessons learned” from the zone model CCFM developed by
Cooper and Forney [3]. Version 2 was released as a component of Hazard 1.2 in 1994 [4, 5].
The first of the 3.x series was released in 1995 and included a vertical flame spread algorithm,
ceiling jets and non-uniform heat loss to the ceiling, spot targets, and heating and burning of mul-
tiple objects (ignition by heat flux, temperature or time) in addition to multiple prescribed fires.
As it evolved over the next five years, version 3 included smoke and heat detectors, suppres-
sion through heat release reduction, better characterization of flow through doors and windows,
vertical heat conduction through ceiling/floor boundaries, and non-rectangular compartments. In
2000, version 4 was released and included horizontal heat conduction through walls, and horizon-
tal smoke flow in corridors. Version 5 improved the combustion chemistry. Version 6, released
in July, 2005, incorporates a more consistent implementation of vents, fire objects, and event pro-
cessing and includes a graphical user interface which substantially improves its usability. Begin-
ning with version 6, a formal revision management system was implemented to track changes to

1
the CFAST source code. The open-source program development tools are provided by GitHub:
https://github.com.
The current version of CFAST, version 7, was released in 2015.
A Wiki file, located on the CFAST web site, maintained by the developers, details the changes
made for each release: https://github.com/firemodels/cfast/wiki/Release-Notes.

1.3 Model Developers

CFAST was developed and is maintained by the Fire Research Division of the National Institute of
Standards and Technology. The developers are Richard Peacock, Glenn Forney, and Paul Reneke.
Kevin McGrattan has participated in the changes leading to CFAST, version 7.

1.4 Relevant Publications

The manuals for CFAST consist of this Technical Reference Guide, a User’s Guide [6], a Verifica-
tion and Validation Guide [7], and a Configuration Management Guide [8]. The Technical Refer-
ence Guide describes the underlying physical principles. The User’s Guide describes how to use
the model. The Verification and Validation Guide documents sensitivity analyses, model verifica-
tion, model validation, and model limitations consistent with ASTM E1355 [1]. The Configuration
Management Guide documents the processes used during the development and validation of the
model.
The U.S. Nuclear Regulatory Commission has published a verification and validation study of
five selected fire models commonly used in support of risk-informed and performance-based fire
protection at nuclear power plants [9]. In addition to an extensive study of the CFAST model,
the report compares the output of several other models ranging from simple hand calculations to
more complex computational fluid dynamics codes such as the Fire Dynamics Simulator (FDS)
developed by the National Institute of Standards and Technology (NIST) [10].

1.5 Governing Equations and Assumptions

The governing equations of CFAST are for conservation of mass and energy within the lower and
upper layers of connected compartments within a building. The momentum within any one zone
is assumed to be zero. Momentum between zones in adjacent compartments is accounted for in
terms of horizontal or vent flow equations (Bernoulli’s law). Other features of CFAST include:

• Compartment geometry: CFAST is generally limited to fire scenarios where the compartment
volumes are strongly stratified. The empirical correlations contained in CFAST were developed
for relatively uncluttered, flat ceilings in compartments that can be characterized as “rooms” as
opposed to corridors or vertical shafts. There are no hard limits on what kind of compartment
can or cannot be modeled in CFAST. The CFAST Validation Guide indicates the accuracy of
its predictions for compartments of various aspect ratios.

2
 • Heat Release Rate: CFAST does not predict fire growth on burning objects. The heat release
rate is specified by the user for one or more fires. There is a simple sub-model to limit the heat
release based on available oxygen.
• Radiation from fires is modeled with a simple point source approximation. This limits the
accuracy of the model within a few diameters of the fire. Calculation of radiative exchange
between compartments is not modeled.
• Mechanical ventilation is modeled by specifying volumetric flow rates into or out of compart-
ments. The overall HVAC (heating, ventilation, air conditioning) system is not modeled.
• Natural Ventilation and Leakage: The flow through vertical openings, like doors and windows,
is modeled using the Bernoulli equation for the pressure difference between two compartments.
Horizontal openings, like hatches, are treated with a single empirical correlation based on
pressure and density differences between upper and lower compartments. Leakage is modeled
by explicitly creating a small vertical or horizontal opening.
• Suppression: CFAST predicts sprinkler activation based on an empirical ceiling jet correlation
and activation model. A simple suppression model decreases the specified heat release rate.
The technical approach and assumptions of the model have been presented in the peer-reviewed
scientific literature [11, 12, 13] and conference proceedings [14]. CFAST has been reviewed and
included in industry-standard handbooks such as the SFPE Handbook [15] and referenced in spe-
cific standards, including NFPA 805 [16] and NFPA 551 [17].
Also, all documents released by NIST are required to go through an internal editorial review
and approval process. This process is designed to ensure compliance with the technical require-
ments, policy, and editorial quality required by NIST. The technical review includes a critical
evaluation of the technical content and methodology, statistical treatment of data, uncertainty anal-
ysis, use of appropriate reference data and units, and bibliographic references. CFAST manuals
are always first reviewed by a member of the Fire Research Division, then by the immediate su-
pervisor of the author of the document, then by the chief of the Fire Research Division, and finally
by a reader from outside the division. These reviewers are technical experts in the field. Once the
document has been reviewed, it is then brought before the Editorial Review Board (ERB), a body
of representatives from all the NIST laboratories. At least one reader is designated by the Board
for each document that it accepts for review. This last reader is selected based on technical com-
petence and impartiality. The reader is usually from outside the division producing the document
and is responsible for checking that the document conforms with NIST policy on units, uncertainty
and scope. This reader does not need to be a technical expert in fire or combustion.
Besides formal internal and peer review, CFAST is subjected to continuous scrutiny because it
is available to the general public and is used internationally by those involved in fire safety design
and postfire reconstruction. The source code for CFAST is also released publicly, and has been
used at various universities worldwide, both in the classroom as a teaching tool as well as for
research. As a result, flaws in the theoretical development and the computer program itself have
been identified and fixed. The user base continues to serve as a means to evaluate the model, which
is as important to its development as the formal internal and external peer review processes.
For each major release of CFAST, NIST has maintained a history of the source code which
goes back to March 1989. While it is not practical to reconstruct the programs for each release for

3
use with modern software tools and computer operating systems, the source code history allows
the developers to examine what changes were made at each release point. This provides detailed
documentation of the history of model development and is often useful to understand the impact
of changes to sub-models as the model continues to evolve.

1.6 Input Data Required to Run the Model

All of the data required to run the CFAST model reside in a single input file that the user generates.
The file consists of the following information:
• material properties (e.g., thermal conductivity, specific heat, density, thickness, heat of com-
bustion)
• compartment dimensions (height, width, length)
• construction materials of the compartment (e.g., concrete, gypsum)
• dimensions and positions of horizontal and vertical flow openings such as doors, windows, and
vents
• mechanical ventilation specifications
• fire properties (e.g., heat release rate, lower oxygen limit, and species production rates as a
function of time)
• sprinkler and detector specifications
• positions, sizes, and characteristics of targets
• specifications for visual output from the model
The input files are provided for the validation exercises described in the Validation Guide [7]. A
complete description of the input parameters can be found in the CFAST User’s Guide [6].
A comprehensive assessment of the numerical parameters (such as default time step or solu-
tion convergence criteria) and physical parameters (such as empirical constants for convective heat
transfer or plume entrainment) used in CFAST is not available in one document. Instead, specific
parameters have been tested in various verification and validation studies performed at NIST and
elsewhere. Numerical parameters are described in this Technical Reference Guide and are subject
to the internal review process at NIST, but many physical parameters are extracted from the litera-
ture and do not undergo a formal review. The model user is expected to assess the appropriateness
of default values provided by CFAST and make changes to the default values, if needed.

1.7 Model Results

The output of CFAST are the sensible variables that are needed for assessing the environment
in a building subjected to a fire. Once the simulation is complete, CFAST produces an output
file containing all of the solution variables. Typical outputs include (but are not limited to) the
following:

4
 • environmental conditions in the room (such as hot gas layer temperature; plume centerline
temperature; oxygen and smoke concentration; and ceiling, wall, and floor temperatures)

• heat transfer-related outputs to walls and targets (such as incident convective, radiative, and
total heat fluxes)

• fire intensity and flame height

• flow velocities through vents and openings

• detector and sprinkler activation times

1.8 Model Scenarios

While the governing transport equations are based on the fundamental conservation laws of mass
and energy, the fire-specific algorithms within CFAST are based on empirical correlations. These
correlations include fire plume and ceiling jet temperatures and velocities, vent flow rates, sprinkler
activation, and so on. These sub-models were developed independently of each other under ideal
conditions. CFAST combines these sub-models in such a way that there are no hard limits on when
a particular sub-model is appropriate or not. The decision as to whether CFAST is appropriate for
a given fire scenario is based primarily on the hundreds of experiments and thousands of point-
to-point comparisons between CFAST and measured quantities that are included in the CFAST
Validation Guide [7]. This document includes a list of the experiments and their important physical
attributes such as the nature of the fire, the aspect ratio of the compartment, the ventilation rate,
and the relative location of targets. For each quantity of interest, such as upper layer temperature or
target heat flux, there is a calculated bias factor and standard deviation that indicates the accuracy
of the model for the particular quantity of interest which is based on measurement uncertainty.
Thus, the CFAST Validation Guide indicates what fire scenarios are appropriate for CFAST, and
the degree of accuracy that can be expected for a particular type of prediction.
In addition to what is included in the CFAST Validation Guide, validation studies have been
performed by NIST grantees, students at universities, and engineering firms using the model. Be-
cause each organization has its own reasons for validating the model, the referenced papers and
reports do not follow any particular guidelines. Some of the works only provide a qualitative as-
sessment of the model, concluding that the model agreement with a particular experiment is “good”
or “reasonable.” Sometimes, the conclusion is that the model works well in certain cases, not as
well in others. These studies are included in the survey because the references are useful to other
model users who may have a similar application and are interested in qualitative assessment. It
is important to note that some of the papers point out flaws in early releases of CFAST that have
been corrected or improved in more recent releases. Some of the issues raised, however, are still
subjects of active research. Continued updates for CFAST are greatly influenced by the feedback
provided by users, often through publication of validation efforts.

1.9 Organization of this Document

The rest of this document is broken into several chapters as follows:

5
Chapter 2 presents the basic transport equations used by CFAST derived from the conservation
laws of mass and energy, along with the ideal gas law.

Chapter 3 describes the algorithms and empirical correlations used to represent fires in the model,
including combustion chemistry and heat release rate, plume entrainment, plume temperature,
and plume velocity.

Chapter 4 includes the empirical correlations used to estimate natural flow through doors and
windows (vertically-oriented vents), floor and ceiling vents (horizontally-oriented vents), and
mechanical ventilation systems.

Chapter 5 documents the calculation of heat transfer including radiation exchange between fires,
walls, gas layers, and objects within compartments, convection between gases and compart-
ment bounding surfaces or objects within compartments, and conduction in compartment bound-
ing surfaces and object within compartments.

Chapter 6 details calculations for fire sprinkler and heat detector activation, smoke detection, fire
suppression, and visibility through smoke.

6
Chapter 2

The Basic Transport Equations

The equations used to model pressure, layer height and temperatures in CFAST take the form of an
initial value problem for a system of ordinary differential equations. These equations are derived
from the principles of conservation of mass and energy using the ideal gas law as an equation of
state to relate temperature and density. These equations predict the evolution in time of the com-
partment pressure (at the floor), upper layer volume, and layer temperatures due to the net gain or
loss of mass and energy in these layers. The primary assumption of a zone model is that properties
such as temperature can be approximated throughout a control volume by a representative average
value. Many formulations based upon these assumptions can be derived [18]. Though equivalent
mathematically, these formulations differ in their numerical solution.
The exchange of mass and energy or heat between zones is due to physical phenomena such
as fire plumes, natural and forced ventilation, convective and radiative heat transfer, and so on.
For example, a vent exchanges mass and heat between zones in connected rooms, a fire plume
typically adds heat to the upper layer and transfers entrained mass and heat from the lower to
the upper layer, and convection transfers heat from the gas layers to the surrounding walls. The
momentum within any one zone is assumed to be zero. Momentum exchange between zones in
adjacent compartments is accounted for in terms of horizontal or vent flow equations (Bernoulli’s
law).

2.1 Model Equations

Each compartment is divided into two control volumes or zones, a relatively hot upper layer and a
relatively cool lower layer, as illustrated in Fig. 2.1. The gas temperature and density are assumed
constant in each layer. The compartment as a whole is assumed to have a single value of pressure,
P. All thermodynamic parameters are assumed to be constant. The specific heat at constant volume
and at constant pressure, cv and cp , the specific1 gas constant, R, and the ratio of specific heats,
γ, are related by γ = cp /cv and R = cp − cv . Regardless of the composition of the gas mixture,
cp = 1012 J/(kg·K) and γ = 1.4; thus,
γ −1 J
R= cp ≈ 289.14 (2.1)
γ kg · K
1 The specific gas constant is the universal gas constant divided by the molar mass of the gas mixture. In CFAST,
it is assumed that the molar mass of the mixture is that of air; thus, the specific gas constant is indeed constant.

7
 Upper Layer

Layer Interface
Natural Vent
Lower Layer Plume

Components of mass and

enthalpy entering or leaving
a control volume

Figure 2.1: Schematic of control volumes in a two-layer zone model.

The mass conservation equation simply asserts that the rate of change of the mass of layer i is
equal to the sum of mass source terms, ṁi :

dmi
= ṁi (2.2)
dt

Mass source terms include plume mass entrainment and supply/exhaust ventilation. Next, the first
law of thermodynamics states that the rate of change in the layer’s internal energy, cv mi Ti , is equal
to the sum of heat source terms, q̇i , minus the work associated with expansion or contraction of the
layer, P dVi /dt:
d dVi
(cv mi Ti ) = q̇i − P (2.3)
dt dt
Included in q̇i are the fire’s heat release rate, convective losses to walls, and radiation exchange.
The layer’s temperature, mass and volume are related to the compartment pressure by the ideal
gas law:
PVi = mi R Ti (2.4)

A system of ordinary differential equations for the compartment pressure, upper layer volume, and

8
upper and lower layer temperature can be derived using Eqs. 2.2–2.4 (for details see Appendix B):
dP γ −1
= (q̇l + q̇u ) (2.5)
dt V
 
dVu 1 dP
= (γ − 1) q̇u −Vu (2.6)
dt Pγ dt
 
dTu 1 dP
= q̇u − cp ṁu Tu +Vu (2.7)
dt cp mu dt
 
dTl 1 dP
= q̇l − cp ṁl Tl +Vl (2.8)
dt cp ml dt
An algebraic equation for each wall surface temperature (ceiling, upper wall, lower wall and
floor) is also solved:
∂ Tw (0,t)
q̇′′ + k =0 (2.9)
∂x
where ∂ Tw (0,t)/∂ x is the temperature gradient at the wall surface, q̇′′ is the net radiative and
convective heat flux from the adjacent gas layer (see Chapter 5 for a description of the heat flux
term) and for brevity the temperature wall profiles are referred to collectively as Tw (x,t). The
spatial coordinate x refers to the depth inside the wall; x = 0 is the wall surface. The temperature
gradient in Eq. (2.9) is evaluated by advancing a known temperature profile from told to t using the
heat conduction equation

∂ Tw k ∂ 2 Tw
= (2.10)
∂t c ρ ∂ x2
with a constant temperature boundary condition
Tw (0,t) = Tw (2.11)
where ρ, c and k are the density, specific heat and thermal conductivity of the wall material.

2.2 Solution Procedure

This section give a brief discussion of the procedure CFAST uses to solve the zone fire modeling
equations given in the previous section (for more details see Appendix C). CFAST uses the differ-
ential / algebraic solver DASSL [19, 20] to solve the differential equations Eqs. (2.5) through (2.8)
and the algebraic equation (once it is discretized) Eq. (2.9) for P, Tu , Vu , Tl and Tw . DASSL solves
these equations by formulating a root finding problem
F t, y, y′ = 0


(2.12)
where F is a vector valued function of residuals given by

 y′ − f(t, y) gas ode′ s
F(t, y, y′ ) = (2.13)
 q̇′′ + k ∂ Tw (0,t) gas/solid constraint
∂x

9
and where f is the right hand side of Eqs. (2.5) through (2.8), t is the time, y is the solution vector
containing the quantities P, Tu , Vu , Tl and Tw at time t and y′ is the derivative of these quantities at
time t.
To start the solution process, all temperatures are initialized to ambient, all (over) pressures are
initialized to zero, and the derivative components of y′ are set to zero. For more complex cases, an
initial y and y′ at t = 0 is calculated and supplied to DASSL.
One step in the solution process then is to advance the solution from tn to tn+1 i.e. to find
the solution of F(tn+1 , yn+1 , y′n+1 ) = 0 given that the solution for F(tn , yn , y′n ) = 0 is known. The
solver estimates values for yn+1 and y′n+1 at tn+1 . CFAST uses this information to compute the
residual F(tn+1 , yn+1 , y′n+1 ). Based upon these residual values, DASSL generates new values of
yn+1 and y′n+1 until F(tn+1 , yn+1 , y′n+1 ) is zero within a specified tolerance. DASSL performs
these iterations using an extension of Newton’s method using derivatives that are approximated
with backward differentiation formulae of orders one through five [19, 20].
The unknown wall surface temperatures, Tw are iterated until the q̇′′ flux term from the gas
layers matches the temperature gradient ∂ Tw (0,t)/∂ x at the wall surface according to Eq. (2.9).
The ∂ Tw (0,t)/∂ x term is determined by advancing the temperature profile from tn to tn+1 using a
value of Tw supplied by DASSL as a boundary condition for the heat conduction problem:

∂ Tw k ∂ 2 Tw
= (2.14)
∂t c ρ ∂ x2
Tw (0,t) = Tw (2.15)

where ρ, c and k are the density, specific heat and thermal conductivity of the wall material.
As discussed in Refs. [18] and [21], these equations are stiff because the compartment pressure
reacts to changing conditions much more rapidly than other solution variables. Stiff problems re-
quire special solution techniques since methods such as Runge-Kutta or Adams-Bashforth require
prohibitively small time steps in order to track the short time scale phenomena (pressure in our
case) found in fire modeling scenarios. Methods that use the Jacobian on the other hand have a
much larger stability region for stiff problems and are, thus, more efficient, stable and accurate
(see, for example, Ref. [22]).

2.3 Species Transport

The equation of state, Eq. (2.4), assumes that the molecular weight of the gaseous mixture through-
out the domain is approximately that of air, 29 g/mol. However, CFAST does track the products
of combustion and the depletion of oxygen in each zone. At the start of the simulation, the com-
position of each layer is set to ambient conditions. The initial temperature is specified by the user.
The oxygen mass fraction is 23 % (21 % volume fraction) and the nitrogen mass fraction is 77 %
(79 % volume fraction). The mass fraction of water vapor is specified by the user in terms of a rel-
ative humidity and the oxygen and nitrogen mass fractions are adjusted accordingly. All other gas
species are initially zero. As fuel is burned, product species are produced in direct proportion to the
rate of fuel consumption (the major products of combustion are determined from the specified fuel
molecule and the minor species yields are specified by the user). The mass fraction of oxygen can
limit the combustion rate as discussed in Section 3.2. Two special separate species are included

10
in the model – a generic toxic species and an arbitrary trace species. Both are excluded from the
overall mass balance, but they are generated by the fire and transported in a manner identical to the
other species.
Each unit mass of a species produced by a fire is carried in the flow to the various rooms and
accumulates in the layers. The species mass divided by the layer volume is the mass concentration.
Filters can be used in mechanical ventilation systems to remove species. The phenomenon has
been implemented in CFAST to remove trace species and soot. It is implemented by modifying
the source terms which describe gas flow. See Ref. [23] for an example on the use of filtering.
The calculation of radiation exchange in CFAST also depends in part on the species concentra-
tions calculated by the model (and thus the user inputs for species yields). There are two separate
radiation calculations performed by CFAST. The first is for thermal radiation as part of the overall
heat transfer calculation, discussed in Section 5.1. The second is for visible light extinction to
determine visibility, discussed in Section 6.2.

11
12
Chapter 3

The Fire Plume

Fires in CFAST are specified by the user in terms of a time-dependent heat release rate (HRR), an
effective fuel molecule, and the yields of the products of incomplete combustion like soot and CO.
Fires can be specified in multiple compartments and are treated as totally separate entities, with no
interaction of the plumes. These fires are generally referred to as “objects” and can be ignited at a
prescribed time, temperature or heat flux.
CFAST does not include a pyrolysis model to predict, as opposed to specify, the growth and
spread of the fire. Rather, the transient pyrolysis rates for each fire are prescribed by the user.
Whereas this approach does not directly calculate the increased pyrolysis due to radiative feedback
from the flame or compartment, in theory these effects could be prescribed by the user. For larger
fires, this can be an important consideration, and the specification used should reflect the actual
conditions as closely as possible.

3.1 Combustion Chemistry

The HRR of the fire is specified by the user, but it may be constrained by the availability of oxygen
in the compartment. The combustion of a hydrocarbon fuel is described by the following single-
step reaction:

CnC HnH OnO NnN ClnCl + νO2 O2 → νCO2 CO2 + νH2 O H2 O + νCO CO

+ νS Soot + νHCl HCl + νHCN HCN + νN2 N2 (3.1)

The user specifies the composition of the fuel molecule and the yields of soot and CO, yS and yCO ,
which are related to their stoichiometric coefficients as follows:
MF
νS = yS (3.2)
MS
MF
νCO = yCO (3.3)
MCO
 
MF
νHCN = min nN , yHCN (3.4)
MHCN

13
Under the assumption that all of the chlorine in the fuel is converted to HCl, the other stoichiomet-
ric coefficients are:

νCO2 = nC − (νCO + νHCN + νS ) (3.5)

nH − (νHCl + νHCN )
ν H2 O = (3.6)
2
νH2 O + νCO − nO
νO2 = νCO2 + (3.7)
2
νHCl = nCl (3.8)
nN − νHCN
ν N2 = (3.9)
2
Note that the nitrogen in the air acts only as a diluent. The yield of hydrogen chloride is based
solely on the composition of the fuel molecule. The yield of hydrogen cyanide is limited by the
nitrogen in the fuel molecule and any excess nitrogen in the fuel that is not converted to HCN
reverts back to elemental nitrogen. Finally, a user-specified trace species can be specified to follow
the transport that results from fire-induced flow for an arbitrary species. This may be of particular
interest for radiological releases [23], but may be useful for any trace amounts released by a fire.

3.2 Heat Release Rate

As fuel and oxygen are consumed, heat is released and various products of combustion are formed.
The heat is released as radiation and convected heat:

Q̇r = χr Q̇ (3.10)

Q̇c = (1 − χr ) Q̇ (3.11)
(3.12)

where Q̇ is the heat released by the fire. The parameters Q̇r and Q̇c are the heat released by radiation
and convection, respectively, and χr is the fraction of the fire’s heat release rate emitted as radiation.
The user specifies the heat release rate, Q̇, as the actual heat released, accounting for combus-
tion efficiency, along with a characteristic base diameter, D, which is used in the plume temperature
and mass entrainment correlations. The combustion efficiency, χa , is the fraction of the theoretical
energy that is actually released during combustion [24]. χa is a function of fuel type, scale, and vi-
tiation. For small fires, Tewarson provides measured values for specific fuels [25]. Within CFAST,
the user also specifies a radiative fraction which takes a default value of 0.35 ; i.e., 35 % of the fire’s
energy is released via radiation. For specific fuels, the work of Tewarson [25], McCaffrey [26],
or Koseki [27] is available for reference. The effects of scale, fuel type, vitiation, and combustion
efficiency are all important to the radiation released by a fire [28, 29]. The typical range for the
radiative fraction is from about 0.05 to 0.4. The assumed and constant values for the combustion
efficiency and radiative fraction may add uncertainty to the calculated results, so the heat release
rate and radiative fraction should be chosen carefully to best model the scenario of interest.

14
 Using the specified heat release rate of the fire, Q̇, and a user-specified the heat of combustion,
∆H, the model calculates the pyrolysis rate of fuel, ṁf :

Q̇
ṁf = (3.13)
∆H
In the event that the HRR is constrained by the availability of oxygen, it is assumed that the
pyrolysis rate does not change. However, only part of the pyrolyzed fuel burns and the HRR
becomes:  
Q̇ = min ṁf ∆H , ṁe YO2 CLOL ∆HO2 (3.14)
where ṁe is the entrainment rate, YO2 is the mass fraction of oxygen in the layer containing the fire,
∆HO2 is the heat of combustion based on oxygen consumption1 , and CLOL (Lower Oxygen Limit)
is the smoothing function ranging from 0 to 1:
 
tanh 800(YO2 −YO2 ,l ) − 4 + 1
CLOL ≈ (3.15)
2
The limiting oxygen mass fraction, YO2 ,l , is 0.15, by default. This value is not a function of temper-
ature.
Any unburned fuel is tracked by the model, and transported to the upper layer via entrainment
in the fire plume or to other compartments through any user-specified vents. Unburned fuel may
burn in the upper layer or at vents if sufficiently hot and if additional oxygen is available.

3.3 Plume Entrainment

A fire pumps mass and energy from the lower layer into the upper layer. The vertical flow of mass
through a horizontal plane at height z above the base of the fire is called the mass entrainment rate,
ṁe (z). The vertical flow of energy through this horizontal plane is given by Q̇c + ṁe (z) cp Tl . The
empirical correlation for the mass entrainment rate depends on whether the plume is unobstructed,
against a wall, or in a corner.

3.3.1 Unobstructed Plumes

The plume mass entrainment, ṁe (z), at a height z above the base of the fire is estimated using
Heskestad’s correlation [31]:
 2 1/3 2/3
!
g ρ∞ 1/3 2.9 Q̇ c
ṁe (z) = 0.196 Q̇c (z − z0 )5/3 1 + √
2/3 (3.16)
cp T∞ g cp ρ∞ T∞ (z − z0 )5/3
where z0 is a virtual origin defined as

z0 Q̇
= −1.02 + 1.4 Q̇∗ 2/5 ; Q̇∗ = √ (3.17)
D ρ∞ cp T∞ g D D2
1 Theheat of combustion based on oxygen consumption is taken to be 13.1 MJ/kg, representative of typical hy-
drocarbon fuels [30].

15
Note that the virtual origin is defined in terms of the total heat release rate of the fire, Q̇. Equa-
tion (3.16) is recommended above the mean flame height, L. Below the flame height, Hes-
kestad [31] recommends the following:

z L
ṁe (z) = ṁe (L) ; = −1.02 + 3.7 Q̇∗ 2/5 (3.18)
L D

where the mean flame height is defined as the distance from the fuel source to the top of the visible
flame where the intermittency is 0.5. A flame intermittency of 0.5 means that the visible flame is
above the mean 50 % of the time and below the mean 50 % of the time.

3.3.2 Wall and Corner Plumes

If the fire is located in a corner or against a wall2 , the unobstructed plume entrainment correlation,
Eq. (3.16), is modified. For a corner, it is assumed that the fire is mirrored in each wall face,
quadrupling the convective HRR, Q̇c , and doubling the base diameter, D. The entrainment rate,
ṁe (z), of this hypothetical larger fire is then divided by a factor of 4. The effective entrainment
rate is approximately 41/3 /4 ≈ 0.40 of its unobstructed value. Of course, the virtual origin, z0 , in
Eq. (3.16) is affected by the change in effective diameter as well.
√
If the fire is against a wall, the convective HRR is doubled, the diameter is multiplied by 2,
and the result of Eq. (3.16) is divided by 2. The effective entrainment rate is 21/3 /2 ≈ 0.63 of its
unobstructed value.

3.3.3 Weak Plumes

In CFAST, there is a constraint on the mass entrainment rate because the plume can rise only
so high for a given HRR. Early in a fire, the plume may not have sufficient energy to reach the
compartment ceiling. Therefore, a limit is placed on the entrainment rate. For the plume to be able
to penetrate the hot upper layer, the density of the gas in the plume must be less than the density
of the gas in the upper layer. This implies that the upper layer temperature must be less than the
plume temperature:
Q̇c + ṁe cp Tl
Tu < Tp ≈ (3.19)
ṁe cp

Rearranging terms yields a limit on the mass entrainment:

Q̇c
ṁe < (3.20)
cp (Tu − Tl )
2 CFAST assumes that the fire is in a corner or against a wall only when the user specifies the exact coordinates of
the corner or wall.

16
3.4 Plume Temperature and Velocity
The centerline plume temperature rise, ∆T0 (z), and velocity, u0 (z), at a height z above the base of
the fire is estimated using Heskestad’s correlations [31]:
 !1/3 
T∞ 2/3
∆T0 (z) = min 900 K , 9.1 2 2
Q̇c (z − z0 )−5/3  (3.21)
g cp ρ∞
"  1/3 #
g 1/3
u0 (z) = min u0,max , 3.4 Q̇c (z − z0 )−1/3 (3.22)
cp ρ∞ T∞
where the virtual origin, z0 , is defined in Eq. (3.17). It is assumed that the temperature and velocity
decrease following a Gaussian profile about the centerline:
"   #
r 2 T0 (z) 1/2
 
∆T (r, z) = ∆T0 (z) exp − ; σ∆T = 0.14 (z − z0 ) (3.23)
σ∆T T∞
"   #
r 2
u(r, z) = u0 (z) exp − ; σu ≈ 1.1 σ∆T (3.24)
σu
Note that the maximum plume temperature of 900 K is suggested by Heskestad [31]. It is also
suggested that the maximum centerline plume velocity is reached when the plume temperature rise
is 650 K, and is a weak function of the convective HRR:
!
g2/5
1/5
u0,max = 2.2 2/5
650 Q̇ c (3.25)
T∞ (cp ρ∞ )1/5

Note that if the velocity is expressed in units of m/s and the convective HRR in kW, then u0,max ≈
1/5
2 Q̇c .
Above the interface between the hot upper layer and the cooler lower layer, the temperature
and density of the entrained air is significantly different than that of the lower layer. To account for
this, the plume temperature correlation [31] is modified for values of z greater than the interface
height, zI :
!1/3
Tu 2/3
∆T0 (z) = 9.1 Q̇c (z − z′0 )−5/3 ; z > zI (3.26)
g c2p ρu2
where the modified virtual origin is given by:
 3/5
Tu
z′0 = zI − (zI − z0 ) (3.27)
Tl
Equations (3.26) and (3.27) are obtained by asserting that ∆T0 is continuous across the layer inter-
face and that Tu ρu ≈ Tl ρl .
The mass entrainment correlation, Eq. (3.16), is modified the same way above the layer inter-
face.

17
3.5 Ceiling Jet
The temperature and velocity just below the ceiling is typically greater than that of the upper layer
due to the presence of a ceiling jet. From the work of Alpert and Heskestad [32], the temperature
of an unconfined ceiling jet is given by:

 6.3
 ; r/H ≤ 0.2
∗ 2/3
Tcj (r) − T∞ = T∞ Q̇H (3.28)
 (0.225 + 0.27 r/H)−4/3 ; 0.2 < r/H < 4.0


To be consistent with the calculated plume temperature, this is adjusted so that the calculated plume
and ceiling jet temperatures are the same for r/H ≤ 0.2:

 ∆T0 (H)
 ; r/H ≤ 0.2
Tcj (r) − T∞ = (3.29)
 0.182 ∆T (H) (0.225 + 0.27 r/H)−4/3 ; 0.2 < r/H < 4.0

0

where ∆T0 (H) is the calculated centerline plume temperature rise at the compartment ceiling from
Eq. (3.21) and the constant 0.182 is (0.225 + 0.27 · 0.2)4/3 .
The radial velocity is given by:

 3.61
 ; r/H ≤ 0.17
p ∗ 1/3
vcj (r) = g H Q̇H (3.30)
 1.06 (r/H)−0.69 ; 0.17 < r/H < 4.0


where the heat release rate of the fire is contained within the non-dimensional expression:
Q̇
Q̇∗H = √ (3.31)
ρ∞ cp T∞ g H 5/2
To account for the presence of an upper layer, the background temperature, T∞ , in Eq. (3.29) is
replaced with the upper layer temperature, Tu . No adjustment need be made to Eq. (3.31) because
ρ∞ T∞ is assumed constant.
An estimate of the ceiling jet thickness, δ , where the excess temperature drops to 1/e of its
maximum value, is given by Motevalli and Marks [32]:
δ h  r i r
= 0.112 1 − exp −2.24 ; 0.26 ≤ ≤ 2.0 (3.32)
H H H
In a corridor of width, W , Delichatsios [32] suggests the following alternative to the ceiling
jet correlation, Eqs. (3.29) and (3.30). For a given distance down the corridor, x, from the plume
centerline, the excess temperature and velocity are estimated as:
 1/3 " #
∆Tcj (x) H  x   W 1/3
= 0.37 exp −0.16 (3.33)
∆Tp W H H
 1/6
q H
vcj (x) = 0.114 H ∆Tcj (x) (3.34)
W

18
where ∆Tp is the excess plume temperature at the ceiling. This correlation is applicable once the
plume has reached the full width of the corridor and the ceiling jet flow is parallel to the corridor
walls so that x > W /2. For x < W /2, the normal correlations apply.

19
20
Chapter 4

Ventilation

CFAST models three types of vent flow: natural flow through vertical vents (such as doors or
windows), natural flow through horizontal vents (such as ceiling holes or hatches), and forced flow
via mechanical ventilation. Forced flow can occur through either vertical or horizontal vents.

4.1 Wall Vents (Doors, Windows, Leakage)

Natural flow through windows and doors is governed by the vertical stratification of the pressure
difference across the opening [33]. The mass flow is calculated by dividing the opening into
discrete horizontal segments, each of which is bounded by the top or bottom of the opening, the
zone interface of either compartment, or the neutral plane, which is where the velocity changes
direction. This is shown schematically in Fig. 4.1.

4.1.1 Calculating Flow Through Wall Vents

Let z = b and z = t denote the height of the bottom and top of the segment, and ∆Pb and ∆Pt denote
the pressure differences at these heights. Because a given segment is either completely above or
completely below the neutral plane, the two pressure differences will have the same sign. The mass
flow through the segment can then be computed by integrating Bernoulli’s equation from b to t:
Z t p
ṁ = C 2ρ ∆P(z) w dz (4.1)
b
r
p Z t |(t − z) ∆Pb + (z − b) ∆Pt |
= C 2ρ w dz (4.2)
b t −b
2 p |∆Pt |3/2 − |∆Pb |3/2
= C 2ρ w (t − b) (4.3)
3 |∆Pt | − |∆Pb |
Here, C is the orifice coefficient taken to be 0.7 [34], ρ is the gas density of the upwind compart-
ment, w is the width of the opening, and ∆P(z) is the pressure across the interface at elevation z.
Note the use of the integral formula
Z √
2
A + Bz dz = (A + Bz)3/2 + constant (4.4)
3B

21
 Compartment 1 Compartment 2

ṁu→u -

Layer Interface
ṁu→l

ṁl→u Neutral Plane

Layer Interface

 ṁl→l

Figure 4.1: Flow patterns for horizontal flow through a vertical vent.

where A = (|t ∆Pt | − b |∆Pb |)/(t − b) and B = (|∆Pt | − |∆Pb |)/(t − b). Equation 4.3 can be written:
p
2 p |∆Pt | + |∆Pt ∆Pb | + |∆Pb |
ṁ = C 2ρ w (t − b) p p (4.5)
3 |∆Pt | + |∆Pb |

This is the way it is written in Ref. [33].

Figure 4.1 indicates schematically how the various mass flows through the opening are dis-
tributed. For the flow originating in the upper layer of the upstream compartment flowing into the
upper layer of the downstream compartment, ṁu→u , or the flow from the lower layer to the lower
layer, ṁl→l , the mass is applied directly to the downstream layer.
The mass flow from the upper layer of the upstream compartment to the lower layer of the
downstream, ṁu→l , is assumed to rise into the upper layer via a spill plume. The heat flow rate of
the plume is:

ḣu→l = cp Tu,1 − Tl,2 ṁu→l (4.6)
Assuming that Tu,1 > Tl,2 , the mass entrainment of the spill plume is given by Poreh et al. [35]:
!1/3
 2/3 2
g ρl,2
Tl,2 1/3
ṁe,p = ṁu→l + Cm ḣu→l w2/3 (zI − zN ) (4.7)
Tu,1 cp Tl,2

where Cm is an empirical constant equal to 0.44, w is the width of the opening, zI is the height of
the layer interface, and zN is the height of the neutral plane. Suggested values for Cm vary between

22
0.44 and 0.66, all of which were determined empirically in a number of different experimental
configurations.
For the mass flow leaving the lower layer of the upstream compartment and entering the upper
layer of the downstream compartment, ṁl→u , the shear flow causes vortex shedding that entrains
upper layer gas and deposits it in the lower layer. It is assumed that the incoming cold plume
behaves like the inverse of the usual jet between adjacent hot layers; forming a descending plume.
The same equations are used to calculate this inverse plume as are used for the upright vent mixing,
above.

4.2 Ceiling and Floor Vents

Cooper [36] developed a semi-empirical correlation for ceiling vents in which the volumetric flow
rates upwards and downwards are functions of both pressure and density differences:
p
V̇up = 0.68 Av 2 | max(∆P, 0)|/ρbot + V̇ex (4.8)
q
V̇down = 0.68 Av 2 | min(∆P, 0)|/ρtop + V̇ex (4.9)

Here, ∆P is the lower compartment pressure minus the upper, Av is the area of the vent, and ρtop
and ρbot are the layer densities adjacent to the vent. The term, V̇ex , represents the density-driven
exchange flow; that is, when ∆ρ = ρtop − ρbot > 0 and |∆P| is relatively small, there is an exchange
of mass and heat at the vent interface where hot gas from below mixes with cooler gases above:
!1/2 
5/2
Cs2 g ∆ρ D5

g ∆ρ Av |∆P|
V̇ex = 0.10 1− ; |∆P| < |∆Pflood | ≡ (4.10)
ρavg |∆Pflood | 2 A2v
p
Here, ρavg is the average of the densities above and below the vent, D = 2 Av /π, and the shape
factor, Cs , is 0.754 for round and 0.942 for square openings.
As with wall vents, mass and heat are exchanged between the layers of the upper and lower
compartments. Take the typical case where the lower compartment has a hot gas layer venting into
the upper compartment. Consider first the relative amounts of mass and heat extracted from the
layers of the lower compartment, and then consider where this mass and heat are deposited in the
upper compartment. First, the mass and heat are extracted from the lower and upper layers of the
lower compartment according to the following weighted average:

ṁup = α ρu V̇up + (1 − α) ρl V̇up (4.11)

ḣup = α cp ρu V̇up Tu + (1 − α) cp ρl V̇up Tl (4.12)

where ρu and ρl refer to the upper and lower layers of the lower compartment.
The weighting factor, α, indicates the relative amounts of mass extracted from the upper and
lower layers of the lower compartment. If the hot gas layer of the lower compartment is sufficiently
deep, all of the gas passing into the upper compartment is extracted from the upper layer. If the
layer is not sufficiently deep, gases from the lower layer are also drawn through the vent, which
reduces the amount of smoke that is exhausted from the compartment. This phenomenon is referred

23
to as plug-holing. Cooper [37] suggests that the degree of plug-holing is a function of the following
form of the Froude number:
  −1/2
5 Tu − T∞
Fr = V̇ g (H − zI ) (4.13)
T∞

where V̇ is the volume flow through the vent and (H − zI ) is the depth of the hot gas layer. Using
the CFD model Fire Dynamics Simulator (FDS), a correlation was developed that defines α as a
function of Fr:
2
α = e−(Fr/2) (4.14)
Figure 4.2 displays the FDS results and correlation.

1
FDS
Fraction of Upper Layer Mass

Best Fit
0.8

0.6

0.4

0.2

0
0 1 2 3 4 5
Froude Number

Figure 4.2: Relative fraction of upper layer gases extracted via a ceiling vent.

The previous paragraph explains from where the gas flowing through the vent is extracted.
Where the gas goes is more simple. If the temperature of the hot gas layer of the lower compartment
is greater than the temperature of the lower layer of the upper compartment, then the mass and heat
are deposited into the upper layer of the upper compartment. Otherwise, the mass and heat are
deposited into the lower layer of the upper compartment.

4.3 Forced Flow

CFAST models mechanical ventilation in terms of user-specified volume flows at various points in
the compartment. The model does not include duct work or fan curves. These equations are high-
order, non-linear and in some cases ill-posed, which caused a great deal of difficulty in reaching a
numerical solution.

24
 The flow through mechanical vents can be filtered. Filtering affects particulates such as smoke
and the trace species. Filtering can be turned on at any time. Effectiveness is from 0 % (no effect)
to 100 % which completely blocks the flow of these two species.

4.4 Leakage
CFAST can automatically calculate leakage between one or more compartments and the outdoors.
Leakage is specified for each desired compartment as either

1. A total leakage area over the wall surfaces or leakage area per unit wall area for wall leakage.
For wall leakage, the model creates a vent on the wall with a sill height of 5 % of the com-
partment ceiling height and a soffit of 95 % of the compartment ceiling height. The width of
the vent is determined from a calculated vent area (the specified area or the specified leakage
area ratio times the total wall area of the compartment) divided by the vent height (90 % of the
compartment ceiling height).

2. A total leakage area over the floor surface or leakage area per unit floor area for floor leaks.
For floor leakage, the model creates a vent on the wall with a width of 90 % of the average wall
width (the sum of the compartment width and the compartment depth divided by two) with
the sill at the floor and the soffit height from a calculated vent area (the specified area or the
specified leakage area ratio times the total floor area of the compartment) divided by the vent
width (90 % of the average compartment wall width).

Flow is calculated from Equation 4.3. CFAST assumes a vent flow coefficient of 0.7.

25
26
Chapter 5

Heat Transfer

This section discusses thermal radiation, convection, and conduction, the three mechanisms by
which heat is transferred between the gas layers and the enclosing compartment walls. Hot gases
exchange heat with solid surfaces via convection and radiation. Heat is transferred through solids
via conduction. Different material properties can be used for the ceiling, floor, and walls of each
compartment (although all the walls of a compartment must be the same). Additionally, each
surface can be composed of up to three distinct layers. This allows the user to deal naturally with
the actual building construction. Material thermophysical properties are assumed to be constant.
Radiative transfer occurs among the fire(s), gas layers and compartment surfaces (ceiling, walls
and floor). This transfer is a function of the temperature differences and the emissivity of the
gas layers as well as the compartment surfaces. Typical surface emissivity values only vary over
a small range. For the gas layers, however, the emissivity is a function of the concentration of
species which are strong radiators, predominately smoke particulates, carbon dioxide, and water.

5.1 Radiation
Radiation heat transfer is calculated between the ceiling, floor, wall layers, and fire, with the inclu-
sion of emission and absorption by the hot gas layer [38]. The following assumptions are made:
• Each gas layer and each wall segment is assumed to be at a uniform temperature.
• The wall and gas layer temperatures are assumed to change slowly over the duration of the
time step of the governing equations.
• The fire is assumed to radiate uniformly in all directions emitting a fraction, χr , of the total heat
release rate. This radiation is assumed to originate from a single point. Radiation feedback to
the fire and radiation from the plume is not modeled in the radiation exchange algorithm.
• The radiation emitted is assumed to be diffuse and gray. In other words, the radiant fluxes
emitted are independent of direction and wavelength. At a solid surface, the emittance, ε, ab-
sorptance, α, and reflectance, ρ, are related via ε = α = 1 − ρ. In the gas phase, the emittance,
ε, absorptance, α, and transmittance, τ, are related via ε = α = 1 − τ.
• Rooms or compartments are assumed to be rectangular boxes. Each wall is either perpendicular
or parallel to every other wall. Radiation transfer through vent openings is lost from the room.

27
 The compartment lining is divided into four surfaces: the ceiling, the floor, and the wall sec-
tions above and below the layer interface. The outgoing radiant flux at surface i consists of an
emissive power and reflectance term given by

q′′out,i = σ εi Ti4 + (1 − εi ) q′′in,i (5.1)

where q′′in,i is a weighted average of all outgoing radiant fluxes from sources such as wall segments,
gas layers and fires. This incoming radiant flux is given by
!
4
1
q′′in,i = ∑ A j q′′out, j Fj−i τ j−i + c′′i (5.2)
Ai j=1

which simplifies to !
4
q′′in,i = ∑ q′′out, j Fi− j τ j−i + c′′i (5.3)
j=1

after noting that Ai Fi− j = A j Fj−i . The term Fj−i is the configuration factor (fraction of radiant
energy emitted by surface j that is intercepted by surface i), τ j−i is the transmittance, σ is the
Stefan-Boltzman constant, εi is the emissivity, Ti is the temperature and c′′i is the radiative flux
from gas layers and fire of surface i .
The problem then is to determine the net radiative flux, q̇′′r,i , at each of these surfaces. This flux
is defined as the difference between the incoming and outgoing radiative flux1 :

q̇′′r,i = q′′in,i − q′′out,i (5.4)

Following Eq. (17-20) in Siegel and Howell [39], Eq. (5.4) is written as a system of four linear
equations (after using Eqs. (5.1) and (5.3) to eliminate q′′out,i and q′′in,i ):

q̇′′r,i 4 1−εj 4
Fi− j τ j−i q̇′′r, j = σ Ti4 − ∑ Fi− j τ j−i σ T j4 − c′′i


− +∑ (5.5)
εi j=1 ε j j=1

The radiation from the gases in the upper and lower layers and the fire is included in the last term:
2
ωi−f χr Q̇
c′′i = ∑ ε j Fi− j σ Tj4 + (5.6)
j=1 Ai 4π

where εj is the emittance (absorptance) of the layer, Fi− j is the view factor between the layer and
solid surface, ωi−f is the normalized solid angle between the fire and wall2 , and Q̇ is the heat
release rate of the fire3 . If the solid surface, i, is the floor or the lower wall and the gas layer, j,
is the upper layer, the view factor, Fi− j , refers to the layer interface. If the solid surface is the
1 The sign for the net radiative flux, q̇′′ , is defined so that positive radiant fluxes heat up wall segments, consistent
r,i
with how convective fluxes are defined. Note that this is opposite to the sign convention used in Siegel and Howell [39].
2 Note that as the area of surface i shrinks to zero, ω /A → 1/R2 , yielding the classic point source radiation
i−f i
model.
3 The location of the fire is taken to be a point located at a height of 1/3 of the height of the fire calculated from

Heskestad’s flame height calculation, Eq. (3.18)

28
upper wall or ceiling, the view factor is 1. Eq. (5.5) is a set of linear equations (4 equations and 4
unknowns) whose solution is discussed in the next section.
The net radiative flux, q̇′′r,i found in Eq. (5.5), along with the convective flux discussed in Sec-
tion 5.2 is used as a boundary condition for computing heat conduction within the solid walls. The
outgoing radiative flux, q′′out,i is used in Section 5.1.5 for computing radiative flux to a target. It
may be rewritten in terms of known quantities, in particular the net radiative flux found in Eq.
(5.5), by eliminating q′′in,i in Eqs. 5.1 and 5.4 and is given by

1 − εi ′′
q′′out,i = σ Ti4 + q̇ (5.7)
εi r,i

5.1.1 Solving the Radiation Equations

Equation (5.5) is a system of linear equations of the form

Aq = Be−c (5.8)

where A and B are 4 × 4 matrices, q is the unknown vector of net radiative fluxes, and e is the
vector of emission terms:

ai j = δi j − Fi− j τi− j (1 − ε j ) (5.9)

bi j = δi j − Fi− j τi− j (5.10)
q j = −q̇′′r, j /ε j (5.11)
e j = σ T j4 (5.12)

The components of the source term vector, c j , are given by Eq. (5.6). The term δi j is the usual
Kronecker delta function:

 1 i= j
δi j =
 0 i ̸= j

5.1.2 Configuration Factors

The configuration factor, Fi− j , is the fraction of radiant energy emitted by surface i that is inter-
cepted by surface j. Sixteen configuration factors are required for each compartment. Figure 5.1
depicts the system of indices for the wall surfaces and layer interface. In general, these factors may
be computed using
1 cos θi cos θ j
Z Z
Fi− j = dA j dAi (5.13)
Ai Ai A j πL2
where L is the distance along the line of integration, θi and θ j are the angles for surface i and j
between the respective normal vectors and the line of integration, and Ai and A j are the areas of
the two surfaces.
A more efficient procedure is to compute configuration factors between the ceiling and floor,
F1−4 , between the ceiling and layer interface, F1−d , and between the floor and layer interface, F4−d ,

29
 1 - Ceiling
2 - Upper Wall

d - Layer Interface

3 - Lower Wall

4 - Floor
Figure 5.1: Schematic diagram of a compartment with indices for computing configuration factors.

using a formula for the configuration factor between parallel plates (Appendix C of Ref. [39]):
(  1/2
2 (1 + X 2 )(1 +Y 2 ) p X
Fi− j = ln 2 2
+ X 1 +Y 2 tan−1 √ +
πXY 1 + X +Y 1 +Y 2

p
−1 Y −1 −1
Y 1 + X 2 tan √ − X tan X −Y tan Y (5.14)
1 + X2
where a and b are the rectangle dimensions, c is the separation distance, X = a/c and Y = b/c.
Algebraic relations in terms of F1−4 , F1−d and F4−d may then be used to determine the 16 re-
quired configuration factors. Two properties are used to derive these relations. First, the symmetry
relation,
Ai Fi− j = A j Fj−i (5.15)
follows from Eq. (5.13). Second, configuration factors for surfaces forming an enclosure satisfy
4
∑ Fi− j = 1 (5.16)
j=1

The configuration factor F1−4 does not change during a simulation since its value depends only
on the compartment height and floor/ceiling area. Therefore, F1−4 need only be computed once.
Configuration factors F1−d and F4−d depend on the layer interface height so need to be determined
each time the radiation exchange is calculated.
F1−4 , F1−d and F4−d are determined using Eq. (5.14). Since A1 = A4 it follows that F4−1 =
F1−4 . The other 14 configuration factors can be calculated using simple algebraic formulas. Since
the floor and the ceiling are assumed to be flat rectangular surfaces, it follows that F1−1 = F4−4 = 0.
From Eq. (5.16) and the symmetry condition, F2−1 = F2−d , it follows that
F1−2 + F1−d = 1 (5.17)
A1
F2−1 + F2−2 + F2−d = 2 F2−1 + F2−2 = 2 F1−2 + F2−2 = 1 (5.18)
A2
Solving equations (5.17) and (5.18) for F1−2 and F2−2 gives
F1−2 = 1 − F1−d
F2−2 = 1 − 2 F2−1

30
Similarly,

F4−3 = 1 − F4−d
F3−3 = 1 − 2 F3−4

Using the above configuration factors and Eq. (5.16), it follows that

F1−3 = 1 − F1−4 − F1−2

F3−2 = 1 − F3−1 − F3−3 − F3−4
F2−4 = 1 − F2−1 − F2−2 − F2−3

5.1.3 Solid Angles

The normalized solid angle ω in Eq. (5.6) is the view factor between a given point, say the fire,
and a solid wall. In other words, it is the fraction of radiant energy emitted by the fire that is
intercepted by a particular wall surface. Normalized solid angles are also used to determine the
fraction of radiation from each wall surface that strikes a target as in Eq. (5.21).

AA
K 
A
A
A
A
A v3
v2A
A
A
A
H
Y
HH A
H A
v1 H A
H
H
HA
Figure 5.2: Solid angle formed by the vectors v1 , v2 , and v3 .

To compute a solid angle, consider the triangle defined by the vertices v1 , v2 , and v3 as illus-
trated in Fig. 5.2. The solid angle of this triangle with respect to the origin is computed
 
1 −1 |v̂1 × v̂2 · v̂3 |
ω= tan
2π 1 + v̂1 · v̂2 + v̂1 · v̂3 + v̂2 · v̂3
where v̂ denotes the normalized vector [40]. The normalized solid angle for a rectangular wall
surface is computed by splitting the surface into two triangles, computing the solid angle for each
triangle and summing.

5.1.4 Transmittance and Absorptance

The transmittance is the fraction of radiant energy that will pass through a volume filled with an
absorbing medium. It is usually expressed in the form:

τ = e−aL (5.19)

31
where a is the absorption coefficient and L is the path length. The absorptance is the fraction of
radiant energy absorbed by that volume. For a gray gas, α + τ = 1.
In general, the transmittance and absorptance are functions of wavelength. This is an important
factor to consider for the major gaseous products, CO2 and H2 O. However, soot has a continuous
absorption spectrum that allows the transmittance and absorptance to be approximated as “gray”
[39] across the entire spectrum. The total transmittance over a path length L through a volume of
combustion products is taken as the product of the transmittance of the soot and major gas phase
products:
τ = e−as L (1 − αH2 O − 0.5 αCO2 ) (5.20)
The factor of 0.5 applied to the absorptance of CO2 accounts for the overlap of the wavelength
bands of the two gases. Tien et al. [41] suggest that the absorption coefficient for soot may be
approximated as as = k fv T where k is a constant that depends on the optical properties of the soot
particles, fv is the soot volume fraction, and T is the (absolute) temperature. Values of k have been
found to be about constant for a wide range of fuels [42].
Absorptance data for H2 O and CO2 are reported in Ref. [43]. For each gas, these data are tab-
ulated and implemented as a two-dimensional array based on temperature and gas concentration.
The effective path length, L, is calculated between the center-points of the exchanging surfaces
(i.e., from the center height of a fire to the center of a wall surface or between the center points of
two compartment surfaces). Path lengths are calculated separately for the upper and lower layers
in a compartment.

5.1.5 Radiation Heat Flux to a Target

Objects, often referred to as “targets,” may be added to any compartment to represent measurement
devices or equipment that might be damaged in a fire. Targets absorb thermal radiation from the
fire, walls and hot gas layer, but targets do not affect the fire simulation. That is, there is a one-way
coupling between the target and the fires, 10 compartment surfaces (ceiling, floor, 4 walls–upper
and lower) and 2 gas layers. Radiation from the fire is taken to be from a point source located
at a height of 1/3 of the height of the fire calculated from Heskestad’s flame height calculation,
Eq. (3.18). The net radiative heat flux to a target is given by:

fires walls
cos(θf )
q̇′′r = εt ∑ 4πR2 u l r f ∑ τuτl Fw εw q′′out,w +
τ τ χ Q̇ +
f f w
 (5.21)
cos(θt ) 4 cos(θt ) 4 4
τu αl σ Tl + τl αu σ Tu − σ Tt
4πRt2 4πRt2
where εt is the emissivity of the target, εw is the emissivity of the wall surface, θf is the angle
formed by the line of sight segment (between the fire f and the target) of length Rf and the normal
direction to the target surface, Fw is the view factor for the wall surface, τu and τl are the values of
transmittance for the path between the target and the fires or walls in the upper and lower layers,
αu and αl are (1 − τu ) and (1 − τl ), and Tt is the temperature of the target4 . The term q′′out,w , given
by Eq. (5.7), is the outgoing radiative flux at the w’th wall surface.
4 In cos(θf ) cos(θf )
the limit where Rf approaches zero, 4πR2f
reduces to 2 .

32
5.2 Convection
The transfer of heat between the gas and solid surfaces is handled slightly differently at the ceiling,
floor and walls, due to the difference in orientation and the presence of a relatively thin hot flow
near the ceiling known as the ceiling jet. The following two sections describe how the convective
heat transfer is done for these different surfaces.

5.2.1 Walls and Floor

In general, the convective heat flux to a solid surface is given by:

q̇′′c = h Tg − Ts


(5.22)

The convective heat transfer coefficient, h, is a function of the gas properties, temperature, and
velocity. In CFAST, simple correlations for natural convection are used, since the gas velocity is
unknown:
h = C|Tg − Ts |1/3 (5.23)
where C is an empirical coefficient (1.52 for the floor and ceiling (in the absence of a ceiling jet)
and 1.31 for the walls [44]), Tg is the average gas layer temperature adjacent to the surface, and Ts
is the surface temperature.

5.2.2 Ceiling
During the early stage of a fire before a hot gas layer has formed, the convective heat transfer to the
ceiling is governed by the temperature and velocity of the ceiling jet. Alpert’s chapter in the SFPE
Handbook [32] presents an empirical correlation for the convective heat flux from the ceiling jet to
a relatively cool surface:
Q̇c  r −1.36
q̇′′c = 1.323 f 2 (5.24)
H H
where f is a friction factor estimated to be 0.03, r is the radial distance to the plume centerline, H is
the ceiling height, and Q̇c is the convective fraction of the heat release rate. The average convective
heat flux to the ceiling can be obtained by integrating this expression over the entire ceiling:
Z 2π Z R
1 0.27 Q̇c
q̇′′c,avg = q̇′′c r dr dθ = (5.25)
LW 0 0 (LW )0.68 H 0.64

Note that the integration is carried out over a circle whose area, πR2 , is taken as equal to the area
of the ceiling, LW .
Equation (5.25) applies to the early stage of the fire; thus, a modified heat transfer coefficient
is used so that there is a transition from the early to later stages when a layer has formed:
 ′′
q̇c,avg

1/3
h = max , C|Tu − Ts | (5.26)
Tu − Ts
Here, Tu is the average temperature of the upper layer and Ts is the ceiling surface temperature.
Notice that the rightmost term is simply the correlation used for the walls and floor.

33
5.3 Heat Conduction within Solid Walls or Targets
The heat conduction equation is solved in the direction normal to solid targets or wall surfaces
using non-uniformly spaced nodes and a second-order accurate central difference scheme for the
spatial derivatives.
Wall surface temperatures are updated in time implicitly using the solver DASSL which cou-
ples wall surface temperatures with gas solution variables (pressures, layer volumes and temper-
atures). Target temperatures on the other hand are updated in time using a semi-implicit time
marching scheme. Since they do not affect compartment conditions significantly, there is no need
to couple them with other solution variables as is done with wall temperatures.
At each time step, the wall temperature profiles are updated in time until the net convective and
radiative heat flux striking the wall equals with the heat flux into the solid [45]:

dT
q̇′′ ≡ q̇′′r + q̇′′c = −k (5.27)
dx x=0

where k is the thermal conductivity of the wall material. This solution strategy requires a differen-
tial algebraic equation (DAE) solver that can simultaneously solve both differential and algebraic
equations. With this method, only one or two extra equations are required per wall segment (two if
both the interior and exterior wall segment surface temperatures are computed). This solution strat-
egy is more efficient than the method of lines since fewer equations need to be solved. Conduction
is then coupled to the gas phase energy exchange.
Wall and target temperature profiles are advanced in time using the heat equation. The heat
conduction equation normal to the solid surface is:

∂T k ∂ 2T
= (5.28)
∂t ρc ∂ x2

where k, ρ and c are the thermal conductivity, density and heat capacity of the target or wall
material. The boundary condition used at the front wall surface is :

Tw (0,t) = Tw (5.29)

The boundary condition used at the target surface is:

dT
q̇′′ = −k (5.30)
dx
where q̇′′ is the net convective and radiative heat flux.
A non-uniform array of internal nodes is used to capture steep gradients in temperature near
the surface. Define a penetration depth of
√
x p = 2 α tend erfc−1 (0.05) (5.31)

where erfc−1 denotes the inverse of the complementary error function. The value x p is the location
in a semi-infinite wall where the temperature rise is 5 % after tend seconds. Eighty percent of the
nodes are placed on the interior side of x p and the remaining 20 % are placed on the exterior side.

34
 The 1-D heat conduction equation can be solved in either Cartesian or cylindrical coordinates.
The solution methodology shall be presented for cylindrical coordinates:
 
∂T k 1 ∂ ∂T
= r (5.32)
∂t ρc r ∂ r ∂r
Dividing the cylinder into N uniformly spaced concentric control volumes, this equation can be
written in discretized form:
" ! ! #
n+1 n+1 n+1 n+1
∆t k Ti+1 − Ti r i Ti − Ti−1 ri−1
Tin+1 − Tin = −
∆r ρc ∆r ri−1/2 ∆r ri−1/2

    
∆t α n+1 i
i−1
− Tin+1 n+1 n+1


= Ti+1 − Ti − Ti−1 (5.33)
∆r2 i − 0.5 i − 0.5

where α = k/(ρc). Defining Ci and Di as

   
α∆t i−1 α∆t i
Ci = 2 ; Di = 2 (5.34)
∆r i − 0.5 ∆r i − 0.5
Equation (5.33) can be written:
 
α∆t
n+1
−Ci Ti−1 + 1 + 2 2 Tin+1 − Di Ti+1
n+1
= Tin i = 1...N − 1 (5.35)
∆r
The boundary condition is applied at control volume N:

α∆t ∆r q̇′′ N
 
n+1 n n+1 n+1 N − 1
TN − TN = 2 − (TN − TN−1 )
∆r k N − 0.5 N − 0.5
or
n+1 ∆r ′′
−CN TN−1 + (1 +CN ) TNn+1 = TNn + DN q̇ (5.36)
k
The internal temperature profile, Ti , is then obtained with a tri-diagonal linear solver.

5.4 Coupling the Gas and Solid Phase Calculations

To illustrate the method, consider a one room case with one active wall. There are four gas phase
equations (pressure, upper layer volume, upper and lower layer temperatures) and one wall temper-
ature equation. Implementation of the gradient matching method requires that storage be allocated
for the temperature profiles at the current time step, t, and at the next time step, t + ∆t. Given the
profile at time t and values for the five unknowns at time t + ∆t (initial guess by the solver), the
temperature profile is advanced from time t to t + ∆t. The temperature gradient at the wall surface
is computed followed by the residuals for the five equations. The DAE solver adjusts the solution
variables and the time step until the residuals for all the equations are below an error tolerance.
Once the solver has completed the step, the array storing the temperature profile for the previous
time is updated, and the DAE solver is ready to take its next step.

35
 Heat transfer between connected compartments is modeled by merging the back surfaces of the
connected ceiling and floor of the compartments or the back wall surfaces of the connected horizon-
tal compartments. A heat conduction problem is solved for the merged walls using a temperature
boundary condition for both the near and far wall. As before, temperatures are determined by the
DAE solver so that the heat flux striking the wall surface (both interior and exterior) is consistent
with the temperature gradient at that surface.
For horizontal heat transfer between compartments, the connections may be between partial
wall surfaces, expressed as a fraction of the wall surface. CFAST first estimates conduction frac-
tions analogous to radiation configuration factors. For example, if only one half of the rear wall
in one compartment is adjacent to the front wall in a second compartment, the conduction fraction
between the two compartments is 0.5. Once these fractions are determined, an average flux, q̇′′avg ,
is calculated using
q̇′′avg = ∑ Fi− j q̇′′j (5.37)
walls

where Fi j is the fraction of flux from wall i that contributes to wall j, q̇′′j is the flux striking wall j.

36
Chapter 6

Fire Protection Calculations

6.1 Sprinkler and Heat Detector Activation

The link temperature of a sprinkler or heat detector is modeled using the differential equation [46]:
√
dTL v

= Tg − TL (6.1)
dt RTI
where TL and Tg are the link and gas temperatures, v is the gas speed, and RTI (Response Time
Index) is a measure of the sensor’s thermal inertia. The gas temperature and speed are obtained
from the plume algorithm (Section 3.4) and the ceiling jet algorithm, described below. Rooms
without fires do not have ceiling jets, in which case the upper layer temperature is used, along with
a fixed speed of 0.1 m/s. This value is simply an order of magnitude estimate. The link and gas
temperatures and the speed are functions of time; the RTI is a constant for a given detector type.
The detector equation is solved numerically using the semi-implicit updating scheme:
√ √ !
TLn+1 − TLn 1 vn n vn+1
T − TLn + Tgn+1 − TLn+1



= (6.2)
∆t 2 RTI g RTI
where the superscript n denotes the value at the current time, and ∆t is the time step.
The gas temperature is estimated at the detector or sprinkler location. If the detector or sprinkler
is located within the plume, the gas temperature and velocity are estimated from section 3.4. If the
detector or sprinkler is located within the ceiling jet, the gas temperature and velocity are estimated
from section 3.5. If the detector or sprinkler is below the ceiling jet layer or in a compartment
without a fire, the gas temperature is taken to be the appropriate layer temperature with a default
speed of 0.1 m/s. The basis of this last value is simply an order of magnitude estimate.

6.2 Visibility
The visibility calculation depends solely on the soot concentration. Soot production is specified
by the user in terms of a yield. Soot is transported like all other gas species and its concentration
is used in the optical density calculation. The optical density per unit distance, Du , is given by the
expression:
Km m′′′
s
Du = (6.3)
ln 10

37
where m′′′
s is the mass of soot per unit volume and Km is the specific extinction coefficient. The
default value is 8700 m2 /kg based on the recommendation of Mulholland [47]. From the optical
density, the percent obscuration per unit distance, Ou , can be calculated from the definitions of
optical density and percent obscuration as a function of light attenuation [46] as
 1 !
10Du − 1
 
1 I0 I d
Du = log10 , Ou = 100 1 − =⇒ Ou = 100 (6.4)
d I I0 10Du

which is used to determine smoke detector activation, below.

6.3 Smoke Detection

CFAST estimates the activation of smoke detectors based on either the smoke concentration or
temperature at the detector location.

• For smoke detector response based on smoke obscuration, the calculated percent obscuration
in the layer where the detector is located is compared to a user-specified value. A default acti-
vation value of 24 %/m (8 %/ft) obscuration is assumed for both ionization and photoelectric
smoke detectors [48]. CFAST does not contain an algorithm that accounts for the time delay
owing to smoke transport from the fire to the detector location, nor to the penetration of smoke
into the detector chamber itself.

• CFAST can also a smoke detector as a very sensitive heat detector with an activation temper-
ature rise of 5 ◦ C and RTI of 5 (m·s)1/2 . The temperature rise estimate is based on a study
performed by Bukowski and Averill [49]. The RTI is not based on experiment; it simply pro-
vides a slight time delay in activation.

Users are cautioned that both these models are very crude and the uncertainty in predictions are
substantial. Consult the CFAST Validation Guide for more details on its validation.

6.4 Fire Suppression

Fire suppression by water is predicted using a simple empirical model developed by Madrzykowski
[50] and Evans [51]. After activation of the sprinkler, t > tact , the heat release rate is assumed to
decrease exponentially:

Q̇(t) = Q̇(tact ) e−(t−tact )/τ ; −1.8

τ = 3 uw (6.5)

where uw is the water spray density, expressed in units1 of mm/s. The product species mass pro-
duction rates are reduced by the same amount as the heat release rate.
There are assumptions and limitations in this approach. Its main deficiency is that it assumes
that sufficient water is applied to the fire to cause a decrease in the rate of heat release. This
1 Note the CFAST graphical interface might use other units, but the program does convert the user-specified value
into mm/s.

38
suppression model cannot handle the case when the fire overwhelms the sprinkler. The suppression
model as implemented does not include the effect of a second sprinkler. Detection of all sprinklers
are noted but their activation does not make the fire go out any faster. Further, multiple fires in
a room imply multiple ceiling jets. It is not clear how the two ceiling jets should interact. When
there is more than one fire, the detection algorithm uses the fire that results in the highest ceiling
jet temperature in order to calculate the sprinkler link temperature.

6.5 Tenability
The International Organization for Standardization (ISO) standard 13571 [52] provides methods
to estimate the time to untenable conditions that may result from a fire. This includes effects from
asphyxiant gases, radiant heat, and convected heat. CFAST includes tenability estimates as part
of its target calculations so that tenability is determined for user-selected locations within one or
more compartments. All of these are expressed as a fractional effective dose (FED) which is an
integrated area under a time-concentration curve for the chosen quantity. The time at which the
FED exceeds a chosen threshold value represents the time to compromised tenability [52]
For asphyxiant gases, this is a function of the CO and HCN concentration,
t2
ϕCO · νCO2 t2
(ϕHCN · νCO2 )2.36
χFED = ∑ ∆t + ∑ ∆t (6.6)
t1 35000 t1 1.2 × 106

where ϕCO is the average concentration of CO over the time increment ∆t in µL/L, ϕHCN is the
average concentration of HCN over the time increment ∆t in µL/L, and νCO2 = eϕCO2 / 5 is used to
account for an increased rate of asphyxiant gases due to hyperventilation where ϕCO2 is the average
molar concentration of CO2 over the time increment ∆t.
For radiant and convected heat, the FED is given by
t2
1 1
χFED = ∑ −1.9
∆t + ∆t (6.7)
t1 4.2 q̇′′ 4.1 × 108 T −3.61

where q̇′′ ia the radiant heat flux in kW/m2 and T is the temperature in ◦ C. For target locations
where the radiant flux is less than 2.5 kW/m2 , the radiant heat term is taken to be zero consistent
with ISO 13571.

39
40
References

[1] American Society for Testing and Materials, West Conshohocken, Pennsylvania. ASTM E
1355-04, Standard Guide for Evaluating the Predictive Capabilities of Deterministic Fire
Models, 2004. vii, 1, 2
[2] W. W. Jones and R. D. Peacock. Technical Reference Guide for FAST Version 18. Technical
Note 1262, National Institute of Standards and Technology, 1989. 1
[3] L. Y. Cooper and G. P. Forney. The Consolidated Compartment Fire Model (CCFM) Com-
puter Application CCFM-VENTS – Part I: Physical Reference Guide. NISTIR 4342, Na-
tional Institute of Standards and Technology, 1990. 1
[4] R. W. Bukowski, R. D. Peacock, W. W. Jones, and C. L. Forney. Technical Reference Guide
for the HAZARD I Fire Hazard Assessment Method. Version 1.1. Volume 2. NIST Handbook
146/II, National Institute of Standards and Technology, 1991. 1
[5] R. D. Peacock, W. W. Jones, G. P. Forney, R. W. Portier, P. A. Reneke, R. W. Bukowski, and
J. H. Klote. Update Guide for HAZARD I Version 1.2. NISTIR 5410, National Institute of
Standards and Technology, 1994. 1
[6] R. D. Peacock, P. A. Reneke, and G. P. Forney. CFAST – Consolidated Fire And Smoke
Transport (Version 7) Volume 2: User’s Guide. Technical Note 1889v2, National Institute of
Standards and Technology, Gaithersburg, Maryland, November 2015. 2, 4
[7] R. D. Peacock and P. A. Reneke. CFAST – Consolidated Fire And Smoke Transport (Version
7) Volume 3: Software Development and Model Evaluation Guide. Technical Note 1889v3,
National Institute of Standards and Technology, Gaithersburg, Maryland, November 2015. 2,
4, 5
[8] R.D. Peacock. CFAST – Consolidated Fire And Smoke Transport (Version 7) Volume 4:
Configuration Management Guide. Technical Note 1889v4, National Institute of Standards
and Technology, Gaithersburg, Maryland, November 2015. 2
[9] M.H. Salley and A. Lindeman. Verification and Validation of Selected Fire Models for Nu-
clear Power Plant Applications, Supplement 1. NUREG 1824, U.S. Nuclear Regulatory
Commission, Rockville, Maryland, 2015. 2
[10] K.B. McGrattan, S. Hostikka, J.E. Floyd, H.R. Baum, and R.G. Rehm. Fire Dynamics Simu-
lator, Technical Reference Guide, Volume 1: Mathematical Model. NIST Special Publication
1018-1, National Institute of Standards and Technology, Gaithersburg, Maryland, 2015. 2

41
[11] W. W. Jones. Modeling Smoke Movement Through Compartmented Structures. Journal of
Fire Sciences, 11(2):172, 1993. 3

[12] W. W. Jones. Multicompartment Model for the Spread of Fire, Smoke and Toxic Gases. Fire
Safety Journal, 9(1):172, 1985. 3

[13] W. W. Jones. Prediction of Corridor Smoke Filling by Zone Models. Combustion Science
and Technology, 35:229, 1984. 3

[14] W. W. Jones and G. P. Forney. Modeling Smoke Movement Through Compartmented Struc-
tures. In Proceedings of the Fall Technical Meeting of the Combustion Institute, Eastern
States Section, Ithaca, NY, 1991. 3

[15] W.D. Walton, D.J. Carpenter, and C.B. Wood. SFPE Handbook of Fire Protection Engineer-
ing, chapter Zone Computer Fire Models for Enclosures. National Fire Protection Associa-
tion, Quincy, Massachusetts, fourth edition, 2008. 3

[16] National Fire Protection Association, Quincy, Massachusetts. NFPA 805, Performance-Based
Standard for Fire Protection for Light Water Reactor Electric Generating Plants, 2001 edi-
tion, 2004. 3

[17] National Fire Protection Association, Quincy, Massachusetts. NFPA 551, Guide for the Eval-
uation of Fire Risk Assessment, 2004 edition, 2004. 3

[18] G. P. Forney and W. F. Moss. Analyzing and Exploiting Numerical Characteristics of Zone
Fire Models. Fire Science and Technology, 14(1 and 2):49–60, 1994. 7, 10

[19] L. R. Petzold. A Description of DASSL: A Differential/Algebraic Sytem Solver. SAND82-

8637, Sandia National Laboratory, Albuqueque, New Mexico, September 1982. 9, 10, 53

[20] K. E. Brenan, S. L. Campbell, and L. R. Petzold. Numerical Solution of Initial-Value Prob-

lems in Differential-Algebraic Equations. Elsevier Science Publishing, New York, 1989. 9,
10, 53

[21] R. G. Rehm and G. P. Forney. A Note on the Pressure Equations Used in Zone Fire Modeling.
NISTIR 4906, National Institute of Standards and Technology, 1992. 10

[22] W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B. P. Flannery. Numerical Recipes: The Art
of Scientific Computing. Cambridge University Press, New York, 3rd edition, 2007. 10

[23] W. W. Jones and R. D. Peacock. Using CFAST to Estimate the Efficiency of Filtering Par-
ticulates in a Building. NISTIR 7498, National Institute of Standards and Technology, 2008.
11, 14

[24] A. Hamins, K. Konishi, P. Borthwick, and T. Kashiwagi. Global Properties of Gaseous Pool
Fires. In Twenty-Sixth Symposium (International) on Combustion, pages 1429 – 1436. The
Combustion Institute, 1996. 14

42
[25] A. Tewarson. Generation of Heat and Chemical Compounds in Fires, chapter 4-3. National
Fire Protection Association and The Society of Fire Protection Engineers, Quincy, MA, 3rd
edition, 2003. 14
[26] B. J. McCaffrey. Entrainment and Heat Flux of Buoyant Diffusion Flames. NBSIR 82-2473,
National Bureau of Standards, 1982. 14
[27] H. Koseki. Combustion Properties of Large Liquid Pool Fires. Fire Technology, 25:241,
1989. 14
[28] A. Hamins, M. Klassen, J. Gore, and T. Kashiwagi. Estimate of Flame Radiance via a Single
Location Measurement in Liquid Pool Fires. Combustion and Flame, 86:223–228, 1991. 14
[29] A. Hamins, M. Klassen, J. Gore, and T. Kashiwagi. Estimate of the Effect of Scale on Ra-
diative Heat Loss Fraction and Combustion Efficiency. Combustion Science and Technology,
96:183–188, 1994. 14
[30] C. Huggett. Estimation of the Rate of Heat Release by Means of Oxygen Consumption
Calorimetry. Journal of Fire and Flammability, 12:61, 1980. 15
[31] G. Heskestad. SFPE Handbook of Fire Protection Engineering, chapter Fire Plumes, Flame
Height, and Air Entrainment. National Fire Protection Association, Quincy, Massachusetts,
fourth edition, 2008. 15, 16, 17
[32] R.L. Alpert. SFPE Handbook of Fire Protection Engineering, chapter Ceiling Jet Flows.
National Fire Protection Association, Quincy, Massachusetts, fourth edition, 2008. 18, 33
[33] H.W. Emmons and T. Tanaka. SFPE Handbook of Fire Protection Engineering, chapter Vent
Flows. National Fire Protection Association, Quincy, Massachusetts, fourth edition, 2008.
21, 22
[34] K. D Steckler, H.R. Baum, and J. G. Quintiere. Fire Induced Flows Through Room Openings
– Flow Coefficients. NBSIR 83-2801, National Bureau of Standards, Gaithersburg, Mary-
land, March 1984. 21
[35] M. Poreh, H.P. Morgan, N.R. Marshall, and R. Harrison. Entrainment by Two-Dimensional
Spill Plumes. Fire Safety Journal, 30:1–19, 1998. 22
[36] L. Y. Cooper. Calculation of the Flow Through a Horizontal Ceiling/Floor Vent. NISTIR
89-4052, National Institute of Standards and Technology, 1989. 23
[37] L.Y. Cooper. SFPE Handbook of Fire Protection Engineering, chapter Smoke and Heat
Venting. National Fire Protection Association, Quincy, Massachusetts, third edition, 2003.
24
[38] G. P. Forney. Computing Radiative Heat Transfer Occurring in a Zone Fire Model. NISTIR
4709, National Institute of Standards and Technology, 1991. 27
[39] R. Siegel and J. R. Howell. Thermal Radiation Heat Transfer. Hemisphere Publishing Cor-
poration, New York, 2nd edition, 1981. 28, 30, 32

43
[40] Van Oosterom A. and Strackee J. The solid angle of a plane triangle. IEEE Trans Biomed
Eng, 30:125–6–188, 1983. 31

[41] C. L. Tien, K. Y. Lee, and A. J. Stretton. SFPE Handbook of Fire Protection Engineering,
chapter Radiation Heat Transfer. National Fire Protection Association, fourth edition, 2008.
32

[42] C. L. Tien and G. Hubbard. Infrared Mean Absorption Coefficients of Luminous Flames and
Smoke. Journal of Heat Transfer, 100:235, 1978. 32

[43] D. K. Edwards. Radiation Properties of Gases. In W. M. Rohsenow, editor, Handbook of

Heat Transfer Fundementals, pages 74–75. McGraw-Hill Book Company, 2nd edition, 1985.
32

[44] J. P. Holman. Heat Transfer. McGraw-Hill Book Company, 1990. 33

[45] W. F. Moss and G. P. Forney. Implicitly coupling heat conduction into a zone fire model.
NISTIR 4886, National Institute of Standards and Technology, 1992. 34

[46] R.L.P. Custer, B.J. Meacham, and R.P. Schifiliti. SFPE Handbook of Fire Protection Engi-
neering, chapter Design of Detection Systems. National Fire Protection Association, Quincy,
Massachusetts, fourth edition, 2008. 37, 38

[47] G.W. Mulholland. SFPE Handbook of Fire Protection Engineering, chapter Smoke Pro-
duction and Properties. National Fire Protection Association, Quincy, Massachusetts, fourth
edition, 2008. 38

[48] J. A. Milke, F W. Mowrer, and P. Gandhi. Validation of a Smoke Detection Performance

Methodology, Volume 3. Evaluation of Smoke Detector Performance. Technical report, The
Fire Protection Research Foundation, Quincy, Massachusetts, October 2008. 38

[49] R.W. Bukowski and J.D. Averill. Methods for Predicting Smoke Detector Activation. In Pro-
ceedings of the Fire Suppression and Detection Research Application Symposium, Research
and Practice, Bridging the Gap, pages 64–72, Orlando, Florida, February 1998. National
Fire Protection Research Foundation, National Fire Protection Association, Quincy, Mas-
sachusetts. 38

[50] D. Madrzykowski and R. Vettori. A Sprinkler Fire Suppression Algorithm for the GSA
Engineering Fire Assessment System. NISTIR 4833, National Institute of Standards and
Technology, 1992. 38

[51] D. D. Evans. Sprinkler Fire Suppression for HAZARD. NISTIR 5254, National Institute of
Standards and Technology, 1993. 38

[52] International Organization for Stanardization, CP 401 - 1214 Vernier, Geneva, Switzerland.
Life-threatening components of fire - Guidelines for the estimation of time to comprimised
tenability in fires, iso 13571:2012(e) edition, September 2012. 39

44
Appendix A

Nomenclature

Note that the units associated with a given symbol are sometimes changed upon input to and output
from the program. In particular, temperatures are typically input in degrees Celsius, converted to
Kelvin, and then converted back to Celsius on output. Energy units involving Joules or Watts are
typically input as kJ or kW, converted to J or W, then converted back to kJ or kW.

A area, m2
Av cross-sectional area of a vent, m2
a absorption coefficient, m−1
C vent constriction (or flow) coefficient, dimensionless
CLOL lower oxygen limit coefficient, dimensionless
cp heat capacity of air at constant pressure, J/(kg·K) or kJ/(kg·K)
cv heat capacity of air at constant volume, J/(kg·K) or kJ/(kg·K)
c heat capacity of a solid, kJ/(kg·K)
D fire diameter, m
Du optical density per unit distance, m−1
D∗ characteristic fire diameter, m
Fk− j configuration factor, surface k intercepted by surface j, dimensionless
fv soot volume fraction, dimensionless
g gravitational constant, 9.8 m/s2
h convective heat transfer coefficient, W/(m2 ·K)
H height of a compartment, m
k thermal conductivity of air, W/(m·K)

45
L length of a compartment, m
mean flame height, m
characteristic length for radiation calculation, m
Mα molar mass of gas species α, g/mol
mi total mass in gas layer i, kg
ṁi rate of mass addition, gas layer i, kg
ṁe plume mass entrainment rate, kg/s
ṁf pyrolysis rate, kg/s
nα number of atoms of element α
Ou Obscuration per unit distance, %/m
P pressure at floor level of a compartment, Pa
q̇i rate of addition of heat into layer i, kW
Q̇ total heat release rate of the fire, kW
Q̇c heat release rate of the fire released as convective energy, kW
q̇′′ heat flux, kW/m2
R specific gas constant, J/(kg·K)
RTI response time index of a sprinkler or heat detector, (m·s)1/2
r radial distance from the fire, m
radial coordinate of cylindrical solid, m
S vent coefficient for vertical flow vents, dimensionless
T∞ ambient gas temperature, K
Ti gas temperature of layer i, K
Tp gas temperature in the plume, K
t time, s
uw water spray density, m/s
V total volume of a compartment, m3
V̇ volumetric flowrate, m3 /s
Vl total volume of lower layer in a compartment, m3
Vu total volume of upper layer in a compartment, m3
v gas velocity, m/s

46
W width of a compartment, m; wall thickness, m
w door or window width, m
x length variable, m
YLOL lower oxygen limit, by mass, dimensionless
YO2 mass fraction of oxygen in a gas layer, dimensionless
yCO carbon monoxide yield, kg/kg
ys soot yield, kg/kg
z height variable, m
z̄ mid-range height of a vertical vent segment, m
z0 height of the virtual origin of fire, m

α gas absorptance, dimensionless

thermal diffusivity in conduction, (m2 /s)
γ ratio of specific heats, 1.4, dimensionless
∆H heat of combustion of the fuel, kJ/kg
∆HO2 energy released per unit mass of oxygen consumed, kJ/kg
∆T temperature rise, K
ε emissivity
ν stoichiometric coefficient, dimensionless
kinematic viscosity, m2 /s
ρ density, kg/m3
ρ∞ ambient density of air, 1.2 kg/m3
σ Stefan-Boltzman constant (5.67 × 10−8 W/(m2 ·K4 )
τ transmittance, dimensionless
χr radiative fraction, dimensionless

47
48
Appendix B

Modeling Equation Derivations

For convenience, Eq. (2.3) (conservation of energy) and Eq. (2.4) (ideal gas law) from Section 2.1
used in the following derivations are re-written here:

d dVi
(cv mi Ti ) = q̇i − P (B.1)
dt dt

PVi = mi R Ti (B.2)

B.1 Pressure Equation

To derive a differential equation for compartment pressure, rewrite Eq. (B.1) for the upper and
lower layer to obtain

d dVl
cv (ml Tl ) = q̇l − P (B.3)
dt dt
d dVu
cv (mu Tu ) = q̇u − P (B.4)
dt dt
From Eq. (B.2)
ml Tl + mu Tu = PVl /R + PVu /R = PV /R (B.5)
Therefore summing Eqs. (B.3) and (B.4) gives

cv V dP
= q̇l + q̇u (B.6)
R dt

since dVu
dt = − dV
dt . Finally R/cv = γ − 1 since R = c p − cv and γ = c p /cv giving
l

dP γ −1
= (q̇l + q̇u ) (B.7)
dt V
the pressure differential equation.

49
B.2 Upper Layer Volume Equation
To derive the upper layer volume differential equation, first note from Eq. (B.2) that
P cv 1
cv mu Tu = cvVu = Vu P = Vu P (B.8)
R c p − cv γ −1
Substituting Eq. (B.8) into the upper layer form of Eq. (B.1) (i = u) gives
1 d dVu
q̇u = (Vu P) + P (B.9)
γ − 1 dt dt
 
1 dVu dP dVu
= P +Vu +P (B.10)
γ −1 dt dt dt
γ dVu Vu dP
= P + (B.11)
γ − 1 dt γ − 1 dt
Multiplying both sides by (γ − 1)/(γP) gives
γ −1 dVu Vu dP
q̇u = + (B.12)
γP dt γP dt
Finally, rearranging terms yields
 
dVu 1 dP
= (γ − 1)q̇u −Vu (B.13)
dt Pγ dt
the upper layer volume differential equation.

B.3 Temperature Equations

To derive the upper layer temperature differential equation note that
d dTu
(cv mu Tu ) = cv ṁu Tu + cv mu (B.14)
dt dt
and from Eq. (B.13) that
dVu γ − 1 Vu dP
P = q̇u − (B.15)
dt γ γ dt
Substituting these two expressions into the upper layer form of Eq. (B.1) (again i = u) gives
d dVu
(cv mu Tu ) + P = q̇u (B.16)
dt dt
dTu γ − 1 Vu dP
cv ṁu Tu + cv mu + q̇u − = q̇u (B.17)
dt γ γ dt
Multiplying Eq. (B.17) by γ gives
dTu dP
cp ṁu Tu + cp mu −Vu = q̇u (B.18)
dt dt

50
 dTu
Finally, solving for dt gives
 
dTu 1 dP
= q̇u − cp ṁu Tu +Vu (B.19)
dt cp mu dt

the upper layer temperature differential equation. A differential equation for lower layer tempera-
ture  
dTl 1 dP
= q̇l − cp ṁl Tl +Vl (B.20)
dt cp ml dt
may be derived similarly.

51
52
Appendix C

Solving the Modeling Equations

As stated in Section 2.2, CFAST uses the differential / algebraic solver DASSL [19, 20] to solve
the zone fire modeling equations Eqs. (2.5) through Eq. (2.9). DASSL solves these equations by
formulating a root finding problem
F t, y, y′ = 0


(C.1)
where F is given by

 y′ − f (t, y) gas ode′ s
F(t, y, y′ ) = (C.2)
 q̇′′ + k ∂ Tw (0,t) gas/solid constraint
∂x

and where f is the right hand side of Eqs. (2.5) through (2.8), t is the time, y is the solution vector
containing pressures, upper layer volumes, layer temperatures and wall surface temperatures and
y′ is the solution vector derivative of these quantities at time t. The solution vector, y is defined as
follows

y = (P1 , P2 , ..., Pn−1 , Pn , pressures for each compartment

Tu1 , Tu2 , ..., Tun−1 , Tun upper layer temperatures for each compartment
Vu1 ,Vu2 , ...,Vun−1 ,Vun upper layer volumes for each compartment
Tl1 , Tl2 , ..., Tln−1 , Tln lower layer temperatures for each compartment
Tw1 , Tw2 , ..., Twk−1 , Twk ) temperatures on each surface of each compartment

where n is the number of compartments and k is the total number of wall surfaces (4n if all wall
surfaces are solved for each compartment). The components of y′ and the residual function F are
arranged analogously. The first n components of F correspond to the dP/dt differential equations,
the next n to the dTu /dt differential equations and so on. Each component of F is a residual which
DASSL drives to zero. For example if dP1 /dt, the differential equation for pressure in the first
compartment, is computed using Eq. (2.5), then F1 (t, y, y′ ), the first component of F, is given
by the difference between y′1 , the estimate of dP1 /dt provided by DASSL, and the dP1 /dt term
computed from Eq. (2.5) or equivalently

F1 (t, y, y′ ) = y′1 − dP1 /dt (C.3)

53
DASSL then determines values for y and y′ so that Eq. (C.3) along with all other components of F
are sufficiently close to zero as determined by a user defined error criteria.
In more detail, F for all components is given by



 y′i − dP
dt
i
1≤i≤n

dT

y′ − dti,u n + 1 ≤ i ≤ 2n


 i


Fi (t, y, y′ ) = dV
y′i − dti,u 2n + 1 ≤ i ≤ 3n (C.4)

dT

y′i − dti,l 3n + 1 ≤ i ≤ 4n






 q′′ (y) + k ∂ Tw (0,t) 4n + 1 ≤ i ≤ 4n + k

∂x

The last equation, q′′ (y) + k ∂∂Txw (0,t), is used to couple the gas with the solid phase. It does this
by requiring that the convective and radiative flux , q′′ (y), striking the wall surface be consistent
with the temperature gradient, ∂∂Txw (0,t), at that surface according to Fourier’s law:

∂T
q′′ = −k (C.5)
∂x
It is implemented as an algebraic equation by discretizing the spatial derivative term ∂∂Txw and
is evaluated by advancing the wall temperature profile from a previously known solution at told to
t using the heat conduction equation

∂ Tw k ∂ 2 Tw
= (C.6)
∂t c ρ ∂ x2
and the boundary condition Tw (0,t) = Tw at the surface provided by DASSL. An outline of the
solution procedure is as follows:

1. Assume that yn and y′n are known at time t = tn

2. For some ∆t guess values for yn+1 and y′n+1 at tn+1 = t + ∆t

3. Using the guessed solution yn+1 , compute mass flow rates: ṁl and ṁu and heat flow rates: q̇l
and q̇u

4. Using the guessed solution yn+1 and the mass and heat flow rates computed in step 3, compute
the right hands side of Eqs. (2.5) through (2.8). Denote these computations ŷ′

5. Compute the gas equation residuals, y′n+1 − ŷ′n+1 , using y′n+1 given in step 2 and ŷ′n+1 computed
in step 4.

6. Compute the temperature gradient residual (last row of Eq. (C.4)):

(a) Using the surface wall temperature found in yn+1 as a boundary condition, advance the
temperature profile Tw from t = tn to t = tn+1 using Eq. (C.6)
(b) Compute the temperature gradient, ∂ Tw /∂ x(0,tn+1 ), at the wall surface using the tem-
perature profile computed in step 6a

54
 (c) Compute the residual q′′ + k ∂∂Txw using heat flow rates computed in step 3 and temperature
gradients computed in step 6b

7. Are the residuals computed in steps 5 and 6 sufficiently small according to a specified error
tolerance? If yes, advance n and go to step 1 to compute a solution at the next time step. If no,
go to step 2 and iterate again.

55