SECTION 3

Chapter 7

Fire Hazard
Analysis Techniques
Chapter Contents
Performing a Fire
Morgan J. Hurley Richard W. Bukowski Hazard Analysis
Developing Fire
Scenarios and Design
Fire Scenarios
Quantification of Design
A vailable methods to estimate the potential impact of fire can be divided into two categories:
risk-based and hazard-based. Both types of methods estimate the potential consequences of
possible events. Risk-based methods also analyze the likelihood of scenarios occurring, whereas
Fire Scenarios
Prediction of Hazards

hazard-based methods do not. Fire risk analysis is described more fully in Section 3, Chapter 8,
Key Terms
“Fire Risk Analysis.” Section 3, Chapter 9, “Closed Form Enclosure Fire Calculations,” provides
simple fire growth calculation methods. bounding condition, design
The goal of a fire hazards analysis (FHA) is to determine the expected outcome of a specific fire curve, design fire
set of conditions called a fire scenario. The scenario includes details of the room dimensions, con- scenario, fire hazard
tents, and materials of construction; arrangement of rooms in the building; sources of combustion analysis, fire model, fire
air; position of doors; numbers, locations, and characteristics of occupants; and any other details scenario, performance-
based design, t-squared fire
that have an effect on the outcome of interest. This outcome determination can be made by expert
judgment, by probabilistic methods using data from past incidents, or by deterministic means such
as fire models. “Fire models” include empirical correlations, computer programs, full-scale and
reduced-scale models, and other physical models. The trend today is to use models whenever pos-
sible, supplemented if necessary by expert judgment. Although probabilistic methods are widely
used in risk analysis, they find little direct application in modern hazard analyses. Probabilistic
models are discussed in Section 3, Chapter 5, “Introduction to Fire Modeling.” Typically, when the
potential impact of fire is estimated, a hazard basis is used. When probabilities or frequencies are
considered, it is usually in the context of determining whether or not a scenario is sufficiently likely
to warrant further analysis.
Hazard analysis can be used for one of two purposes. One is to determine the hazards that are
present in an existing or planned facility. The other use is for design, where trial design strategies
are evaluated to determine whether they achieve a set of fire safety goals. Hazard analysis can be
thought of as a component of risk analysis. That is, a risk analysis is a set of hazard analyses that
have been weighted by their likelihood of occurrence. The total risk is then the sum of all of the
weighted hazard values. In the insurance and industrial sectors, risk assessments generally target
monetary losses, since these dictate insurance rates or provide the incentive for expenditures on
protection. In the nuclear power industry, probabilistic risk assessment has been the basis for safety
regulation. Here the risk of a release of radioactive material to the environment is commonly exam-
ined, ranging from a leak of contaminated water to a core meltdown.
Available fire hazard calculation methods range from relatively simple equations that can be
performed with a hand calculator to complex methods that require powerful computers, and many
methods that fall between.
See also Section 2, Chapter 1, “Physics and Chemistry of Fire”; Section 3, Chapter 4, “Use of
Fire Incident Data and Statistics”; Section 3, Chapter 5, “Introduction to Fire Modeling”; Section 3,
Chapter 8, “Fire Risk Analysis”; Section 3, Chapter 9, “Closed Form Enclosure Fire Calculations”;

Morgan J. Hurley, P.E., is the technical director of the Society of Fire Protection Engineers, where he is
responsible for advancing the science and practice of fire protection engineering internationally. He is also
adjunct faculty in the University of Maryland’s Department of Fire Protection Engineering.
Richard W. Bukowski, P.E., is a senior research engineer and coordinator of codes and standards activities for
the Building and Fire Research Laboratories at NIST.

3-121
3-122 SECTION 3 ■ Information and Analysis for Fire Protection

Section 3, Chapter 10, “Performance-Based Codes and Stan- DEVELOPING FIRE SCENARIOS
dards for Fire Safety”; and Section 3, Chapter 11, “Overview of
Performance-Based Fire Protection Design.”
AND DESIGN FIRE SCENARIOS
Fire Scenario and Design Fire Scenario Defined
Determining the fire source is one of the most important parts of
PERFORMING A FIRE HAZARD ANALYSIS
performing a fire hazard analysis. To determine the fire source,
Steps of a Fire Hazard Analysis a design fire scenario must be developed. A fire scenario is a set
of conditions that defines the development of fire and the spread
Performing an FHA is a fairly straightforward engineering anal- of combustion products. Fire scenarios comprise three sets of
ysis. The steps include the following: features: building characteristics, occupant characteristics, and
1. Selecting a target outcome fire characteristics. Building characteristics describe the build-
2. Determining the scenario(s) of concern that could result in ing features that could affect fire development and the spread
that outcome of combustion products. Occupant characteristics describe the
3. Selecting an appropriate method(s) for prediction of growth state(s) of occupants at the time of the fire. Fire characteristics
rate of fire effects describe the ignition and growth of the fire. A design fire sce-
4. Calculating the time needed for occupants to move to a safe nario is a set of conditions that defines the critical factors for
place determining the outcomes for trial fire protection designs of new
5. Analyzing the impact of exposure of occupants or property buildings or modifications to existing buildings.2 Design fire
to the effects of the fire scenarios are the fire scenarios that are selected to analyze a trial
6. Examining the uncertainty in the hazard analysis design. They are generally a subset of the fire scenarios.
7. Documentation of the fire hazard analysis process, includ- The design fire scenario is based on a fire that has a rea-
ing the basis for selection of models and input data sonable likelihood of developing from a series of events. Fire
scenarios need to be based on reality and should be developed
Fire hazard analysis can also be used as part of the accordingly. For example, the occupancy, the purpose for which
performance-based design process. For more information the design is being developed, the fuel load, potential changes
on performance-based design, see Section 3, Chapter 10, in the property, the presence of sprinklers and fire detection, the
“Performance-Based Codes and Standards for Fire Safety,” and presence of alarm and notification systems, and smoke manage-
Section 3, Chapter 11, “Overview of Performance-Based Fire ment should be considered. Design fire scenarios differ by oc-
Protection Design.” cupancy and should be based on reasonably expected fires and
worst-case fires. Although this chapter deals with hazard-based
Selecting a Target Outcome approaches, some risk must be included in the analysis when
developing design fire scenarios. For instance, if a fire may be
The target outcome most often specified is avoidance of occu- technically plausible but is extremely unlikely, that scenario
pant fatalities in a building. Another might be to ensure that fire may not be necessary to include in the design fire scenarios.
fighters are provided with protected areas from which to fight
fires in high-rise buildings. The U.S. Department of Energy Determining the Scenario(s) of Concern
(DOE) requires that FHAs be performed for all DOE facilities.1
Their objectives for such FHAs, as stated in DOE 5480.7A, in- Records of past fires, either for the specific building or for simi-
clude the following: lar buildings or class of occupancy, can be of substantial help in
identifying conditions to be avoided. Statistical data from NFPA
• Minimizing the potential for the occurrence of fire
or from the National Fire Incident Reporting System (NFIRS)
• No release of radiological or other hazardous material to
on ignition sources, first items ignited, rooms of origin, and the
threaten health, safety, or the environment
like can provide valuable insight into the important factors con-
• An acceptable degree of life safety to be provided for DOE
tributing to fires in the occupancy of interest. (See also Section
and contractor personnel and no undue hazards to the pub-
3, Chapter 4, “Use of Fire Incident Data and Statistics.”) Anec-
lic from fire
dotal accounts of individual incidents are interesting but might
• Critical process control or safety systems are not damaged
not represent the major part of the problem to be analyzed.
by fire
Murphy’s Law (“if anything can go wrong, it will”) applies
• Vital programs are not delayed by fire (mission continuity)
to major fire disasters; that is, significant fires seem to involve a
• Property damage does not exceed acceptable levels (e.g.,
series of failures that set the stage for the event. Therefore, it is
$150 million per incident)
important to examine the consequences of things not going ac-
An insurance company might want to limit the maximum cording to plan. In DOE-required FHAs, one part of the analy-
probable loss to that on which the insurance rate paid by the sis is to assume both that automatic systems fail and that the
customer is based; a manufacturer might want to avoid failures fire department does not respond. This is used to determine a
to meet orders to avoid erosion of its customer base; and some worst-case loss and to establish the real value of these systems.
businesses might want to guard their public image of provid- The 2006 edition of NFPA 101®, Life Safety Code®, includes a
ing safe and comfortable accommodations. Any combination of performance-based design option containing a basic set of de-
these outcomes could be selected as appropriate for an FHA. sign fire scenarios. Scenario 8 is a common fire that starts while
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-123

either the fire alarm system or the sprinkler system (in turn) is 2. An ultrafast fire in a primary means of egress would likely
rendered ineffective. Given the normal high reliability of these mean a flammable liquid fire in the corridor near one of the
systems, it is not required for the performance objectives to be exit doors.
met fully under these conditions, but stakeholders should feel 3. Fire in a normally unoccupied room would likely include
that the resulting losses are not catastrophic or otherwise un- a fire in a janitor’s closet, started by oily rags or ignition of
acceptably severe. In a risk assessment, the consequences of some cleaning fluid.
such failures would be weighted by the probability of failure and 4. Fire in a concealed space, particularly if the hotel were of
added into the total risk. In a hazard analysis, the objective is combustible construction, might occur in the drop ceiling
hazard avoidance, so the contribution of low probability events above the bathroom. This would likely be an electrical
is more subjective. Scenarios must be translated into design fires fire.
for fire growth analysis and occupant evacuation calculation. 5. A shielded fire near occupied space might be in a maid’s
See the discussion in the Quantification of Design Fire Scenar- cart or under a display table in a meeting room.
ios section later in this chapter. 6. The most severe fire from the largest fuel load typical to the
building might occur during remodeling or might occur due
to storage of furniture in one room or storage of chairs in a
NFPA 101 Design Fire Scenarios
meeting room.
NFPA 101 provides eight design fire scenarios that should be 7. The outside exposure fire could include other buildings,
considered in the development of a performance-based design. skylights in the roof of a low-rise building nearby, or a
Briefly, these design fire scenarios are as follows: wildland fire. This fire would be specific to the occupancy
and building being considered.
1. An occupancy-specific design fire scenario that is represen-
8. Failure of a system would need to include looking at rated
tative of a typical fire for the occupancy
walls, rated floors, as well as sprinkler and fire alarm sys-
2. An ultrafast-developing fire in the primary means of egress,
tems. When looking at these systems, one should consider
with interior doors open at the start of the fire (for a discus-
what might fail rather than failure of the entire system. For
sion of fire development, see Section 2, Chapter 4, “Dy-
instance, failure of a sprinkler system might mean failure
namics of Compartment Fire Growth”)
of the entire water supply or it might mean failure of a
3. A fire that starts in a normally unoccupied room that may
single sprinkler to react when expected. By providing re-
endanger large numbers of occupants
dundancy into water supply and fire pumps, and monitor-
4. A fire that originates in a concealed wall or ceiling space
ing main valves, failures could be limited as a part of this
adjacent to a large occupied room
evaluation.
5. A slowly developing fire, shielded from fire protection sys-
tems, in close proximity to a high-occupancy area
Bounding Conditions
6. The most severe fire resulting from the largest possible fuel
load characteristic of the normal operation of the building During development of the fire scenarios and design fire sce-
7. An outside exposure fire narios, the allowable future changes in the facility must also
8. A fire originating in ordinary combustibles with each pas- be considered. The extent of the changes that are considered
sive or active fire protection system individually rendered by the design become bounding conditions for the analysis and
ineffective; this scenario is not required where it can be subsequent use of the building. One can expect that a design fire
shown that the level of reliability and the design perfor- scenario is not exactly what will happen and that the building
mance in the absence of the system are acceptable to the as originally designed and anticipated will not remain exactly
authority having jurisdiction (AHJ) as analyzed. Therefore, as one develops design fire scenarios
and one calculates the expected fire response, some amount of
Although only eight scenarios are listed in the performance
change in those scenarios must also be considered.
option of NFPA 101, more than eight scenarios will be devel-
When conducting a hazard analysis, it is important to con-
oped and analyzed. For most building designs, for example,
sider the types of changes that may occur. If the hazard analysis
there will usually be far more than a single scenario that is rep-
only considered a specific set of initial conditions, then it would
resentative of a typical fire in a given occupancy.
be necessary to revise the fire hazard analysis any time changes
were made in the future. The range of changes that will be con-
Applying NFPA 101 Design Fire Scenarios sidered by the hazard analysis is a judgment call between the
designers and the owner.
For a typical building, what happens when each of these eight
For example, a hotel room floor might become a meeting
general scenarios is applied to what might occur as a reasonable
room floor; a meeting room area might become an exposition
design fire in that building? For the purposes of this illustration,
center; occupant loads could be greater than expected or calcu-
a multistory hotel building with some meeting rooms on lower
lated; movable walls could create simultaneous use when nonsi-
floors is considered. The following fires might be used as design
multaneous use was expected; or the space between a ballroom
fires in meeting the eight-scenario criteria of NFPA 101:
ceiling and the floor above might be used for storage. All of
1. A typical fire based on the occupancy might include a pa- these events are reasonably foreseeable, but some may fall out-
tron smoking in bed, or a sterno-initiated fire in a meeting side of the bounding conditions. Bounding conditions must be
room or restaurant area. clearly identified because changes in the building may occur.
3-124 SECTION 3 ■ Information and Analysis for Fire Protection

Other situations that might occur on a more general basis, The first step is to investigate potential fires that might
for any occupancy, include the response of a fire department and occur so that the design fire scenarios can be chosen. Based
cutbacks in fire department funding or unwanted alarms causing on the scenarios from NFPA 101, the scenarios examined for a
deactivation of a system. Some of these bounding assumptions typical convention center might be as follows:
can be addressed specifically—for instance, maximum fuel load
1. The occupancy-specific design fire scenario might in-
or occupant characteristics.
clude a fire in an exhibit booth or a fire in auditorium seating.
Implied Risk 2. The ultrafast fire might involve a fire in a plastic boat
display located near the main exit.
Although this chapter addresses fire hazard analysis, there is 3. A fire in a normally unoccupied room could occur in the
some implied risk in any such analysis. The primary risk fac- storage of stacked chairs in an exhibit hall next to the exhibit
tors involved are included in the design fire development. The hall being considered. There could be a show in one exhibit hall
design fires described for the hotel building did not include such with large numbers of people, and the adjacent hall might be
accidents as gasoline tanker trucks crashing into the side of the used as temporary storage during that event.
building or bombs ignited at the base of the building. There is
always the risk that these events could happen, but the engineer 4. A fire in a concealed space is unlikely in Type 1 con-
must evaluate the likelihood of these events. For example, build- struction but could occur in electrical or insulation areas.
ings are typically not designed to survive the impact and ensuing 5. A shielded fire could occur in the plastic boat previously
fire of a missile strike. If this were to occur, achievement of the mentioned or in a covered exhibit space. More and more juris-
design goals and objectives might not be expected. Similarly, dictions are requiring automatic sprinklers in covered exhibit
it is conceivable that simultaneous fires could occur, although spaces, but that is not yet universal.
prescriptive building codes such as NFPA 101 explicitly exclude 6. The most severe fire to be considered would likely be
such an event. These might be limitations described in the fire the boat fire previously mentioned.
strategy report to clarify what is covered and what is not.
7. An outside exposure fire would typically not be consid-
When proposing to exclude a scenario from further consid-
ered for this occupancy because convention centers are gener-
eration, it is important to ensure that stakeholders understand
ally surrounded by parking lots and other open areas. However,
the implications of excluding the scenario. For example, if the
if the loading dock is considered outside, the scenario might in-
fire scenario associated with a gasoline tanker truck crashing
volve fire in a truck waiting to unload at the convention center.
into the side of the building is dismissed, and the building is
located on a highway leading to a major oil refinery, stakehold- 8. A typical fire with failure might include failure of the
ers would need to understand and accept that if a gasoline tanker sprinkler or fire alarm system or perhaps failure of the smoke
truck did crash into the side of the building, goals and objectives control system.
might not be met. If the purpose of the example is to perform an egress analysis,
the worst-case fire may be all that is necessary for evaluation.
Data Sources The worst-case fire would likely be the shielded boat fire at peak
In developing design fire scenarios, it is useful to have data on rate of heat release. To quantify the fire, users might look at the
which to base future quantification. Members of the NFPA Life fuel load and estimate the rate of burning, they might look at
Safety Code Technical Committees developed the design fire plastic fires and extrapolate, they might look at fast or ultra-
scenarios based on statistical analyses prepared by the NFPA fast fires and assume the fire peaks at the estimated sprinkler
Fire Analysis and Research Division and also on past fires that response time, they might assume the fire is shielded on the
have occurred in different occupancy types. inside of the boat and so not have the fire peak at the estimated
The NFPA One Stop Data Shop provides much information sprinkler response time, or they might specify sprinklers inside
regarding fire statistics and results. Other sources addressing the boat and limit the fire size. The user would likely try a com-
typical fires in occupancies include Factory Mutual data, state bination of these factors to see the effects.
or local jurisdiction data for various occupancies, the National Once the fire scenario is developed, smoke-filling calcula-
Fire Incident Reporting System, or past fire history published tions can be performed to determine the clear height of a smoke
in the NFPA Journal. Other possibilities include fire test results layer over time. Those calculations would be compared to the
(many of which can be found on the National Institute of Stan- timed evacuation analysis. Both calculations would likely start
dards and Technology Fire Internet site), manufacturers’ data without suppression or smoke control to see whether the evacu-
regarding specific fire performance of materials, or listings of ation can occur without those two systems. If so, the analysis is
materials by recognized test labs. It can be reasonably expected simplified.
that the amount of data to develop a design fire will not be suf- Finally, the user would identify bounding conditions via a
ficient to exactly predict what will happen in all cases. sensitivity analysis. For instance, is the size of the boat important?
How about the materials of the boat? Has the fuel been removed
from the boat? If smoke control is necessary to make the design
Overall Example
work, that smoke control needs to be identified as a critical system.
The following example develops scenarios for a large exhibit Similarly, the occupant load, the exit sizes, the number of disabled
hall at a convention center and describes some of the work that persons, and the availability of an alarm system as well as its audi-
might be done using the scenarios that have been developed. bility must all be considered in the sensitivity analyses.
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-125

Once all of these factors have been considered and dealt ing an evacuation analysis might require quantification of the
with, the hazard-based analysis is complete. The documentation growth and fully developed stages.
of the analysis is the next important part and cannot be omitted
from any fire hazard analysis. The assumptions, bounding condi- Ignition. The design fire curve starts at ignition. A simple
tions, scenarios considered, and limitations should be identified approach to developing a design fire curve is to assume that
to the AHJs, the owner, and other interested parties. See Section an ignition source of sufficient intensity is available to instan-
3, Chapter 11, “Overview of Performance-Based Fire Protection taneously ignite the initial fuel package to establish burning.
Design,” for more details on documentation requirements. However, if the heat transfer to a combustible object or the tem-
perature of the object is known, calculations can be performed
to predict whether the object will ignite. Calculations to deter-
QUANTIFICATION OF DESIGN mine whether ignition occurs depend on the state of the fuel:
FIRE SCENARIOS solid, liquid, or gas.
Ignition can be divided into two categories: piloted and
Quantification of design fire scenarios involves two steps. The
nonpiloted. In the case of piloted ignition, a “pilot” such as a
first step is to develop the design fire curve for the design fire
spark or flame initiates flaming. For nonpiloted ignition, flam-
scenario or portion of the design fire scenario of interest. The
ing occurs spontaneously as a result of heating in the absence of
design fire curve represents the heat release rate over time for
flame or spark.3 Except for piloted ignition of gases and liquids
the fire in question. Once the design fire curve is estimated, the
that are at a temperature above their flashpoint, all materials
second step, predicting the fire effects, is then possible.
must first be heated before ignition can take place.3
The purpose of the design fire is similar to the assumed
loading in a structural analysis—that is, to answer the question
of whether the design will perform as intended under the as- Solids. With the exception of smoldering combustion, for
sumed challenge. Keeping in mind that the greatest challenge is a solid to ignite it must first be heated sufficiently to release
not necessarily the largest fire (especially in a sprinklered build- flammable vapors. Flammable vapors can be given off either by
ing), it is helpful to think of design fires in terms of their growth pyrolysis or by melting and subsequent vaporization. Pyroly-
phase, steady-burning phase, and decay phase (Figure 3.7.1). sis occurs when a material is heated and decomposes, releasing
vapors known as pyrolyzates. Unlike melting and vaporization,
Design Fire Curves in which no molecular changes occur, the vapors given off are
different from the material that was originally heated.4 The pro-
The design fire curve is a description of the intensity (heat release cess of pyrolysis can be viewed as “thermal cracking,” in which
rate) of a fire as a function of time. The design fire curve can larger molecules are broken into smaller molecules.
be divided into four phases: ignition, growth, steady-burning, Piloted ignition occurs if the concentration of pyrolysis
and decay. Because there is not a single framework for develop- gases is above the lower flammable limit and a “pilot” is pres-
ing the entire design fire curve, each step is typically developed ent. For nonpiloted ignition to occur, the pyrolysis gases must
separately and then brought together as a single curve. be at a concentration above the lower flammable limit and they
It is not always necessary to quantify each phase of a design must be above their autoignition temperature. Because of this,
fire curve, depending on the goals of the analysis. For example, it requires less energy for piloted ignition to occur than for non-
to predict when a fire detection or suppression system would ac- piloted ignition.4
tivate, it might only be necessary to quantify the growth phase. Methods of predicting ignition of solid materials exposed
For sizing a smoke control system, only the maximum heat re- to thermal radiation differ depending on whether a solid is ther-
lease rate might be needed. A structural analysis might need the mally thin or thermally thick. A thermally thick material is one
peak burning rate and the duration of peak burning. Perform- in which a temperature rise will not be perceived on the unex-
posed surface when the material is heated. Wood is a typical
example of a thermally thick material, whereas sheet metal is a
Growth phase of fire
good example of a thermally thin material.
An engineering guide4 published by the Society of Fire
Protection Engineers (SFPE) focusing on piloted ignition con-
Flashover tains six methods for predicting the piloted ignition of solid ma-
terials under radiant exposure as follows.
For thermally thin materials, the method of Mikkola and
Temperature rise

Fully developed fire

Wichman can be used:
Growth
(Tig – T0)
tig = ρL0c · (1)
Ignition (q″r – q· ″crit)

Decay where

Time
Tig = Ignition temperature (°C)
T0 = Initial temperature (°C)
FIGURE 3.7.1 Design Fire Structure
tig = Time to ignition (sec)
3-126 SECTION 3 ■ Information and Analysis for Fire Protection

ρ = Density of the material (kg/m3) Ignition of a liquid at its flashpoint is analogous to piloted ig-
nition of a solid, in that for ignition to occur, a pilot must be present.
c = Specific heat of the material (kJ/kg·C)
The analogy for nonpiloted ignition of liquids would be ignition
L0 = Thickness of the material (m) at the autoignition temperature. Values for flashpoints and autoig-
nition temperatures for some common materials can be found in
q· ″ = External heat flux (kW/m2)
r NFPA 497, Recommended Practice for the Classification of Flam-
q· ″crit = Critical heat flux for ignition (kW/m2) mable Liquids, Gases, or Vapors and of Hazardous (Classified)
Locations for Electrical Installations in Chemical Process Areas.
For thermally thick materials, the following methods can be
used: Gases. For ignition of a flammable gas to occur, it must be
• Mikkola and Wichman mixed with a sufficient quantity of oxygen for a reaction to take
(Tig – T0)2 place. Concentrations where this occurs are represented by a
π
tig = kρc · (2) flammability range, which corresponds to gas/air concentrations
4 (q″r – q· ″crit)2 that are at or above the lower flammable limit and not exceeding
where k is thermal conductivity (W/m·K). the upper flammable limit. Flammability limits for a variety of
gases can be found in NFPA 497.
• Tewarson For mixtures of flammable gases, Le Chatelier’s principle
π (TRP)2 can be used to determine the lower flammable limit.3 Le Chat-
tig = (3) elier’s law states that
4 (q· ″r – q· ″min)2
where TRP is thermal response parameter (kW·sec1/2/m2) 100
Lm = (7)
and q· ″min is minimum heat flux for ignition (kW/m2). }P
i
• Quintiere and Harkleroad i Li
‹ ·
2 where
q″min
tig = for t ≤ tm (4)
b · q· ″r Lm = Lower flammability limit of the mixture
where b is a constant related to kρc (sec–1/2)
and tm is char- Pi = Volume fraction of gas i
acteristic time to reach thermal equilibrium (sec).
Li = Lower flammable limit of gas i
• Janssens
‹
‹ ·
–1.83 For more information, see Section 6, Chapter 10, “Gases.”
kρc q″r
tig = 0.563 – 1 (5)
h2ig q· ″crit
Fire Growth. Following ignition, a fire might grow as it de-
velops on the first item ignited or spreads to additional items.
where hig is heat transfer coefficient at ignition, which in-
To determine whether spread would occur to adjacent items,
corporates both the radiative and convective components
the problem can be approached from the perspective of whether
(W/m2 ·C).
or not these items would ignite. For growth involving a single
• Toal, Silcock, and Shields item, the fire could spread to unignited portions of the item. This
(FTPn) could either lead to the entire item burning, or earlier ignited
tig = · (6) portions might burn out before the fire spreads to involve the
(q″r – q· ″crit)n entire item, such that the entire item is never fully involved.
where FTPn is flux time product and n is flux time product
index is greater than or equal to 1. Prediction of Fire Effects
See SFPE’s engineering guide4 for additional information The primary importance of the appropriate selection of the de-
on applying these methods as well as the appropriateness of sign fire’s growth is in obtaining a realistic prediction of detector
these methods for different situations. and sprinkler activation, time to start of evacuation, and time to
initial exposure of occupants. In 1972, Heskestad first proposed
Liquids. For a liquid to ignite, it must be at a temperature that is that for the early fire growth period the assumption that fires
equal to or greater than its flashpoint. NFPA 30, Flammable and grow according to a power law relation works well and is sup-
Combustible Liquids Code, defines flashpoint as “the minimum ported by experimental data.5 He suggested fires of the form
temperature of a liquid at which sufficient vapor is given off to
Q = αtn (8)
form an ignitable mixture with air, near the surface of the liquid
or within the vessel used.” where
A number of test methods can be used to measure the flash-
Q = Rate of heat release (kW)
point of a liquid. Flashpoint is not a physical property and is in-
stead a model of physical phenomena associated with vaporization α = Fire intensity coefficient (kW/secn)
of a sufficient quantity of fuel to establish a gaseous mixture that
t = Time (sec)
is at the lower flammable limit at a distance above the fuel surface
and therefore can change with the test method employed.3 n = 1, 2, 3
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-127

Later, it was shown that for most flaming fires (except 22 lb (10 kg) of wood arranged in a solid cube, as sticks ar-
flammable liquids and some others) n = 2, the so-called t- ranged in a crib, and as a layer of sawdust (Figure 3.7.3). These
squared growth rate.6 A set of specific t-squared fires labeled three arrangements would have significantly different growth
slow, medium, and fast, with fire intensity coefficients such that rates although representing identical fuel loads.
the fires reached 1000 Btu/sec (1055 kW) in 600, 300, and 150
seconds, respectively, were proposed for design of fire detection Steady Burning. Where a fire scenario involves a fire in an
systems.7 Later, these specific growth curves and a fourth called enclosure, fire growth might continue until all the combustible
“ultrafast,”8 which reaches 1055 kW in 75 seconds, gained favor items within the room are involved. Once this occurs, the rate
in general fire protection applications. of burning is influenced by one of two factors: (1) the available
This set of t-squared growth curves is shown in Figure ventilation or (2) the available fuel. Calculation of fire tempera-
3.7.2. The slow curve is appropriate for fires involving thick, tures within the room is easily accomplished by use of simple
solid objects (e.g., solid wood table, bedroom dresser, or cabi- algebraic equations. Although computer models are frequently
net). The medium growth curve is typical of solid fuels of lower used in hazard analyses, they are generally no more accurate
density (e.g., upholstered furniture and mattresses). Fast fires (and indeed may be less accurate) than simple hand calcula-
are thin, combustible items (e.g., paper, cardboard boxes, drap- tions for prediction of temperature and burning rate during fully
eries). Ultrafast fires are some flammable liquids, some older developed burning.10 For example, for postflashover fires, hand
types of upholstered furniture and mattresses, or materials con- calculation methods are generally used to estimate compartment
taining other highly volatile fuels. temperatures.11
These t-squared curves represent fire growth starting with SFPE’s engineering guide11 on fire exposures of structural
a reasonably large, flaming ignition source. With small sources, elements provides calculation methods for predicting fire tem-
there is an incubation period before established flaming, which peratures and burning rates in fully developed compartment
can influence the response of smoke detectors. During this in- fires. Some of these methods are based on an assumption of
cubation period, the fire may not significantly grow in size, al- ventilation-limited burning, and others model fuel-controlled
though smoke would still be produced in quantities potentially conditions. For most cases, the method developed by Law was
sufficient to activate smoke detectors. found to provide bounding predictions when the “Ψ” factor was
This specific set of fire growth curves has been incorporated not used and the predicted burning duration was increased by a
into several design methods, such as that for the design of fire factor of 1.4.
detection systems in NFPA 72®, National Fire Alarm Code®. Law’s method is as follows:
They are also referenced as appropriate design fires in several
T = Tgm(1 – e–0.05Ψ )(ºC) (9)
international methods for performing alternative design analyses
in Australia and Japan and in a product fire risk analysis method ¡ ¢
A
published in this country.9 Although in the Australian method- –0.1
Ao√Ho
ology the selection of growth curve is related to the fuel load 1–e
(mass of combustible material per unit floor area), this is not
Tgm = 6000 … (ºC)
A
justified, since the growth rate is related to the form, arrange-
ment, and type of material and not simply its quantity. Consider
£ Ao√Ho ¤

Π
‹
A
W –0.036
Ao√Ho
Corrugated cardboard cartons
4.6 m (15 ft) high
Wood pallets Cotton/polyester m· f = 0.18Ao √Ho 1–e (kg/sec)
1.5 m (5 ft) high innerspring mattress D
various contents
Full mail bags where
Thin plywood wardrobe 1 m (3 ft) high
Solid wood
Methyl alcohol pool Upholstered cabinetry Tgm = Maximum compartment temperature (ºC)
furniture
6000 A = Surface area of interior of enclosure (m2)

5000 Ultra Ao = Area of ventilation opening (m2)

Heat release rate (kW)

fast Fast Slow

4000 Medium

3000

2000

1000

0
0 200 400 600 800 Solid cube Sticks (crib) Sawdust
Time from ignition (sec)
FIGURE 3.7.3 Dependence of Fire Growth on Fuel Form
FIGURE 3.7.2 Set of t-Squared Growth Curves and Arrangements
3-128 SECTION 3 ■ Information and Analysis for Fire Protection

Ho = Height of ventilation opening (m) 1

mf
Ψ=
√ A × Ao
mf = Mass of fuel (kg)
m· = Mass burning rate of fuel (kg/sec)
f

Fraction of initial HRR

W = Width of wall containing ventilation opening (m) 0.4

D = Depth of compartment (m)

0.3
Law reports that the correlation for predicting burning rate
is valid for
‹

m· f D ½ 0.2
< 60
Ao√Ho W
In some cases, it may only be desired to predict whether flash-
over is possible for a given fire scenario involving a fire in an
enclosure. In such cases, the approach described in the section 0.1
Prediction of Flashover can be used.12 0 200 400 600 800
Time (sec)
Decay. All fires eventually decrease in size. A fire can decay e [-0.0023(t-t act )] Paper cart fuel package
for one of three reasons: consumption of available fuel, oxygen Secretarial desk fuel package Executive desk fuel package
depletion, or suppression. Because the hazards posed during the Office II fuel package Office I fuel package
Sofa fuel package Workstation I fuel package
decay phase are typically insignificant in comparison to the haz- Workstation II fuel package Wood cribs fuel package
ards posed during the fully developed phase, decay is typically
omitted from analysis. An exception is in calculations involving
FIGURE 3.7.4 Decay Rates for Various Fuels
structural fire resistance of concrete or insulated steel. Where
test data are available, they might include decay.
If decay occurs due to the exhaustion of fuel, Table 3.7.1
shows the expected temperature change as fuel is depleted.11 to occur in a room. The time at which flashover occurs can be
For fires with a predicted duration of less than 60 minutes, a estimated by determining when the fire is predicted to reach this
decay rate of 10°C/min can be used. In other case, a decay rate minimum size. The following methods can be used to predict
of 7°C/min can be used. the minimum heat release rate necessary for flashover.12
Decay could also occur in the event that a sprinkler system
is present and activated. A simple assumption is that the fire im- • Method of Babrauskas
·
mediately goes out, but this is not conservative. A National In- Q = 750Ao√Ho (9)
stitute of Standards and Technology (NIST) study documents a
(conservative) exponential diminution in burning rate under the where
application of water from a sprinkler (Figure 3.7.4).13 Since the ·
Q = Minimum heat release rate required for flashover
combustion efficiency is affected by the application of water, (kW)
the use of values for soot and gas yields appropriate for post-
flashover burning would represent the conservative approach in Ao = Area of opening into compartment (m2)
the absence of experimental data. Ho = Height of opening into compartment (m)

Prediction of Flashover. Flashover occurs when a fire grows • Method of McCaffrey, Quintiere, and Harkleroad
·
to such a size that it involves all combustible items within an en- Q = 610(hk AT Ao√Ho )½ (10)
closed room. Although occurrence of flashover is not a hazard in
itself, flashover would affect the occurrence of other hazards as where
k
described in the next section. Several correlations are available hk =
to predict the minimum heat release rate necessary for flashover δ
k = Thermal conductivity of compartment surface
(kW/m·K)
δ = Thickness of compartment surface (m)
TABLE 3.7.1 Rate of Decrease in Temperature
AT = Total area of compartment surfaces (m2)
Temperature Decay (°C/min) Restrictions
• Method of Thomas
10 Duration < 60 minutes ·
Q = 7.8AT + 378Ao√HO (11)
7 Duration > 60 minutes
where the variables are as defined above.
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-129

PREDICTION OF HAZARDS T1 = Temperature of gas (K)

Fire is a dynamic process of interacting physics and chemis- T2 = Temperature of target (K)
try, so predicting what is likely to happen under a given set of The equation is only applicable for instantaneous calcula-
circumstances is daunting. The simplest predictive methods are tions, as the temperature of the target will rise as a function of
algebraic equations. Computer models are used to automate fire the thermal radiation that it receives.
hazard calculations and are particularly useful where many re-
peated calculations must be performed. Smoke Production. When calculating smoke production rates,
smoke is usually defined as the products of combustion and the
Simple Fire Hazard Calculations air entrained into the fire plume. Therefore, the amount of smoke
Once the design fire curve has been developed, it is then pos- produced is a function of the height above the fire. Section 3,
sible to predict the hazards that would result. The types of haz- Chapter 9, “Closed Form Enclosure Fire Calculations,” pro-
ards that might be of interest include the following: vides a number of equations that can be used to predict smoke
production.
• Radiant heat flux, which affects the potential for ignition of
materials or thermal injury to people Fire Plumes and Ceiling Jet Temperatures and Velocities. A
• Smoke production, which dictates the volume of smoke fire will produce a plume of hot gas that will rise and contact the
produced ceiling of a compartment, forming a ceiling jet. The temperature
• Fire plume and ceiling jet temperatures and veloci- and velocity of a plume can be calculated as described in Sec-
ties, which could cause weakening of exposed structural tion 3, Chapter 9, “Closed Form Enclosure Fire Calculations.”
elements Similarly, the temperature and velocity of a ceiling jet can be
• Species production, which affects the rate at which an un- calculated in accordance with the following equations:14
tenable environment could be created ·
• Depth of upper layer, which can be used as a surrogate for r 16.9Q 2/3
For 0.18 ≥ ∆T = (13)
an untenable environment H H 5/3
·
As was the case with the stages of design fire curves, it is r 5.38(Q/r)2/3
For > 0.18 ∆T = (14)
not always necessary to quantify all of the hazards that result H H
from a design fire scenario. The hazards that are quantified are a ‹ ·
1/3
r Q
function of the goals of the analysis. For example, if the purpose For 0.15 ≥ U = 0.96 (15)
of the analysis is to determine whether a thermally activated H H
detection or suppression system activates, only the plume and ·
r 0.195Q 1/3H1/2
ceiling jet temperatures and velocities might be determined. For For > 0.15 U= (16)
H (r/H)5/6
analysis of a smoke control system, only the smoke production
rate might be determined. A structural analysis might only re- where
quire calculation of the heat transfer to the structure. An evacu- ∆T = Temperature rise over ambient (°C)
ation analysis might require quantification of all of the hazards
listed. U = Ceiling jet velocity (m/sec)
H = Height above fire (m)
Radiant Heat Flux. Radiant heat flux is a measure of the rate
of radiative heat transfer per unit area. An example of radiant r = Horizontal distance from fire centerline (m)
heat transfer is the heating that can be felt from exposure to the ·
Q = Total heat release rate (kW)
sun on a hot day (although the intensity of thermal radiation in
sunlight is too small to be of concern from a fire standpoint). When using these equations, it must be cautioned that they
The radiant heat flux from a single burning item can be pre- are only valid for horizontal, unobstructed ceilings where there
dicted as a function of the distance from the item in accor- is no smoke layer present. In cases where a layer forms, higher
dance with Equation 18 of Section 3, Chapter 9, “Closed Form temperature rises can be expected.
Enclosure Fire Calculations.” For radiant heat fluxes resulting
from fire gases, such as in a compartment fire, the radiant heat Species Production. Fires can create a number of products of
flux can be calculated if the gas temperature and the tempera- combustion that can be toxic or corrosive, including carbon di-
ture of the target object are known by applying the following oxide (CO2), water vapor (H2O), carbon monoxide (CO), and
equation: many others that vary with the fuel and burning conditions.
· Species production rates can be calculated from the following
Qr = εσ(T14 – T24) (12)
equation:15
where ·
· Q
· Gj = yj (17)
Qr = Rate of radiant heat transfer (kW) ∆Hc
ε = Emissivity of gas (0–1) (—) where
·
σ = Stephan-Boltzmann constant (5.67 × 10–11 kW/m2·K4) Gj = Smoke production rate of species j (kg/sec)
3-130 SECTION 3 ■ Information and Analysis for Fire Protection

yj = Yield fraction of species j (–) FLCirr = The fraction of the incapacitating dose from all
· incapacitating products (HCl, HBr, etc.)
Q = Heat release rate (kW) ‹

[CO2]
VCO2 = exp , where [CO2] is the concentration of
∆Hc = Heat of combustion of fuel (kJ/kg) 5
carbon dioxide in percent
Yield fractions for several fuels are available in the SFPE
Handbook of Fire Protection Engineering.15 FEDI = {exp[8.13 – 0.54(20.9 – [O2])]}–1, where [O2] is
o
the concentration of oxygen in percent

Depth of Upper Layer. As smoke is produced in a compart- FI = {exp[6.1623 – 0.5189[CO2]]}–1, where [CO2] is
co2
ment, it forms a layer that descends as a function of time. This the concentration of CO2 in percent
is analogous to filling a bowl of water. Section 3, Chapter 9, It should be noted that the equations for FED and the com-
“Closed Form Enclosure Fire Calculations,” provides an equa- ponents of FED are based on a one minute exposure. For expo-
tion that can be used to estimate the velocity of descent of the sures to constant concentrations of fire products, the FED can
smoke layer. However, when applying this equation, it should be determined by multiplying the value determined using the
be noted that the mass production rate of smoke is not constant, previous equations by the exposure time in minutes. For expo-
since as the layer descends, the smoke production rate decreases sures where the concentrations vary with time, the total FED
due to the reduced vertical distance available to entrain air into can be calculated by discretizing the exposure (determining the
the plume. (See the equations for predicting smoke production average exposure at each one minute interval and summing the
rate in the same chapter.) FED determined for each one minute interval).
It should be noted when applying the previous correlations
Toxicity. Toxic gases produced by a fire can incapacitate or kill that some populations are more susceptible than others to fire
people who are exposed to them. A commonly used approach products (e.g., asthmatics, the old, and the young). Additionally,
to determine whether the fire-induced environment is potentially no single FED value for design has been widely agreed on even
harmful to people exposed is the “fractional effective dose” (FED) for “average” populations. A report by the National Institute of
model developed by Purser.16 This can be expressed as follows: Standards and Technology investigates this subject in detail.17
FIN = [(FI + FI + FLCirr) × VCO2 + FEDI ]
co cn o Egress Calculations. Evacuation calculations are sometimes
or FIco (18) simple enough to be done by hand. The most thorough presentation
2
on this subject (and the one that is most often used in alternative
where
design analysis) is that of Nelson and Mowrer.18 See also Section
FIN = Fraction of an incapacitating dose of all asphyxiat- 4, Chapter 2, “Calculation Methods for Egress Prediction.”
ing gases
FI
Simple Analytical Solution Techniques
= Fraction of an incapacitating dose of CO
co
Simple computer programs and spreadsheets can be used to
FI = Fraction of an incapacitating dose of HCN
cn perform simple fire hazard calculations. In the case of the equa-
FLCirr = Fraction of irritant dose tions listed previously or referenced in other chapters, this is a
relatively straightforward task. However, many fires and fire ef-
VCO2 = Multiplication factor for CO2-induced hyper-
fects are not steady state. An example is smoke filling within an
ventilation
enclosure. The smoke production rate is a function of the smoke
FEDI = Fraction of an incapacitating dose of low-oxygen layer height, so the rate of smoke layer descent is not constant.
o
hypoxia In such instances, spreadsheets can be used to develop solutions
to differential equations for which developing an exact solution
FI = Fraction of an incapacitating dose of CO2
co2 is nontrivial.
Purser gives the following equations for calculation of the For a differential equation of the following form:
individual fractional effective doses:16
dy
8.2925 × 10–4 × [CO]1.036 = f(y, t) (21)
FI = (19) dt
co 30
where the initial value y(t = 0) is y0.
where [CO] is the concentration of CO, expressed in parts per The Euler method is a numerical technique for solving dif-
million. ferential equations of this form, and can be stated as
exp (CN/43) yn+1 = yn + hf(yn, tn) (22)
FI = (20)
cn 220
where
where
yn = Value of equation y at time step n
CN = The concentration of HCN in parts per million
yn+1 = Value of equation y at time step n + 1
added to the concentration of other nitriles minus
the concentration of NO2 h = Time step size
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-131

This process can be iterated over the desired length of time Equation 24 could easily be programmed into a spread-
to obtain the desired solution. Since the Euler method deter- sheet or simple computer program, along with the ceiling jet
mines the value of equation y at time step n + 1 based on the temperature and velocity correlations expressed in Equations 13
value at time step n and the slope of the tangent to y at time step through 16, to calculate Ug and Tg. It is also necessary to include
n, errors can be introduced based on the nonlinearity of equation a method of calculating the heat release rate at each time step.
y. There are methods available to reduce this error, such as the The following example illustrates the use of this method in
improved Euler method. estimating thermal detector response. A heat detector is located
However, another method of reducing the error is to re- on a 3 m high ceiling, 2 m away from a fire located on the floor
duce the size of the time step, recognizing that as the size of with a constant heat release rate of 1000 kW. The room has
the time step approaches zero, the difference between the pre- an ambient temperature of 20°C and is sufficiently large that
dicted value of y and the actual value of y also approaches a smoke layer will not form quickly. The heat detector has a
zero. The computational power offered by modern computers temperature rating of 75°C and an RTI of 50 m½sec½. When
allows very small time steps to be used and to still get a solu- would the detector operate? Using Equation 24 and Equations
tion rapidly. 13 through 16, the solutions in Table 3.7.2 are obtained when
It should be noted that the default for many spreadsheets is different time steps are used.
to not permit iterative calculations. (The spreadsheet views the
“circular reference” as an error.) Spreadsheets for which this Computer Models
is the case would need to be configured to allow iteration. The
spreadsheet’s user’s manual or help function can be consulted Fire Models. A survey21 documented available models and
for assistance. calculation methods that could be applied to FHA. The key
to determining which are appropriate to a given situation and
example. Thermal detector response can be used to illus- which are not is a thorough understanding of the assumptions
trate application of the Euler method to a fire protection prob- and limitations of the individual model or calculation and how
lem. This example uses an algorithm similar to that used by these relate to the situation being analyzed. Single-room models
the computer fire model DETACT-QS19 to predict the time to are appropriate where the conditions of interest are limited to a
activation of a thermal detector for a heat release rate that fol- single, enclosed space. Where the area of interest involves more
lows a power law curve. Two calculations will be performed. than one space, and especially where the area of interest extends
First, the instantaneous ceiling jet velocities are calculated in beyond a single floor, multiple-compartment models should be
accordance with Equations 13 through 16. A quasi-steady as- used. This is because the interconnected spaces interact to influ-
sumption is made, which means that transport delays from the ence fire development and flows.
fire to the detector are ignored. Then, based on the tempera- Many single-compartment models assume that the lower
ture and velocity of the ceiling jet and the thermal response layer remains at ambient conditions (e.g., ASET).22 Since there
characteristics of the sensor (response time index), the tem- is little mixing between layers in a room (unless there are me-
perature change at the detector will be calculated. See Sec- chanical systems), these models are appropriate. However, sig-
tion 16, “Water-Based Fire Suppression Equipment” for more nificant mixing can occur in doorways, so multiple-compartment
information. The change in temperature of the detector can be models should allow the lower layer to be contaminated by en-
expressed as20 ergy and mass (Figure 3.7.5).
dTd √Ug(Tg – Td) The model should include the limitation of burning by
= (23) available oxygen. This is straightforward to implement (based
dt RTI on the oxygen consumption principle) and is crucial to obtain-
where ing an accurate prediction for ventilation-controlled burning.
For multiple-compartment models, it is equally important for
Td = Temperature of detector (°C)
the model to track unburned fuel and allow it to burn when it
Ug = Ceiling jet velocity at detector (m/sec) encounters sufficient oxygen and temperature. Without these
features, the model concentrates the combustion in the room
Tg = Ceiling jet temperature at detector (°C)
of origin, overpredicting conditions there and underpredicting
t = Time (sec) conditions in other spaces.
RTI = Detector response time index (m1/2 × sec1/2)
A Euler solution to this expression can be expressed as
√Ug(Tg – Td ) TABLE 3.7.2 Time Step(s) Predicted Activation Time(s)
Td T + ∆t n (24)
n+1 + dn
RTI Time Step(s) Predicted Activation Time(s)
where
1 24.000
Td = Temperature of detector at time step n + 1 (°C) 0.1 24.200
n+1
Td = Temperature of detector at time step n (°C) 0.01 24.230
n 0.001 24.229
∆t = Size of time step (sec)
3-132 SECTION 3 ■ Information and Analysis for Fire Protection

Smoke layer mixing at doorway available to any user of the model, and a copy should be sup-
plied with the analysis as an important supporting document.
Tµ,i Tµ,j
Input Data. Even if the model is correct, the results can be seri-
ously in error if the data that are input to the model do not rep-
Doorjet resent the condition being analyzed. The FHA should include a
Tl,j Plume listing of all data values used, their source (i.e., what apparatus
or test method was employed and what organization ran the test
and published the data), and some discussion of the uncertainty
Tl,j
of the data and its result on the conclusions. The National Insti-
tute of Standards and Technology’s (NIST) website contains a
section of well-documented data for use in calculations, called
Fire on the Web (http://fire.nist.gov). A much larger database,
called FASTDATA, is available from NIST on a CD-ROM (see
the URL above for information). (See also the subsection en-
titled “Accounting for Uncertainty” later in this chapter.)
FIGURE 3.7.5 Assumption of Zone Models That Fire Gases
Collect in Internally Uniform Layers
Egress Models. The prediction of the time needed by the build-
ing occupants to evacuate to a safe area can be performed and
compared to the predicted available safe egress time. Whether
Heat transfer calculations take up a lot of computer time, the evacuation calculation is done by model or hand calculation,
so many models take a shortcut. The most common is the use it must account for several crucial factors. First, unless the oc-
of a constant “heat loss fraction,” which is user-selectable (e.g., cupants see the actual fire, time is required for detection and no-
ASET or CCFM23). The problem is that heat loss can vary dur- tification before the evacuation process can begin. Next, unless
ing the course of the fire. the information is compelling (such as seeing the actual fire), it
Another problem can occur in tall spaces, for example, atria. takes time for people to decide to take action. The action they
The major source of gas expansion and energy and mass dilu- choose may or may not be evacuation. Finally, the movement
tion is entrainment of ambient air into the fire plume. It can be begins. All of these factors require time, and that is the critical
argued that in a very tall plume, this entrainment is constrained. factor. No matter how the calculation is done, all of the factors
However, most models do not include this constraint, which can must be included in the analysis to obtain a complete picture. An
lead to an underestimate of the temperature and smoke density excellent discussion of this topic is found in SFPE’s Engineer-
and an overestimate of the layer volume and filling rate—the ing Guide—Human Behavior in Fire.26
combination of which may give predictions of available safe The process of emergency evacuation of people follows the
egress times that are either greater or less than the correct value. general concepts of traffic flow. A number of models perform
In the model CFAST,24 this constraint is implemented by stop- such calculations and may be appropriate for use in certain oc-
ping entrainment when the plume temperature drops to within cupancies. Most of these models do not account for behavior
1°C of the temperature just outside the plume, where buoyancy and the interaction of people (providing assistance) during the
ceases. event. The literature reports incidents of providing assistance to
disabled persons, again especially in office settings.27 If such
Documentation. Only models that are rigorously documented behavior is expected, it should be included, as it can result in
should be allowed in any application involving public health, significant delays in evacuating a building.
safety, or welfare, such as in code enforcement or litigation. Crowded conditions, as well as smoke density, can result in
This means that the model should be supplied with a technical reduced walking speeds.28 A person’s walking speed decreases
reference guide that includes a detailed description of the in- in dense smoke until he or she moves as if blindfolded (Figure
cluded physics and chemistry, with proper literature references; 3.7.6). Care should be exercised in using models relative to how
a listing of all assumptions and limitations of the model; and they select the path (usually the shortest path) that the person
estimates of the accuracy of the resulting predictions, based on travels. Some models are optimization calculations that give the
comparisons to experimental data. Public exposure and review best possible performance.
of the exact basis for a model’s calculations, internal constants,
and assumptions are necessary for it to have credibility in a
Analyzing the Impact of Exposure
regulatory application.
ASTM publishes a Standard Guide for Documenting Com- In most cases, the exposure will be to people, and the methods
puter Software for Fire Models, ASTM E1472-05.25 Documen- used to assess the impacts of exposure of people to heat and
tation for any model used in a regulatory application should combustion gases involve the application of combustion toxi-
comply with this guide. Although it may not be necessary for cology models. The HAZARD I software package contains the
the full source code to be available, the method of implementing only toxicological computer model, called TENAB,29 that is
key calculations in the code and details of the numerical solver based on research at NIST on lethality to rats and by Purser on
used should be included. This documentation should be freely incapacitation of monkeys. TENAB accounts for the variation
 CHAPTER 7 ■ Fire Hazard Analysis Techniques 3-133

1.5 analysis (i.e., varying the data to the limits and seeing whether
Irritating smoke the conclusions change) should be performed. This is also a
Nonirritating smoke good time to justify the appropriateness of the model or calcu-
lation method. For more information, see Section 3, Chapter 10,
Walking speed (m/sec)

1.0 “Performance-Based Codes and Standards for Fire Safety.”

Final Review
If a model or calculation produces a result that seems counter-
0.5
intuitive, there is probably something wrong. Cases have been
seen in which the model clearly produced a wrong answer (e.g.,
Walking speed of the temperature predicted approached the surface temperature
blindfolded person of the sun), and there have been others in which it initially
0 looked wrong but was not (e.g., a dropping temperature in a
0 0.2 0.4 0.6 0.8 1.0 1.2
space adjacent to a room with a growing fire was caused by cold
Extinction coefficient, α (1/m) air from outdoors being drawn in an open door). Conversely, if
the result is consistent with logic, sense, and experience, it is
FIGURE 3.7.6 Reduced Walking Speeds Resulting from probably correct. This is also a good time to consider whether
Crowded Conditions and Smoke Density the analysis addressed all of the important scenarios and likely
events. Were all the assumptions justified and were uncertainties
addressed sufficiently to provide a comfort level similar to that
obtained when the plan review shows that all code requirements
in exposure to combustion products as people move through a have been met?
building, by reading the concentrations from the fire model in
the occupied space during the time the person is in that space.
If the person moves into a space with a lower concentration of SUMMARY
carbon monoxide, the accumulated dose can decrease. Details
such as these ensure that the results are reasonable. Quantitative fire hazard analysis is becoming the fundamental
Assessing the impact of exposure to sensitive equipment is tool of modern fire safety engineering practice and is the en-
more difficult, since little data exist in the literature on the effects abling technology for the transition to performance-based codes
of smoke exposure on such equipment. Of particular importance and standards. (For more information on performance-based
here is the existence of acid gases in smoke, which are corrosive codes, see Section 3, Chapter 10, “Performance-Based Codes
and especially harmful to electronics. Fuels containing chlorine and Standards for Fire Safety.”) The tools and techniques de-
(e.g., polyvinyl chlorides) have been studied. However, unless scribed in this chapter provide an introduction to this topic and
the equipment is close to the fire, acid gases, especially HCl, the motivation for fire protection engineers to learn more about
deposit on the walls and lower the concentration to which the the proper application of this technology.
equipment may be exposed. CFAST in the HAZARD I package Predicted fire hazards are a function of the design fire
contains a routine that models this process and the associated scenarios analyzed. Therefore, when performing a fire haz-
diminution of HCl concentration. ard analysis, it is important to select design fire scenarios that
are challenging enough to represent a realistic “worst case,”
Accounting for Uncertainty but not so challenging that the likelihood of occurrence is too
remote.
Uncertainty analysis refers to dealing with the unknowns and There are many fire hazard calculations that can be per-
variation inherent in any prediction. In the calculations, this un- formed with a hand calculator, a simple spreadsheet, or a com-
certainty is derived from assumptions in the models and from the puter program. In some cases, these simple methods would not
representativeness of the input data. In evacuation calculations, be sufficient, for example, in cases where compartment geom-
there is the added variability of any population of real people. In etry is complex, where it is desired to optimize cost/benefit, or
building designs and codes, the classic method of treating uncer- where predicted hazard values are very close to acceptable lim-
tainty is with safety factors. A sufficient safety factor is applied its. However, even in these types of cases, simple methods can
such that, if all of the uncertainty resulted in error in the same be used for initial predictions or as a reality check of results
direction, the result would still provide an acceptable solution. from more complex models.
In the prediction of fire development/filling time, the intent is In any engineering analysis, it is incumbent on the user
to select design fires that provide a worst likely scenario. Thus a to understand the application and limitation of any methods
safety factor is not needed here, unless assumptions or data are used. This chapter has outlined a number of simple fire haz-
used to which the predicted result is very sensitive. ard calculation methods, but the applicability and limitations
The FHA report should include a discussion of uncertainty. of the methods were not included. Users are referred to the
This discussion should address the representativeness of the documents referenced in the text for information regarding the
data used and the sensitivity of the results to data and assump- applications and limitations of any of the methods included in
tions made. If the sensitivity is not readily apparent, a sensitivity this chapter.
3-134 SECTION 3 ■ Information and Analysis for Fire Protection

ACKNOWLEDGMENT DiNenno et al. (Eds.), National Fire Protection Association,

Quincy, MA, 2002.
James R. Quiter, P.E., wrote portions of the discussion of de- 17. Gann, R., et al., “International Study of the Sublethal Effects of
sign fire scenarios, which originally appeared in the “Simple Fire Smoke on Survivability and Health (SEFS): Phase I Final
Report,” NIST Technical Note 1439, National Institute of Stan-
Fire Hazard Calculations” chapter in the 19th edition of this dards and Technology, Gaithersburg, MD, 2001.
handbook. 18. Nelson, H. E., and Mowrer, F. W., “Emergency Movement,”
SFPE Handbook of Fire Protection Engineering, 3rd ed., P. J.
DiNenno et al. (Eds.), National Fire Protection Association,
BIBLIOGRAPHY Quincy, MA, 2002.
19. Evans, D. D., and Stroup, D. W., “Methods to Calculate the Re-
References Cited sponse Time of Heat and Smoke Detectors Installed Below Large
Unobstructed Ceilings,” NBSIR 85-3167, Building and Fire Re-
1. Department of Energy General Memorandum, U.S. Department search Laboratory, U.S. Department of Commerce, Gaithersburg,
of Energy, Germantown, MD, Nov. 7, 1991. MD, 1985.
2. SFPE Engineering Guide to Performance-Based Fire Protection, 20. Evaluation of the Computer Fire Model DETACT-QS, Society of
National Fire Protection Association, Quincy, MA, 2006. Fire Protection Engineers, Bethesda, MD, 2003.
3. Drysdale, D., An Introduction to Fire Dynamics, 2nd ed., Wiley, 21. Olenick, S. M., and Carpenter, D. J., “An Updated International
Chichester, UK, 1998. Survey of Computer Models for Fire and Smoke,” Journal of
4. Engineering Guide: Piloted Ignition of Solid Materials Under Fire Protection Engineering, Vol. 13, No. 2, 2003, pp. 87–110.
Radiant Exposure, Society of Fire Protection Engineers, 22. Cooper, L. Y., and Stroup, D. W., Calculating Available Safe
Bethesda, MD, 2002. Egress Time (ASET)—A Computer User’s Guide, NBSIR 82-
5. Heskestad, G., Similarity Relations for the Initial Convective 2578, National Bureau of Standards, Gaithersburg, MD, 1982.
Flow Generated by Fire, FM Report 72-WA/HT-17, Factory 23. Cooper, L. Y., and Forney, G. P., The Consolidated Compart-
Mutual Research Corporation, Norwood, MA, 1972. ment Fire Model (CCFM) Computer Code Application CCFM.
6. Schifiliti, R. P., Meacham, B. J., and Custer, R. L. P., “Design VENTS—Part 1: Physical Basis, NISTIR 4342, National Insti-
of Detection Systems,” SFPE Handbook of Fire Protection En- tute of Standards and Technology, Gaithersburg, MD, 1990.
gineering, National Fire Protection Association, Quincy, MA, 24. Peacock, R. D., Forney, G. P., Reneke, P., and Jones, W. W.,
2002. CFAST, the Consolidated Model of Fire Growth and Smoke
7. Heskestad, G., and Delichatsios, M. A., Environments of Fire Transport, NIST Technical Note 1299, National Institute of Stan-
Detectors—Phase 1: Effect of Fire Size, Ceiling Height, and dards and Technology, Gaithersburg, MD, 1993.
Material, Vol. 2, Analysis, NBS-GCR-77-95, National Bureau of 25. ASTM E1472, Standard Guide for Documenting Computer Soft-
Standards, Gaithersburg, MD, 1977. ware for Fire Models, American Society for Testing and Materi-
8. Stroup, D. W., and Evans, D. D., “Use of Computer Models als, W. Conshohocken, PA, 2005.
for Analyzing Thermal Detector Spacing,” Fire Safety Journal, 26. Engineering Guide—Human Behavior in Fire, Society of Fire
Vol. 14, 1988. Protection Engineers, Bethesda, MD, 2003.
9. Bukowski, R. W., “A Review of International Fire Risk Predic- 27. Julliet, E., “Evacuating People with Disabilities,” Fire Engineer-
tion Methods,” Interflam ’93, Interscience Communications, ing, Vol. 146, No. 12, 1993.
Ltd., London, UK, 1993. 28. Jin, T., “Visibility Through Fire Smoke,” SFPE Handbook of
10. Hurley, M., “Evaluation of Models of Fully Developed Post- Fire Protection Engineering, 3rd ed., P. J. DiNenno et al. (Eds.),
Flashover Compartment Fires,” Journal of Fire Protection Engi- National Fire Protection Association, Quincy, MA, 2002.
neering, Vol. 15, No. 3, 2005, pp. 173–197. 29. Peacock, R. D., Jones, W. W., Bukowski, R. W., and Forney,
11. Engineering Guide: Fire Exposures to Structural Elements, Soci- C. L., Technical Reference Guide for HAZARD I Fire Hazard
ety of Fire Protection Engineers, Bethesda, MD, 2004. Assessment Method, Version 1.1, NIST Hb 146, Vol. 2, National
12. Walton, W., and Thomas, P., “Estimating Temperatures in Com- Institute of Standards and Technology, Gaithersburg, MD, 1991.
partment Fires,” SFPE Handbook of Fire Protection Engineer-
ing, 3rd ed., P. J. DiNenno et al. (Eds.), National Fire Protection NFPA Codes, Standards, and Recommended Practices
Association, Quincy, MA, 2002.
13. Madrzykowski, D., and Vettori, R. L., Sprinkler Fire Suppres- Reference to the following NFPA codes, standards, and recommended
sion Algorithm for the GSA Engineering Fire Assessment System, practices will provide further information on fire hazard analysis
NISTIR 4833, National Institute of Standards and Technology, techniques discussed in this chapter. (See the latest version of The
Gaithersburg, MD, 1992. NFPA Catalog for availability of current editions of the following
14. Alpert, R., “Ceiling Jet Flows,” SFPE Handbook of Fire Protec- documents.)
tion Engineering, 3rd ed., P. J. DiNenno et al. (Eds.), National NFPA 30, Flammable and Combustible Liquids Code
Fire Protection Association, Quincy, MA, 2002. NFPA 72, National Fire Alarm Code
15. Tewarson, A., “Generation of Heat and Chemical Compounds in NFPA 101®, Life Safety Code®
Fires,” SFPE Handbook of Fire Protection Engineering, 3rd ed., NFPA 497, Recommended Practice for the Classification of Flam-
P. J. DiNenno et al. (Eds.), National Fire Protection Association, mable Liquids, Gases, or Vapors and of Hazardous (Classified)
Quincy, MA, 2002. Locations for Electrical Installations in Chemical Process Areas
16. Purser, D., “Toxicity Assessment of Combustion Products,”
SFPE Handbook of Fire Protection Engineering, 3rd ed., P. J.