&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
FIRE RISK ASSESSMENT
A symposium
sponsored by
ASTM Committee E-5
on Fire Standards
Hilton Head, S.C, 4 June 1980
ASTM SPECIAL TECHNICAL PUBLICATION 762
G. T. Castino, Underwriters Laboratories Inc.,
and T. Z. Harmathy, National Research
Council of Canada, editors
ASTM Publication Code Number (PCN)
04-762000-31
€b
1916 Race Street, Philadelphia, Pa. 19103
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
Copyright © by A M E R I C A N S O C I E T Y FOR T E S T I N G AND M A T E R I A L S
1982
Library of Congress Catalog Card Number: 81-68807
NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.
Printed in Baltimore, Md.
Marcli 1982
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
Foreword
The symposium on Fire Risk Assessment was held 4 June 1980 at Hilton
Head, S.C. ASTM Committee E-5 on Fire Standards sponsored the event.
The symposium chairmen were G. T. Castino of Underwriters Laboratories
Inc. and T. Z. Harmathy of the National Research Council of Canada, both
of whom also served as editors of this publication.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
Related
ASTM Publications
Design of Buildings for Fire Safety, STP 685 (1979), $28.00, 04-685000-31
Fire Standards and Safety, STP 614 (1977), $27.75, 04-614000-31
Ignition, Heat Release, and Noncombustibility of Materials, STP 502 (1972),
$10.00, 04-502000-31
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
A Note of Appreciation
to Reviewers
This publication is made possible by the authors and, also, the unheralded
efforts of the reviewers. This body of technical experts whose dedication, sacrifice of time and effort, and collective wisdom in reviewing the papers must
be acknowledged. The quality level of ASTM publications is a direct function
of their respected opinions. On behalf of ASTM we acknowledge with appreciation their contribution.
ASTM Committee on Publications
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
Editorial Staff
Jane B. Wheeler, Managing Editor
Helen M. Hoersch, Senior Associate Editor
Helen P. Mahy, Senior Assistant Editor
Allan S. Kleinberg, Assistant Editor
Virginia M. Barishek, Assistant Editor
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
Contents
Introduction
1
Assessing the Risk of Fires Systemically—w. D. ROWE
3
A Discussion of Fire Risk Assessment—H. J. ROUX
16
Formulating Acceptable Levels of Fire Risk—B. M. COHN
28
Using Fire Tests for Quantitative Risk Analysis—w. c. T. LING
AND R. B. WILLIAMSON
38
Legal and Economic Criteria for Test-Based Fire Risk Assessment—
V. M. BRANNIGAN AND RACHEL DARDIS
59
Assessment of Fire Risk for Exterior Structural Members—
R. G. GEWAIN
75
Summary
95
bidex
99
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
STP762-EB/Mar. 1982
Introduction
This Special Technical Publication is the result of a symposium on Fire Risk
Assessment held at Hilton Head, S.C, during the June 1980 meetings of
ASTM Committee E-5 on Fire Standards. This subject has for some time
been in the forefront of the interest of several subcommittees of Committee
E-5, among them Subcommittee E05.91 on Planning and Review. The symposium and this publication were the primary elements of a project on the
assessment of fire risk carried out by Subcommittee EOS.32 on Research.
ASTM-sponsored symposia have provided up-to-date information, methodology, and test data for use by the various committees of ASTM and the
community in general. ASTM and Committee E-5 have been conducting
symposia of this kind since 1%1 and continue to find them to be an effective
means of communication between the diverse interests within the fire community. The papers contained in this publication represent the statements of
selected experts from various interest groups and cover a range of important
and topical subjects relating to the assessment of fire risk; the topics were
also designed to provide material that could significantly influence the development and use of fire risk assessment standards and test methods that
would qualify for inclusion in fire risk assessment standards developed by
ASTM Committee E-5 and other standards organizations.
The public perception of fire risk, as with any other risk, depends upon a
complex array of factors, including historical, consequential, economic, and
societal considerations—all of which play roles of varying importance in
establishing "an acceptable level of risk" in any community. Acceptable
levels of fire risk can be obtained only through careful examination of the
nature of fire risk through systemic assessment (Rowe), specific definition of
the risks under consideration (Roux), formulation of new methods for setting
risk levels (Cohn), and applying quantitative test-based analyses (Ling and
Williamson, Brannigan and Dardis, and Gewain). The papers presented in
this publication provide such careful examination.
The fire community is made up of numerous interest groups working on
one or more aspects of the risks involved and, at the same time, functioning
independently and in interrelated programs—all aimed at reducing risks
associated with the use of materials, products, and systems and reducing the
loss of life and property caused by unwanted fires.
Considerable effort has been put forth by Committee E-5, other ASTM
committees, and a broad range of interests in the development of a risk assessment protocol. Subcommittees of Committee E-5—Subcommittee EOS.91
on Planning and Review, Subcommittee EOS.35 on Guide Criteria, Subcom1
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
2
FIRE RISK ASSESSMENT
mittee E05.31 on Terminology, and Executive Subcommittee E05.90—have
studied a variety of metiiodologies and treatises on the subject of risk assessment. These studies have led to the statement of certain definitions of terms
and general guidelines for use in the process of writing fire risk assessment
standards. It is understood that more is needed in the way of specific guidelines before the first comprehensive fire risk assessment standard—one with
sufficient end-use applicability to the majority of diverse interests that make
up the fire community—can be generated.
Some consider that fire risk can be most properly assessed as a scalar
quantity; others define it based upon the nature of the consequences, namely,
the severity of the loss; still others assess fire risk on the basis of societal preferences; finally, the methodological approach is used in the assessment of
fire risk. While the papers presented in this publication deal with these considerations, they are by no means a complete treatise on the subject. Rather,
they represent a unique compilation of careful and topical treatments of the
assessment of fire risk. With this thought in mind. Subcommittee EOS.32
points to two underlying themes that run through all of the papers:
1. Gaining insight into the relative merits of materials, products, and systems with respect to fire behavior requires a coherent fire risk assessment
philosophy, which is prerequisite to the development of sound test standards
and building regulations. Understanding of and adaptation to known fire
performance characteristics of materials, products, and systems, measured
by relevant tests, will permit their use in appropriate environments, commensurate with assigned levels of risk.
2. The illusion of a risk-free society is likely to be dissipated by future
needs; cost-benefit analyses and related economic considerations will indicate our ability or inability to afford extensive reductions of fire risks. Awareness of the need for optimizing our fire risk assessments against social or economic realities is essential.
These papers deal meaningfully with existing risk issues by describing risk
assessment, in relation to fire, as a quantifiable process that can be factored
relevantly into standards development and regulatory processes, thereby providing for the future development of standards and regulations that are less
arbitrary and that interrelate with other material, product, and system standards and regulations, forming a comprehensive fire risk assessment system.
It is the hope of the symposium committee that the information imparted in
this publication will prove immediately useful and. at the same time,
stimulate further review and study.
G. T. Castino
Underwriters
nderwriters Laboratories
Laborat
Inc., Northbrook.
III. 60062; syniposiuni chairman and editor.
T. Z.
Harmathy
National Research Council of Canada. Ottawa, Ontario, Canada KIA, 0R6; symposium chairman and editor.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
W. D. Rowe^
Assessing the Risk of Fires
Systemically
REFERENCE: Rowe. W. D., "Assessing the Risk of Fires Systemically," Fire Risk
Assessment. ASTM STP 762. G. T. Castino and T. Z. Harmathy, Eds., American Society for Testing and Materials, 1982, pp. 3-15.
ABSTRACT: The risks of fire, from either natural or man-made causes, can never be
completely eliminated. The objective of control of risk from fires is to reduce the probability and consequences of events leading to fires to a level where the residual risks are
commensurate with the benefits of society's undertakings. Involuntary and inequitably
distributed risks must be considered as well as voluntary ones.
On a systemic level, the causes of fires may be identified and the risk of their occurrence estimated. This need not be done in detail, but major contributors to accidental
and purposeful fires can be identified, especially in buildings. A second step is addressing
control through prevention and mitigation during and after occurrence. The costeffectiveness of the risk reduction of these approaches provides a first-level ordering of
where to apply resi)urces. The ability to implement such control actions realistically
through standards, inspections, building codes, and so on, provides a means for reordering priorities.
The impact of acceptable levels of risk, established by societal requirements through
the voluntary standard-setting process, is also addressed. Both acceptance and achievement standards are discussed.
KEY WORDS: fire, fire risk assessment
The risk of fire, from either natural or man-made causes, can never be
completely eliminated. The objective of control of risk from fires is to reduce
the probability and consequences of events leading to and resulting from fires
to a level where the residual risks, which are involuntary and inequitably
distributed, are commensurate with the societal benefits of activities leading
to exposure to fires.
The author, a risk analyst rather than an expert in any aspect of fire safety,
believes that a systemic approach to fire safety from a risk assessment point
of view may provide an alternative perspective to the problem of fire safety.
'Director, Institute for Risk Analysis, American University, Washington, D.C. 20016.
3
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
4
FIRE RISK ASSESSMENT
Such a perspective may help in focusing on the roles and needs for controls
and standards of different types in the area of fire safety. The value of such
an approach is considered briefly by examining several aspects of the problem in a sequential manner.
First, a brief overview of the fire safety problem as a system made up of
subsystems will be given. Next, a brief overview of risk assessment methodologies will be described as a base line for understanding what is meant by
risk assessment. Then several important systems and subsystems will be considered in risk assessment terms. Finally, the range of reasonable solutions,
in terms of standards and availability of resources, will be considered.
Fire Safety from a Systems Viewpoint
The problem of fire safety basically involves a set of identifiable systems,
loosely connected in what might be called a metasystem. First, there is a
dichotomy between technological and human approaches to safety. In the
first case, engineering design is used to aid in the prevention, detection, containment, and extinguishing of fires. In the latter, a myriad of "players" are
involved with differing objectives and motives [7].^ For example, there are
people who are involuntarily at risk from fires who may or may not be willing
to invest in increased protection because of occupational considerations,
profit objectives, special interests associated with the application of codes,
and cost considerations.
A second category involves classes of fires characterized by extent, human
loss, damage, building use, and construction characteristics. Another category includes fires characterized by the type of combustion, the combustibles involved, and the progression of the fire. Other subsystems involve
such characteristics as extinguishment and rescue, early detection and containment, or construction for fire resistance.
The risk analyst must put these together in some structured manner in
order to achieve a reasonable approach to risk assessment. The objective is to
determine the spectrum of risks and degrees of control possible, as well as to
determine where resources can best be applied to improve safety to a predetermined level. Since nothing is risk free, the establishment of an objective,
in terms of a risk level to be achieved, is an integral part of the process.
Therefore, it is important to understand what is meant by risk assessment
before one can apply it to fire safety.
Tlie Risk Assessment Process
Risk is the potential for harm. It may more formally be defined as "the
potential for realization of unwanted, negative consequences of an event"
^The italic numbers in brackets refer to the list of references appended to this paper.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
5
[2]. The word "potential" refers to the probabilistic nature of risk. It is
negative because risk is only one part of a decision process. One only takes a
risk to obtain some benefit, that is, one gambles. We gamble all the time as
individuals deciding upon risks and benefits for ourselves on a voluntary
basis. Evaluating such gambles is an important part of risk assessment. In
these cases one takes a risk to achieve a benefit or is willing to undertake effort to reduce one's own risk. However, another major concern for risk
assessment is involuntary risk imposed inequitably upon those at risk (risk
agents), that is, the risk takers are not the beneficiaries of the risk. Although
such risks can never be entirely reduced, the concern is to keep them at a low
residual level (an acceptable level of risk) [3].
The term "risk assessment" is used here to describe the total process of
risk analysis, which embraces both the determination of levels of risk and the
social evaluation of risks. Risk determination consists of both identifying
risks and estimating the likelihood and magnitude of their occurrence. Risk
evaluation measures both risk acceptance^the acceptable levels of societal
risk—and risk aversion—the methods of avoiding risk—as alternatives to involuntarily imposed risks. The relationship among the various aspects of risk
assessment is illustrated in Fig. 1.
Risk identification and risk estimation are extremely important aspects of
risk assessment and involve scientific and technical determinations. These
often involve considerable uncertainty, and the interpretation of the impact
of uncertainties relies less on scientific consideration than on technical and
social value judgments.
In regard to risk identification, changes in the levels of risk are identifiable
in three circumstances: when a new risk is created, when the magnitude of an
existing risk changes, and when perception of an existing risk changes. All
three circumstances may occur simultaneously [4].
A new risk is a risk that did not exist previously, not a new identification of
an existing risk. Most new risks result from the activities of mankind, usually
as a result of a new technology, and often nature has no natural defense
against these risks [5]. For example, the sea transport of large volumes of liquefied natural gas (LNG) represents a new class of fire risks.
A new perception of existing risks may occur because the hazard was
previously unidentified, because the magnitude of the hazard changes suddenly, or because a slow change in the hazard's magnitude crosses some
threshold of societal concern. Alternatively, public perceptions of existing
risks may change whether or not the level of risk is changing in an objective
sense [6]. This change may occur when a more dominant risk is limited or
reduced, such as when the reduction of contagious diseases transfers public
concern to chronic diseases. Or public concern with risk may be stimulated
by the communications media or by the threat certain hazards pose to particular individuals or groups rather than statistically to the population at
large [7]. Individuals satisfied with the status quo may view the hazard as a
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
FIRE RISK ASSESSMENT
IS
1" I/)
v\
r! ..
Ill Q
i°
t OZ
^ X
111 ^ E ^ z - z ^ ^ i u
isis|2|2i
°
0-5*uj"=!ii<
C IL V
to"
U. u. U. lU
iS > =
Ul ^
im
a.'
2 ^ uS
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
7
threat to their present condition. For example, intentional arson for profit
has recently become a national concern; not only has the rate of and insurance value of such fires increased, but the media also have focused on the
problem at the national level.
One particular problem in risk identification concerns the source of
catastrophes. These large, disruptive events result from natural or man-made
hazards, as well as a combination of the two. Natural disasters generally
result from local and regional changes in energy balance patterns, which
result in storms, tornadoes, floods, earthquakes, drought, famine, and other
occurrences [8]. In many cases, technological approaches to mitigate the impact of natural events have reduced the number of events occurring but increased the magnitude of the consequences of those that do occur [9].
However, in some instances, the impact of large events, those that exceed
design limits, are accentuated as a result of "attractive exposure." For example, public auditoriums are designed with a certain level of fire protection
built into the structure and with additional procedures for prevention and
mitigation. As crowds gather and exceed capacity, both procedures for control and control systems can become ineffective, in which case a fire could be
catastrophic [10]. As a result, when an event exceeds its design capacity, the
magnitude of the accompanying disaster can be greater.
Man-made catastrophes occur as a result of technological systems which
meet at least one of three conditions. These conditions address systems that
involve the following factors:
1. Stored potential energy—for example, petroleum and natural gas storage, water in dams and tanks, heat in nuclear reactors, combustibles in
general.
2. Potential release of toxic materials—for example, chemicals such as
chlorine, cyanide, mercury, radioisotopes, pesticides, toxins, or DNA
derivatives.
3. Kinetic energy (inertial)—transportation, missiles, satellites, and debris
from space aircraft.
These, along with some of their properties, are illustrated as three large
circles in Fig. 2. Moreover, cases where these systems exist in combination
are particularly susceptible to catastrophes. Transportation of fuels is one of
the major sources of catastrophic accidents. Transportation of hazardous
chemicals and radioisotopes provides unusual pathways for catastrophic exposure. Many perceive the coupling of stored energy systems with toxic materials, such as those found in nuclear reactors or stored nerve gas, explosive
shells, and rockets, as a major potential problem. Although the historical
record for catastrophes of this type is nil, the concern for potential catastrophes is high [11]. Nuclear war with ballistic missiles is an example in which
all three systems are combined.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
8
FIRE RISK ASSESSMENT
Man Made Catastrophles
With Delayed Consequences
Man Made Natural Catastrophles
With Immediate Consequences
FIG. 2—Systems prone to catastrophic events.
The process of risk estimation has five steps: identifying the cause of risk;
measuring its effects; determining the risk exposure; defining the consequences of exposure; and valuing the consequences of exposure. These five
steps are illustrated in Fig. 3.
The first step in risk estimation is to identify causative events, or events
that create a probability of risk occurrence. Each causative event may lead to
several possible outcomes. In the second step, these outcomes are defined,
and their relative probability is determined. The causative event and outcome do not, of themselves, constitute risk until the exposure of humans, institutions, and the natural environment to the outcomes is considered. The
probabilistic relationship between events and their outcomes can be measured through scientific experiments, the statistical design of experiments,
and hypothesis testing.
In fire prevention, the causative event usually involves an initiator of a fire
and its outcome—the many possibilities that might result should a fire
start—in a certain situation.
The third step in risk estimation defines exposure pathways, the means by
which risks are transmitted. Risks cannot occur unless there is some exposure pathway to people. We live with fire and energy all the time. We use
it, control it, and make it work for us. The concern is with uncontrolled pathways whereby people and property are exposed.
The fourth step is to define the possible consequences of risk exposure and
to determine, for each risk, the probability that consequences will occur. The
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
II
•4!
"s^
0
1
0
It.
t
1t11
/
o
/
ii
!
ii
' suoiteiati >isiti -
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
9
10
FIRE RISK ASSESSMENT
final step in risk estimation considers the value placed by affected individuals
on the consequences of risk exposure. Some individuals, for example, may be
unconcerned about consuming trace amounts of pesticides in food because
they consider the effects remote; others may be sufficiently concerned that
they eat only organically grown foods.
The probability of risk occurrence and the value placed on risk consequences by affected individuals determines the public's response to risks.
Thus, the process of risk estimation requires two basic determinations: a
consequence probability determination and a consequence value determination. The steps comprising each determination are bracketed on the right
side of Fig. 3; the determinations overlap at Step 4, the consequences.
Fire prevention can be considered in this light. First, the causes of fires can
be examined (as they already have been) and their probabilities (or attractiveness for purposefully ignited fires) can be estimated from historic records
to the extent possible. Various outcomes can be considered, and the probabilities of such outcomes can be controlled by early warning containment systems, extinguishing systems, and response systems. Exposure involves the
possibility of exposure to people and property. Warning, protective, and
escape systems can reduce exposure to people while protective, containment,
and response systems can reduce exposure to property. Consequence mitigation involves a variety of options, including rescue and medical systems, firefighting methods and equipment, insurance for spreading risks, and loans
and loan guarantees for recovery from disasters.
Consequence values are subjective and can be changed. Fire safety and
prevention programs aimed at the public and industry can sometimes increase awareness of safety. Both individual action for increased safety (such
as in the home) and community action can result.
Risk aversion involves the application of controls to avert risk. However,
since risks cannot be eliminated, acceptable levels of risk must be established
either formally or by tacit acceptance of specific status quo situations. The
risk aversion process is well established in the fire safety area on a subsystem
basis. Considerable progress and technical knowledge are available for
preventing and mitigating fires. Building design for fire resistance, automatic early warning and response systems, new materials and better equipment for prevention and fighting fires, improved fire codes, and better trained
and better equipped fire-fighters are some examples. The problem is that all
of these approaches require resources. The major question yet to be addressed involves where resources can be most effectively used and at what
level of resources risk aversion should be supported. Should more be put into
prevention or into response systems? Who should pay for these? When are
mandated controls too expensive?
In many respects, the problems within subsystems involve risk trade-offs.
The homeowner must decide how much he wants to spend for fire ex-
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
11
tinguishers, smoke detectors, escape devices, fire insurance, and so on. The
commercial or industrial building owner must make the same kinds of decisions, but on a different scale and with different motivation. In the first case,
the homeowner is making a purely voluntary decision—his money for his own
safety and protection. In the second case, resources must be used nonproductively to achieve some degree of required protection. The decision involves
taking the risk of underinvestment, just meeting requirements, or exceeding
them. The trade-off involves not only protection but possible future liability.
Should more be spent to protect property or life?
For methods which are too expensive for individual operation, reliance on
municipal, state, and federal organizations becomes necessary. Such organizations can provide funds for support and ensure equitable access to
benefits. Since most subsystems for fire prevention are localized, state and
federal subsystems are minimal, except for a few states such as Ohio. As a
result, each community has its own system of fire codes, fire prevention and
fighting, and level of resource balancing. The result is a confusion of codes
and applications. Federal or state codes to standardize in this area would
have to deal with the differing needs of communities, such as rural versus urban, small versus large, and so on.
Acceptable levels of risk are both targets to be achieved and visible
measures of residual risk. The existence of specific levels of acceptable risk
can often be an embarrassment to public officials, those who are responsible
for public safety and those who must balance tax levels against the many
sources seeking to use these funds. When stated levels for acceptable risk are
not being achieved, for whatever reason, many public officials would like to
keep such information in implicit, rather than explicit, form. The problem
translates into standards and their application and enforcement.
What Is an Acceptable Risk?
Unquestionably, some risks are acceptable. Some conditions that support
this contention are evident:
1. Threshold condition—a risk is perceived to be so small that it can be ignored.
2. Status quo condition—a risk is uncontrollable or unavoidable without
major disruption in life-style.
3. Regulatory condition—a credible organization with responsibility for
health and safety has, through due process, established an acceptable risk
level.
4. De facto condition—a historic level of risk continues to be an acceptable
one.
5. Voluntary balance condition—a risk is deemed worth the benefits by a
risk taker.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
12
FIRE RISK ASSESSMENT
There have been many definitions of acceptable risk, unacceptable risk,
unreasonable risk, and so on. Most of these definitions involve the methods
of establishing references for comparing risks, such as revealed, implied, and
expressed preferences [9], or evaluation of a spectrum of possibilities for different situations [/].
To circumvent all the arguments involved in establishing a precise definition, a working or operational definition may be useful: a risk is acceptable
when those affected generally are not. or are no longer, apprehensive about
it.
As a general statement, this definition does not include all those affected.
Nor does it matter by what process the level of acceptability is achieved. It
does, however, include the "regulators" and "experts," since they must be
satisfied that the risks are low enough. It relates to society's propensity for
anxiety aversion as well as its propensity for risk aversion. The statement addresses voluntary and involuntary risks. The broadness of the definition
reflects that there is no single, universal method of arriving at an acceptable
level of risk.
Two variables exist: (1) how much risk is acceptable for an entirely new
undertaking, that is, a go-no go decision; and, (2) which alternative should
be selected to reduce risk to a residual level. This last variable is most often
the consideration in fire safety. A range of alternative controls for reducing
risk can be established for any existing or prospective case. Moreover, these
controls can be ordered in terms of their cost-effectiveness of risk reduction
[12], as shown in Fig. 4 (which assumes that all cost-ineffective solutions, lying above the curve, are ignored). The question that remains is when does
one stop spending to reduce risk [13]? By putting a dollar value on life [14],
by best practical technology, best available technology, and so on, as shown
in Fig. 5? These are societal decisions, which appropriately can be set only by
society and its designated representatives [15]. At the present time, acceptable levels of risks have not been set formally other than in an imputed manner through the establishment of fire codes. Acceptable levels of risk may be
established tacitly by the voluntary standard process but may be abrogated
by the formal consideration of regulatory bodies.
Standards in the System Process
There are a variety of standards necessary in the fire safety control system.
One kind of standard involves specification of an acceptable level of risk,
either explicitly or tacitly. This involves a social decision as to what level of
fire protection is needed and how and how much to pay for it. A second kind
of standard describes the minimum level certain kinds of subsystems must
achieve to be considered as meeting a definitional requirement. Different requirement levels can be specified, such as different levels of flame retardant
capability for materials. The American Society for Testing and Materials
(ASTM) standards are of this latter kind. They are "intended for analysis
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
13
ORDERED BY VALUE OF AR,/ACi
5i = Risk reduction action.
Affi = Changes in risic for Si.
ACi = ctiange in cost for S,.
AC,
^' SIMOOTHEO COST-EFFECTIVENESS CURVE
COST OF RISK REDUCTION
FIG. 4—The cost-effectiveness of the risk-reduction ordered relationship for discrete actions.
S; = risk reduction action: R; = change in risk for S;,- C; = change in cost for S,.
and assessment of the fire perfonnance of materials, products, and systems
within their relevant environment" [76]. As such, they represent both a
classification system and a specification of the levels to be achieved.
Standards of the first kind are called "acceptability standards" while the
second kind are called "achievement standards" in order not to confuse them
with performance and design standards, which have other meanings in different contexts. Acceptability standards must be set by individuals at risk or
by public officials responsible for public safety. These can be set in terms of
performance, that is, levels of risk to be achieved, or in terms of designs
which specify approaches to be used to meet performance criteria. Models of
risk behavior for different designs in different conditions are used to
translate design standards into performance standards and vice versa. These
models then become critical in determining levels of risk. Who is responsible
for evaluating and standardizing such models? At present, there seems to be
no particular coverage in this area.
Achievement standards are the responsibility of experts for materials, for
procedures, for training, and a host of other subsystems. Responsible
technical bodies can specify such standards in terms of levels of protection afforded by particular designs; selection of designs to be used to achieve a level
of protection at some cost is part of the acceptance standard process. Perfor&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
14
FIRE RISK ASSESSMENT
NO RISK REDUCTION
AS LOW AS PRACTICABLE
RANGE
BEST PRACTICABLE TECHNOLOGY
I— MEASURED ZERO
-ABSOLUTE ZERO
COST OF RISK REDUCTION
FIG. 5—Some criteria for llie acceptance levels of the cost-effectiveness of risk reduction.
mance standards of the achievement type specify levels of achievement in
protection to be obtained independent of design. Design standards specify
particular designs that can reach the protection level, and, again, models are
needed for translation. However, in this area, considerable attention is given
to such models. Certainly, more can be done, but organizations such as
ASTM provide responsible focal points.
Summary
There is little that is novel in looking at fire safety from a systems point of
view. However, the very process of structuring at a systems level can help in
understanding the many component subsystems, their interaction, and their
weak and strong points. A view toward systematic use of resources in an effective manner must also consider the scope of responsibility and motivation
of the many "players" in the system. Recognition of the different types of
standards involved can help provide for a better understanding of the differing objectives and roles involved in each type.
The major question left unanswered is whether a substantial analysis at the
systems level would be useful and warranted. If one expects final solutions.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROWE ON ASSESSING RISK OF FIRES SYSTEMICALLY
15
the answer is "no." If one is looking for better insight to focus resources more
effectively, locate weak points for corrective action, and make the "players"
involved aware of their responsibilities, limits of action, and ability to respond
to the system, then it might be of considerable merit. If so, the analysis ought to
proceed in stages to prevent the obtaining of too much detail too soon.
References
| / | Lowrance, W. W.. Of Acceptable Risk, Kaufman, Los Angeles, 1976.
b i Rowe, W. D., All Anatomy of Risk. Wiley. New York, 1977.
|J1 Okrent, D. and Whipple, H., "An Approach to Societal Risk Acceptance Criteria and Risk
Management," UCLA-ENG-7746, University of California. Los Angeles, June 1977,
[4] Rowe, W. D.. "Government Regulation of Societal Risks." The George Washington Law
Review. Vol, 45. No, 5, Aug. 1977. pp, 944-968.
[.5] Krutilla. J. V, and Fisher. A, C The Economics oj Natural Environments. Johns Hopkins
Press, Baltimore, Md,, 1975.
|6| Kiesler. C. A.. Collins, B. E., and Miller, N,, Attitude Change. Wiley. New York, 1969.
[7] Bern, D., Beliefs. Attitudes and Human Affairs, Brooks/Cole, Belmont, Calif.. 1970.
\H\ Cochrane. H, C "Natural Hazards and Their Distributive Effects," National Technical
Information Service, PB-262-021, U.S. Department of Commerce, Springfield, Va., 1975.
|91 Kates. R. W.. Risk Assessment of Environmental Hazard. Scope fi. Wiley. New York.
1978.
[10] Kupperman. R. H., Wilcox. R, A., and Smith. H, A., Science. Vol. 187, 7 Feb, 1945.
| / / | Dworkin, Judith. "Global Trends in Natural Disasters, 1947-1973," Working Paper No.
26, Natural Hazard Research, University of Colorado, Boulder, Colo., 1974.
\I2] Rowe, W. D., "Assessing Risk to Society," presented at the Symposium on Risk Assessment and Hazard Control, American Chemical Society, New Orleans, La., March 1977.
[13] Otway, H. J. and Fishbein. M., "Public Attitudes and Decisionmaking." International Institute of Applied System Analysis Research Memorandum, RM-77-54, Loxenburg,
Austria, 1977,
[14] Otway. H. J., Pahner, P. D.. and Linnerooth, J,, "Social Values in Risk Acceptance," International Institute of Applied System Analysis Research Memorandum, RM-75-54, Loxenburg, Austria, 1975.
]I5] Otway, H. J., "Risk Assessment and Societal Choices," International Institute of Applied
System Analysis Research Memorandum, RM-75-2, Loxenburg, Austria, 1975.
1/6] ASTM Directory. American Society for Testing and Materials, Philadelphia, 1979,
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
H. J. Roux^
A Discussion of Fire Risk
Assessment
REFERENCE: Roux, H. J., "A Discussion of Hue Risi< Assessment," Fire Risk Assessment, ASTM STP 762, G. T. Castino and T. Z. Harmathy, Eds., American Society for
Testing and Materials, 1982, pp. 16-27.
ABSTRACT: Fire standards have been characterized in the ASTM Policy on Fire Standards (1979) as either fire risk assessment standards or fire test standards. Although the
need for fire risk assessment standards is evident, there is a lack of understanding of what
a fire risk assessment standard is and how it is obtained. The role of ASTM Subcommittee E05.91 on Planning and Review, a subcommittee of ASTM Committee E-5 on Fire
Standards, and its success in answering these questions is described. Related subjects
such as definitions of risk, hazard, and safety, the effect of proximity, and determination
of the potential for harm are discussed.
KEY WORDS: fire risk, fire risk assessment, fire hazard, fire safety
In the musical The King and /, by Richard Rodgers and Oscar Hammerstein, there is a song titled "Is a Puzzlement." Fire risk assessment is a
puzzlement!
It has been said that the impossible takes just a little bit longer. It has also
been said that the best wine comes from vines that have struggled to grow
and produce the grapes. I have no argument with the validity of these sayings, nor with their application to fire risk assessment. Indeed, it has been
many years since the ASTM Board of Directors originally issued the ASTM
Policy on Fire Standards [7],^ in which the idea of fire risk assessment (at that
time known as fire hazard assessment) was embodied.
I have often wondered whether ASTM Subcommittee EOS.91 on Planning
and Review of ASTM Committee E-5 on Fire Standards would ever develop
an understanding of fire risk assessment; but I do believe it has now done so
to a measurable degree. The efforts of all of the members of ASTM Subcom'Coordinating manager, Product Fire Performance, Armstrong World Industries, Inc., Lancaster, Pa. 17604.
^The italic numbers in brackets refer to the list of references appended to this paper.
16
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
17
mittee E05.91 in this regard should be applauded, particularly those of Dr.
Clay Hathaway and Dr. Tibor Harmathy.
Definition of Risk
To forgo a recitation of the history of this development, risk is now defined
by ASTM Subcommittee E05.91 as the combination, probably the product,
of the expected frequency of the event, the expected degree of exposure, and
the potential for harm.
Risk = expected frequency of the event X expected degree of exposure
X potential for harm
In the case of fire risk, the event is the fire itself; the exposure is of the people, property, or operation exposed to the fire; and the potential for harm is
the consequences of the applicable products of the fire—for example, heat,
flame, smoke, and toxic gases—to those exposed. It is probably correct to assume that the elements of risk are multiplied by each other, as shown in the
previous equation, for if any element is equivalent to zero, the risk is then
zero.
The definition of risk has been rather simply stated; therefore, a few words
about a very real application of this definition are in order. The three elements of risk—the expected frequency of the event, the expected degree of
exposure, and the potential for harm—were all, on reflection after the fact,
included in the 1979 nuclear power plant accident at Three Mile Island
(TMI), near Harrisburg, Pa., as evidenced by the subjective response of the
people, including me, who live near the plant.
It is my perception that before the accident and, interestingly, after it, the
risk was presented to these people principally in terms of the expected frequency of the event—not so much as if it were the sole element of risk, but
rather with an emphasis that precluded recognition of the other elements.
Potential for harm was mentioned, much as the Rasmussen report [2] mentions both the likelihood and the consequence of an event in its definition of
risk. True, the expected frequency of the event was declared to be extremely
low. However, the occurrence of the TMI accident itself suggested to some
the expected degree of exposure ("we live next door, downwind") and of the
potential for harm ("a devastated area for 40 years"). And these words were
expressed loudly enough that, although quieted by the low expected frequency of the event, these people now have an innate understanding of a level
of risk higher than that previously understood. In other words, risk is not just
the element of the expected frequency of the event; it also includes the
elements of the expected degree of exposure and the potential for harm!
For those who wish to apply their scholastic abilities in this area, I purposely used the potential for harm of an operation—that is, "a devastated
area for 40 years"—as my example, rather than the potential for harm to
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
18
FIRE RISK ASSESSMENT
people or property, simply to give a broader frame of reference to the potential for harm and, consequently, the risk.
In the continuing response of people to the TMI accident, which now includes a concern about the procedure of repair of the reactor, specifically the
venting of radioactive gases from the containment building, there is the further indication that the three elements of risk are not necessarily equal in importance. The equation that was given earlier for risk may consequently need
to be modified. However, further study of this idea by using the TMI accident as an example is hampered by the question of proximity, both spatial
and temporal, as it applies to risk. For many people, the understanding of
risk is a feeling of being included in the risk. The expected frequency of the
event and the expected degree of exposure are already total in such cases,
and, therefore, their thoughts are focused solely on the potential for harm.
This is clearly true for those people near TMI who have been told that
venting the radioactive gases will probably occur soon. The apparent emphasis, therefore, on the potential for harm as venting is being considered, is
not necessarily an indication of its importance in relation to the three
elements of risk, because of the temporal and spatial proximities in this
situation. One should be aware that a similar error of perception on this
point could well occur in other situations, such as a fire.
It is my thought that we should not consider risk from the view of those
close to the event. Fire is usually not proximate, in the sense that it is certainly not planned or expected. Rather, there is a finite expected frequency of
the event that is less than total, and there is also a finite expected degree of
exposure that is less than total.
I have found, though, at least one reference that applies to this question of
relative importance. The specific statement was offered as a point of common
law in the report on the well-known Beverly Hills Supper Club fire in Kentucky. This statement is found in Section G of the report [3] by the state
police commissioner: "The gravity of the possible harm and not the probability of the occurrence must be given the most weight in considering
whether someone has been negligent, although both factors must be balanced against the burden of adequate precautions." Although the latter part
of this statement bears more on the subject of cost-benefit analysis, the first
part certainly touches on the idea of relative importance.
The same consideration of relative importance is also implied in a news article [4] that appeared in the 27 March 1980 issue of Engineering News
Record. The article reported that the Council of Environmental Quality had
sharply criticized the Nuclear Regulatory Commission's failure to analyze
potential reactor accidents, especially severe accidents, in its environmental
impact statements. The council specifically noted that the TMI environmental impact statement exemplifies the deficiencies of a narrow approach to
major accidents. One excerpt from the TMI statement, offered as evidence of
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
19
this deficiency, stated that while the consequences of a Class 9 accident could
be severe, its probability is so small that the environmental risk is extremely
low. Apparently, the council objected to this conclusion because the potential
for harm was considered to be equal to or of less importance than the expected frequency of the event.
Alternative Definition of Risk
So far, I have offered a definition of risk (the combination of three
elements), a warning about the confusing aspects of proximity in assessing
risk, and the suggestion that the potential for harm may be of greater importance than the other two elements. I have probably been remiss, though, in
not offering further evidence of the acceptability of this definition of risk.
The British Standards Institution, for one, has been reported-' as having
stated that "the integration of the results from testing procedures for flammability, flame spread, smoke obscuration, and heat contribution, together
with the degree of usage and probability of fire from past experience, can be
used to produce some form of (risk) index." In this statement, the "results
from testing procedures" are the equivalent of our potential for harm, "the
degree of usage" is similar to our expected degree of exposure, and "the
probability of fire" is like our expected frequency of the event.
Much earlier, in the beginning of the Consumer Product Safety Commission (1974), Locke [5] had written of risk as being equal to the probability of
injury per unit time of exposure times the severity associated with the injury.
Here, "severity" is similar to our potential for harm, and the "probability of
injury per unit time of exposure" combines our expected frequency of the
event and expected degree of exposure.
In 1977 at the National Fire Protection Association's (NFPA) annual
meeting in Washington, D.C., W. T. Fine of the U.S. Navy Naval Surface
Weapons Center had offered the concept of a "risk score" in evaluating and
controlling safety [6]. This "score" was equal to the product of consequences, exposure, and probability. "Consequences" in this case is the same
as our potential for harm, "exposure" like our expected degree of exposure,
and "probability" like our expected frequency of the event.
One fire protection consultant"* has defined risk as the probability of an
event resulting in injury, illness, or damage. The "probability of event" is
echoed in our definition of risk, while the word "resulting" is equal to our expected degree of exposure. The "injury, illness, or damage" is, in essence,
the result of our measure of the potential for harm.
The International Standards Organization (ISO), in their Technical
^Loosemore, G. W., Armstrong Europe Services, Ltd., private communication, 25 Feb. 1980.
''Cohn, Bert, Gage-Babcock & Associates, personal communication, 4 Aug. 1977.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
20
FIRE RISK ASSESSMENT
Report No. 6585 [7\ on "Fire Hazard and the Design and Use of Fire Tests,"
also refers to three elements of hazard: "risk of initiation" (which matches
our expected frequency of the event), "consequential hazard" (which
matches our potential for harm), and "degree of exposure of property or people" (which matches our expected degree of exposure). Unfortunately, this
document does use the term "hazard" instead of "risk," which is our preferred term.
Much the same error in terminology, if I may call it this, occurs in the
document "Benefits of Environmental, Health and Safety Regulation,"
which was recently (25 March 1980) issued by the U.S. Senate Committtee on
Governmental Affairs [8]. In this document the three elements of risk are
identified as an "activity to create the presence of hazard" (which is our expected frequency of the event), "exposure of people or natural systems in the
surroundings" (which is our expected degree of exposure), and "damage"
(which is our potential for harm).
In concluding this section, I repeat that the ASTM Policy on Fire Standards defines fire risk as the probability that a fire will occur and the potential for harm to life and damage to property resulting from its occurrence.
This exercise in citing other sources for a definition of risk has permitted a
presentation of alternative terms for the three elements of risk, which may be
more acceptable to some people than our own terms. There are many more
that could be given—for example, "occurrence" in place of our combination
of expected frequency of the event and expected degree of exposure. It was
not my intention to use a flood of semantics to obtain order out of chaos,
although I do believe in the power of dimensional analysis to derive relationships—for example, the earlier observation that the elements of risk are
probably multiplied by each other.
Definition of Hazard
Turning now to another subject, ASTM Subcommittee E05.91 has further
identified risk as the qualitative measure for which we are seeking a quantitative dimension. In other words, risk is the scalar quantity that extends
from zero risk to total risk. At some level of risk, which can vary for a variety
of reasons, it is expected that the authority having jurisdiction, or society's
representative, will define the area of risk above this level as hazardous and
will accept the area of risk below this level as safe. This, then, becomes our
definition of "hazard," and consequently of "safe." It is of extreme importance to note that "safe" does not refer solely to a zero level of risk. Safe can
include a very measurable amount of risk.
The setting of this level of risk, which differentiates between hazardous
and safe, is, I believe, truly a function of the authority having jurisdiction or
a representative of society, if not the person or people involved. I do not
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
21
believe it is the responsibility of an organization such as ASTM, for example,
although it has been suggested that ASTM should develop the tools to
measure risk, and possibly should even offer guidance on how the level between hazardous and safe should be determined. As a matter of fact, there
are probably many among us who would wish to offer this guidance if only to
preclude "bureaucratic fiat" or "prejudiced opinion."
Interestingly, and using the TMI accident again as a very real application
of this concept, in more than one survey conducted after the TMI accident a
majority of people have expressed a favorable attitude toward nuclear power,
provided there are appropriate controls. To me, this indicates an understanding by the people of the level of risk that defines hazardous and safe,
and a demand that the risk remain below the hazardous level.
If you are beginning to believe that I feel people do understand risk and
hazard, even though they may not be able to define these terms in so many
words, and that I have confidence in the correctness of their understanding,
you are correct.
In the Massachusetts Department of Public Health regulations [9], there is
a definition of hazardous substance that complements the concept which I
have just expressed. A hazardous substance is defined here as one that is
either extremely flammable or just flammable, for which a label of "danger"
or "warning," respectively, is required. A nonhazardous substance is one
that is combustible but for which there is no requirement of a label. The
quantitative dimension of risk, in this case, is the open cup flash
point—greater than 27°C (80°F) for the nonhazardous substance, and below
27°C (80°F) for the hazardous substance.
Further guidance toward defining hazardous versus safe is found in the
work of Starr [10], although the definition in this case is based on the concept
of voluntary and involuntary risk (Fig. 1). In this approach, the area above
the appropriate curve is deemed to be unacceptable, and consequently
hazardous. The reverse is true for what is deemed safe. I have plotted the expected value for death due to fire in the United States (this is for the entire
population), which can be seen to be acceptable.
Other methods of differentiating safe from hazardous are the comparative
method, in which the level of risk for one type of situation is compared with
that for another type of situation; the expectancy method, in which the level
of risk is compared with the expected length of life of a person in his home (or
the working time of a person in his office); and the natural method, in which
the level of risk is compared with the risk of death due to disease.
Dimensions of Rislc
The final area of concern is the quantitative dimension of risk. This is truly
a difficult problem to solve, but it is one that must be solved if we are to
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
22
FIRE RISK ASSESSMENT
o
o
1
M
t-
ro
o
Q
•a
t>
_>
o
>
c
o
(A
tV
Q.
\
u
c
t>
m
"(5
3
C
C
O)
n
>
<
( d j n s o d x 3 ^o j n o H - u o s J S d / s s D J i e ^ e j )
I
o
i^
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
23
understand risk and, subsequently, fire risk assessment. Lord Kelvin [11] expressed the problem so well:
I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot
measure it, when you cannot express it in numbers, your knowledge is of a
meagre and unsatisfactory kind; it may be the beginning of knowledge, but you
have scarcely, in your thoughts, advanced to the stage of Science, whatever the
matter may be.
It is in the work of the U.S. Energy Research and Development Administration (ERDA), specifically their Risk Management Guide, published
in June 1977 [12], that I found possible dimensions for risk and its three
elements. The ERDA's definition of risk is the probability of an accident
multiplied by the consequence of the accident. This fits our definition of risk,
"probability of an accident" fits the expected frequency of the event, and
"consequence" combines the expected degree of exposure and potential for
harm.
According to this definition, then, risk is equal to the loss per unit time or
unit activity; the expected frequency of the event is equal to the event frequency per unit time or unit activity; the expected degree of exposure is equal
to the exposure per event; and the potential for harm is equal to the loss or
cost per exposure (Table 1). The key dimension is the last one, potential for
harm.
As early as 1973, the British Fire Research Station attempted to quantify
risk (see Ref 13). The use of a similar statistic for the United States has been
contemplated by the NFPA Systems Concepts Committee for their Firesafety
Concepts (Decision) Tree [14]. The particular dimension for fire risk in this
reference is the number of casualties (fatal or nonfatal) per worker hour of
exposure in a given class of occupancy. This takes into account the number
of casualties, the number of people at risk, and the length of time they are at
risk. It corresponds with our proposed dimension for risk of loss per unit activity. This, by the way, is not unlike the more often seen statistic in the
United States of the number of deaths per so many million passenger miles
for commercial aviation—again, the loss per unit of activity.
TABLE 1—Dimensions of risk.
Risk
loss
=
\
unit time /
expected frequency
ot the event
^ expected degree
ot exposure
/ event frequency \
\
unit time
-)/
^
potential for harm
/ exposure \
/
loss
\
\ exposure
X
event
/
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
24
FIRE RISK ASSESSMENT
Potential for Harm
The reason I have focused on the dimension of potential for harm is to consider the complete development of fire risk assessment. ASTM Subcommittee EOS.91 has a task group at work trying to develop a recommended practice for a fire risk assessment standard. However, the work has not progressed so far that the following question could not be asked. Should ASTM
Committee E-5 limit its efforts solely to developing a value for the potential
for harm? Is not the task of developing values for the expected frequency of
the event and the expected degree of exposure more appropriately placed
with some other body, within or outside of ASTM? I do not have a specific
answer to these questions to recommend at this time. I would suspect,
though, that other ASTM Committees are interested in the general subject of
risk assessment, for whatever purpose. Would we not be better served, in
ASTM Committee E-5, as well as in the rest of ASTM, by the creation of a
new committee with the scope to develop and provide guidance in risk assessment methods, much as is done for statistical analysis?
If we do accept the charge that our role is to develop a value for the potential for harm, there are still many considerations before effective guidance
can be given. So far, we have stated that the potential for harm is the result of
the applicable products of the fire—for example, heat, flame, smoke, and
toxic gases. The dimension for measuring the potential for harm that has
been offered is the amount of loss or cost per exposure. How do we get from
the products of fire to the loss or cost?
One suggestion, which relates to people, is to accumulate the products of
the fire within the space in which the object of interest is located and count
one loss for that exposure if the accumulation of the products exceeds in any
respect the tenability limits for the space. The actual presence of a person in
that space is then considered under the expected degree of exposure, which is
an entirely independent consideration. Note that a time element might be involved in both of these considerations.
Tenability limits could be developed on the basis of resulting injury or
death. The space could be defined in relation to the object of interest—for
example, a sofa in a living room or a floor in the patient room of a nursing
home. I have said that the expected degree of exposure is an independent
consideration. What of the expected frequency of the event? In another activity in which I am involved, the Building Firesafety Model [15] and the
Firesafety Concepts (Decision) Tree [14] of the NFPA Systems Concepts
Committee, it has been proposed that the goal and the measure of success of
fire safety is to keep the exposed (people) from ever being in the same space
at the same time as the fire and its products (heat, flame, smoke, and toxic
gases). This proposal can be extended to our risk definition, if we consider
that the combination of the expected frequency of the event and the potential
for harm are equivalent to the fire and its products and that the expected
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
25
degree of exposure is the same as the exposed. The expected degree of exposure is, therefore, independent of the other two elements, while the expected frequency of the event and potential for harm are interdependent.
This is fine, for it permits us to use the fire scenario as the connecting link
between the latter two elements, particularly to design the fire to which the
object of interest will be subjected, and all elements are tied together for a
given occupancy/given environment/given space. The task group on
upholstered furniture of ASTM Subcommittee EOS. 15 on Building Content
is proceeding in this very direction of determining risk as a function of occupancy. This idea also allows the use of available fire statistics to estabUsh
both the expected frequency of the event and the appropriate scenario.
Various concepts have been offered on how we obtain the appropriate
products of the fire, which are then used to establish the potential for harm.
In an early paper in 1976 [16], I had offered two distinct techniques. The first
technique was simply to conduct a room fire test either with or both with and
without the object of interest involved. The actual result or difference in
results, respectively, in terms of the products of the fire, is a measure of the
potential for harm for the object of interest. There is, of course, an expense
and encumbrance (complexity, limited testing facilities) involved with this
form of testing. Alternatively, one might hope to obtain the potential for
harm of the object of interest directly, albeit by induction. This proposed
technique would involve the use of an integration equation.
In addition to the suggestion I made in the 1976 paper, Herpol of Belgium
had offered the same approach earlier [17\. Harmathy has most recently
taken a similar approach [18]. In all of these references, the suggested procedure would make use of an integration equation to bring together the
results from one or more fire-performance test methods, plus other
parameters, in an appropriate order that generates (induces) the potential
for harm of the object of interest. The general form of this equation for
potential harm (PH) is given in Eq 1.
PH = FP," X FP2'' X FPn" X OPi» X OP2'' X O P /
(1)
where FP is the fire performance parameter, and OP indicates other
parameters. Interestingly, this type of equation can be traced back to earlier
work done by Harrington in 1965 on the desirability function [19].
The tools of dimensional analysis and empirical integration may well serve
us initially in the development of appropriate integration equations, at least
until mathematical fire modeling has progressed further.
Siunmaiy
In summary, I have offered other, matching, definitions of risk; identification of risk as a scalar quantity; a definition of "hazardous" and of "safe" as
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
26
FIRE RISK ASSESSMENT
they relate to a particular level of risk; a quantitative dimension for risk; and
the suggestion of directing ASTM Committee E-5 solely toward developing a
complete understanding of potential for harm either by full-scale testing, by
integration equations, or by some other technique. This suggestion should be
the focus of the future for ASTM Committee E-5.
In making this suggestion, I am also stating that this presentation is not a
definitive solution to the problem of fire risk assessment. It is only a start.
To demonstrate that it is only a start, there is the question, actually more
of an item of concern, of whether the control of risk is offered after the risk
has been created or before it occurs by controlling the elements of risk. Actually, the answer is a matter of perspective.
For example, the use of smoke detectors to warn people of a fire and its
potential for harm will reduce the expected degree of exposure because people will react to the alarm by evacuation. Similarly, the use of an automatic
extinguishing system will reduce the potential for harm. Therefore, rather
than consider these fire protection features as controlling the risk after the
fact, it is more helpful to consider them as controlling the elements of risk. In
this way, one is able to see more clearly how the elements of risk can be
manipulated in a given situation to minimize the risk. A third question, offered here only as an exercise in thinking about the details of risk, is this:
should ease of ignition be considered part of the potential for harm or part of
the expected frequency of the event? I do not have an answer.
References
[/] ASTM Policy on Fire Standards, Revised, American Society for Testing and Materials,
Philadelphia, 16 Jan. 1979.
[2] Rasmussen, Norman C , A Method for Risk Analysis of Nuclear Reactor Accidents,
Massachusetts Institute of Technology, Boston, Jan. 1978.
[3] "Investigative Report to the Governor—Beverly Hills Supper Club Fire, May 28, 1977,"
Commonwealth of Kentucky, 16 Sept. 1977.
[4] Engineering News Record, 27 March 1980, p. 19.
[5] Locke, John W., Operations Research. Vol. 22, No. 6, Nov.-Dec. 1974.
[6] Fine, William T., NFPA News Report. June 1977.
[7] "Fire Hazard and the Design and Use of Fire Tests," Technical Report 6585, International
Standards Organization, Geneva, Switzerland, 1979.
[8] "Benefits of Environmental, Health and Safety Regulation," U.S. Senate Committee on
Governmental Affairs, Washington, D.C., 25 March 1980.
[9] Regulations Concerning Hazardous Substances, 105 CMR 650, Department of Public
Health, Commonwealth of Massachusetts, 14 Nov. 1979.
[10] Starr, Chauncey, Science, Vol. 165, 19 Sept. 1969.
[//] Kelvin. W. T., Popular Lectures and Addresses. 1891-1894.
[12] Risk Management Guide. ERDA 76-45/11, SSDC-11, U.S. Energy Research and Development Administration, Washington, D.C., June 1977.
[13] "The Estimated Fire Risk of Various Occupancies," Fire Research Note 989, Fire
Research Station, Department of the Environment, Borehamwood, England, 1973.
[14] Firesafety Concepts Tree. National Fire Protection Association, Quincy, Mass., 1980.
[15\ "Firesafety Systems Analysis for Residential Occupancies," National Fire Proection Association, Quincy, Mass., March 1977.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
ROUX ON DISCUSSION OF FIRE RISK ASSESSMENT
27
[16] Roux, H. J. in Fire Standards and Safety. ASTM STP 614. American Society for Testing
and Materials, Philadelphia, 1976, pp. 194-205.
[17] Herpol, G. A., Ignition. Heat Release, and Noncombustibility of Materials. ASTM STP
502. American Society for Testing and Materials, Philadelphia, 1972, pp. 99-111.
[18] Harmathy, T. Z., Fire and Materials, in press.
[19] Harrington, E. C , Jr., Industrial Quality Control. April 1965, pp. 494-498.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
B. M. Cohn^
Formulating Acceptable Levels
of Fire Risk
REFERENCE: Cohn, B. M., "Formulating Acceptable LeyeU of Fire Risk," Fire Risk
Assessment. ASTM STP 762. G. T. Castino and T. Z. Harmathy, Eds., American Society for Testing and Materials, 1982, pp. 28-37.
ABSTRACT: Building codes and standards traditionally have set levels of performance
based on the experienced judgment of the code writer. Occasionally, a certain level of
safety is envisioned, but the objective tends to be expressed in imprecise terms, such as
the "minimum" or "reasonable" level of safety. There is a need to set goals for fire safety
in terms which can be measured so that they can be specified in building codes and standards. With such quantification, systematic fire safety analyses will be more meaningful.
A probabilistic approach can be used to set fire safety goals. Sufficient statistics are
available today to set realistic goals, and the state of the art in fire safety systems analysis
is sufficiently advanced to develop the methodology needed to calculate the failure probabilities for specific events. Such a methodology is described. It assigns probabilities to
events associated with fire development in a building and allows these probabilities for a
specific set of conditions to be related to the risk being evaluated (loss of life, property
damage, business interruption, and so on). Statistics of industry experience help to relate
the relative safety levels to the actual loss forecasts. Similar statistical analysis can be
used to establish goals for fire safety in building codes, regulations, and standards.
KEY WORDS: fire risk assessment, risk levels, building codes, fire development, fire
spread, probabilistic methodology
Once upon a time, it was popular to call a steel-and-concrete building
"fireproof." That term finally was considered to be sufficiently misleading
that the term "fire resistive" was adopted to suggest that nothing is totally
impervious to the effects of the most severe fire which can be envisioned. Today, we are making the same mistake by encouraging the use of the terms
"safe" and "safety" in connection with the fire-protection features of a
building. High-rise buildings are being loaded up with electronic gadgetry
called "life safety" systems, which presumably will do wonders to cut down
on the horrible risk of being burned alive on the 9th, 19th, or 90th floor of
our modem apartment and office buildings.
'Senior vice president, Gage-Babcock & Associates, Elmhurst, III. 60126.
28
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
COHN ON FORMULATING ACCEPTABLE LEVELS OF FIRE RISK
29
What is the risk of being killed in a high-rise without these special "life
safety" features? What is the risk after the typical building code "high-rise
package" has been added? Less? Probably, yes. But safe? No.
To the person who promotes high-rise sprinklers and computer-controlled
fire management systems, the risk could appear unacceptable without these
features. To the enforcing authority it must be unacceptable; otherwise there
would be no legal or moral justification for forcing their installation. But
what is acceptable? Who decides what is acceptable? How does a person
decide that sufficient gadgetry has been added to make the building "safe,"
if he is foolish enough to make that claim, or "reasonably safe," if he is a little smarter?
For the building owner or architect, there is essentially only one measure to
use in deciding whether an acceptable level of fire risk has been achieved.
This acceptable level is what is "acceptable to the authority having jurisdiction." Codes and standards are the basic tools of the authority that has
jurisdiction in deciding the acceptability of a particular construction or arrangement. Experienced judgment may also be used, and more and more
often, today, a fire safety analysis may be commissioned by the owner to show
that the intent of the code has been met.
The degree of safety—the level of risk—obtained when the requirements of
a code are complied with is unknown. This is true because (1) we have no
tools with which to measure the degree of risk with any kind of accuracy, and
(2) even if we had, we would not use them because of a reluctance to admit
that there is an acceptable level of risk where the safety of human lives is involved.
Codes and standards would be greatly improved if their goals were better
defined. While initially the persons preparing a code or standard will have
definite objectives in mind and will try to achieve a certain level of safety, the
goals often are obscured within a short period of time. Even before its initial
adoption, pressures are exerted to ease the impact on a variety of special interest groups, and by the time the code has been through several revisions, it
becomes difficult to imagine why certain requirements exist, let alone what
their incremental benefit is. Hit-or-miss changes are made which upset the
interrelationships between the various sections of the code.
Despite the often-heard claim that today's building codes are performance
codes, they only permit the construction of a traditional building. The construction types—fire resistive, ordinary, mill, wood frame, and so on—reflect
the types of buildings that were popular in the late 1800s. Innovative arrangements and modern construction techniques are difficult to classify and
tend to be discouraged. For instance, the atrium building still violates most
building codes; it just cannot be fit into a mold which says there can be no
unprotected openings between floors. Where they have been built, the codes
have been modified after the fact to accommodate them, or variances have
been granted, usually after the owner has agreed to some extra protection
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
30
FIRE RISK ASSESSMENT
which presumably makes the big hole in the middle equivalent to a traditional building. We don't know what the real risks are in an atrium
building—some studies Gage-Babcock engineers have made indicate that the
big, open space provides better life safety protection than the traditional,
enclosed corridor as a path of egress.
A true performance code should permit unlimited variations in the construction and arrangement, provided that the end result conforms to some
stated goals of safety, property conservation, public access, and energy conservation. Such a code is not practical today because of the difficulties in its
administration and enforcement. An easy-to-apply method would have to be
available with which to compare the degree of safety attained by the various
combinations of materials and constructions to ensure that the intent of the
code is met. The authorities who approve the plans for the construction of an
innovative building need sufficient guidance to allow them to decide easily
and quickly whether the new concept is acceptable under the code.
Expressing the level of safety in terms such as "minimum," "reasonable,"
or "acceptable" is not adequate. Goals for fire safety need to be stated in
terms which can be measured. Only with such quantification will a meaningful, systematic fire safety analysis be possible. Without such analysis, it is
doubtful that ASTM Committee E-5 on Fire Standards will ever issue a fire
risk assessment standard which is worth the paper it's written on.
For an obvious reason, assigning real numbers to the results of a life safety
probability analysis would be unacceptable. It would lead to the situation of
seeming to weigh money spent against lives lost. While people appear willing
to accept certain risks for themselves, they seem unwilling to have less-thanperfect safety levels set as a matter of public policy. Setting risk levels in
terms such as the number of lives lost, or saved, seems inappropriate. Alternatively, it is practical to agree on a certain building arrangement as fulfilling our expectations of reasonable safety. Other arrangements can then be
compared with this standard.
As an example, Gage-Babcock recently submitted a proposal for evaluating the deployment of forces of a large-city fire department. In the usual
approach, fire stations are located so as to be within an arbitrary travel
distance (or travel time) from most buildings, to provide a level of safety
which is considered reasonable by the authorities. Any recommendation for
closing an existing station is met with opposition because people perceive
that they will be getting less protection than before. The analysis we proposed
would first analyze the level of protection which the city has been providing
and which the citizens consider acceptable. Any proposed changes would be
compared with the existing, acceptable arrangement rather than with an arbitrary standard.
Most of our modem buildings today provide what we would usually consider to be an acceptable degree of life safety protection from fire. With low
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
COHN ON FORMULATING ACCEPTABLE LEVELS OF FIRE RISK
31
flammability and good exiting facilities, fire fatalities in occupancies such as
offices, schools, places of assembly, factories, and warehouses are rare.
When fatalities do occur, they are often the result of an abnormal, temporary, higher hazard condition, such as construction, cleaning with flammable liquids, a natural gas leak, or flammable decorations. In fact, it appears that in places where practically everyone is awake, mobile, and can
escape from a fire environment without assistance, fire fatalities are unusual.
The acceptable level of fire risk can be set quite low.
A somewhat higher level of risk may have to be accepted where there is a
relatively high number of occupants who have physical or mental impairments or who otherwise are restrained from escaping a fire without
assistance. Included in this group are the institutionalized, the inebriated,
and the drugged. The level of risk appears to be in direct proportion to the
level of impairment and in inverse proportion to the number of rescuers
potentially available.
At the recent Federal Emergency Management Administration/Department of Health, Education, and Welfare/National Bureau of Standards symposium in Washington on fire safety for the handicapped, an early conclusion reached by the panel on detection and alarm was that a higher level of
risk appears to be inevitable for the incapacitated, helpless person left alone
in a residence. With no one to assist, the very young, the senile, the crippled,
and the heavily drugged or inebriated stand comparatively little chance of
escape. Fire statistics seem to bear out a preponderance of fire fatalities in
this group. The sterility of surroundings needed to lower the hazard effectively under these circumstances is not likely to be acceptable to society.
About 4 years ago John CampbelP and I started work on a methodology
for systematically tracing the alternative paths of fire development in
buildings from incipiency to full-room involvement and, from there, to other
spaces. In this critical path technique, the paths radiate out from a point
representing the initial contact of a source of ignition with a source of fuel
(Fig. 1). In most instances, fire does not result. The igniting source may be
too brief or of too low an intensity; there may be inadequate fuel or the fuel
may be too difficult to ignite; or somebody may notice the igniting flame or
spark and remove the fuel from the danger zone before it can ignite. All of
these possibilities can be shown by paths radiating out from the initial point.
One of the paths represents the successful ignition of the fuel and the progression of the fire to an incipient stage.
A methodology for quantifying the relative probability of the occurrence of
each event was then developed. Formulas, tables, and graphs were prepared
to show the apparent relative hazard, based on statistical data, observations,
test results, handbook information, and other sources. A couple of the
^Manager, Chicago office, Gage-Babcocl« & Associates.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
32
FIRE RISK ASSESSMENT
Ignition Source
Contacts Fuel
Incipient Fire
Occurs
Meaningful Fire
Occurs
FIG. 1—Diagram of the critical path for fire development in a building, for use with the critical path method for fire hazard analysis, developed by Gage-Bahcock & Associates in 1978,
graphs which were prepared are shown in Fig. 2, and the type of input
needed to complete the worksheet and calculations is given in Table 1. At
this time, the methodology is useful primarily to demonstrate the feasibility
of the approach. Trial applications which have been made were judged to
have accurately quantified the relative hazard presented by alternative fire
protection schemes, as a tool for justifying equivalencies and trade-offs in
code compliance.
A probabilistic approach to fire risk analysis is not new. Rating systems for
calculating relative fire hazard are more than 70 years old, and the concept of
probabilities is inherent in every building code. Probabilities may not be expressed in quantitative terms, but the very fact that codes develop a range of
requirements for different loss expectancies is indicative of a thought process
which includes consideration of the probabilities of various events occurring.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
COHN ON FORMULATING ACCEPTABLE LEVELS OF FIRE RISK
33
Meaningful Fire
Occurs
Fire Involves
Room/Space
Fire Penetrates
Barrier
Ignicion of Fuel
In Next Space
7^
FIG. 1—Continued.
The critical path method for fire hazard analysis can be expanded to include probabilities of human fatalities at each point in the chain of events. As
now conceived, the method considers real time only in relation to the probability of certain interactions occurring in the fire development sequence
(such as fire department response time). However, escape from fire is usually
thought of in real-time increments. "You have only 2 min to escape" is a
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
34
FIRE RISK ASSESSMENT
QJ
III
• ^ ^
E S g §.
W
(U m
[fi
ss
52 o s 5 55 535
( M/qi) P'^o'l 3->TJ
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
COHN ON FORMULATING ACCEPTABLE LEVELS OF FIRE RISK
35
^1
!W -
S 3
'*3 .'i*
<u -o
"5 .s
"5 "<.
"IK"
1*^
£ g
s-
j j
oc-a
s sa
Ki •P
e
VI
o u
CI,
^
•^
o
.> ?0
«a
>* H
-£
en
S
^
u.
n
H
• ^
3-S
o
=3•w§
1
<N
^
N
o -^
• ^
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
36
FIRE RISK ASSESSMENT
TABLE 1—Input required for the critical path method for fire hazard analysis.
Variable
1. Room contents
(a) Ease of ignition
(b) Spatial distribution
(c) Class of combustible (A or B)
(d) Fire load
(e) Flash point of Class B combustibles
2. Physical dimensions
(a) Room or space
(b) Openings in room/space enclosure
3. Fire behavior classifications
(a) Flame spread ratings
(b) Fire resistance ratings
4. Human response for fire suppression
(a) Percentage of time room is occupied
(b) Response time from other spaces
(c) Fire department or fire brigade response time
5. Type and quality of fire protection equipment
(a) Automatic fire detection
(b) Fire suppression
(c) Nonfire instrumentation
How Obtained"
C or M
M
C
C or M
C
M
M
C
C
M
M
M
C
C
C
"Key to abbreviations:
C = classified on the basis of existing standards or guides
M = measured (or estimated)
commonly used warning, but since fire does not develop in any consistent
pattern, the probability of having only 2 min to escape in any given fire incident is actually quite low. Relatively consistent with fire development,
though, are carbon monoxide (CO) generation and human susceptibility to
CO concentrations. It appears to be feasible to develop probabilities for the
presence of people in a fire environment coincident with the occurrence of
critical CO levels.
For the most part in fire hazard analysis, sufficient test data or experience
statistics are not yet available to develop reliable probabilities for the occurrence of events. Instead, a subjective, or personalistic, concept of probability
is utilized. Under this concept, the probability of an event is the degree of
belief or degree of confidence placed in the occurrence of an event by a particular individual based on the evidence available to him. The evidence may
consist of relative frequency of occurrence data to the extent available and
any other quantitative or nonquantitative information. If the individual believes the occurrence of an event to be unlikely, he assigns a probability close
to zero. If he believes that it is very likely to occur, he assigns a probability
close to 1.
The subjective probability concept is particularly suited to fire protection
systems analysis and the development of codes and standards. Probabilities
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
COHN ON FORMULATING ACCEPTABLE LEVELS OF FIRE RISK
37
can be assigned for events occurring only rarely and in the future and for
which no objective data may be available. The techniques for applying these
concepts are available to us today. ASTM standards can help in validating
data by standardizing test fires and by developing relationships between test
fires and real fires in buildings. As an example, there is a need to establish
valid probabilities for the failure of Vi-h, 1-h, 2-h, and 4-h rated wall and
floor assemblies under actual fire exposures of varying severities. It is generally
known that an assembly that has obtained a 2-h rating under standardized
fire test conditions may fail when exposed to a building fire of similar intensity, or it may fail prematurely under shorter exposure to higher
temperatures than that of the test fire. Knowing the probability of such
failures tends to produce more accurate risk assessment.
An analysis method such as the one developed by Gage-Babcock & Associates could be the missing link to formulating acceptable levels of fire risk.
Using this approach, a probability rate can be determined for the occurrence
of a fatality during a fire event in any particular type of building. This rate
can be calculated for a building arrangement which is considered to meet an
acceptable level of safety. Using this number as the norm, other building arrangements can be analyzed and a determination made as to whether the fire
fatality rate is as low as the norm. Preliminary indications are that this
analytical method can be used to verify that codes and standards set
minimum requirements, at a level of fire risk which is socially acceptable.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
W. C. T. Ling^ and R. B. Williamson^
Using Fire Tests for Quantitative
Risk Analysis
REFERENCE: Ling, W. C. T. and Williamson, R. B,, "Using Fire Tests for Quantitative Rislj Analysis," Fire Risk Assessment, ASTM STP 762. G. T. Castino and T. Z.
Harmathy, Eds., American Society for Testing and Materials, 1982, pp. 38-58.
ABSTRACT: Fires can be considered a causa/ chain-of-evenfs in which the growth and
spread of fire may cause damage and injury if it is rapid enough to overcome the barriers
placed in its way. Fire tests for fire resistance of the barriers can be used in a quantitative
risk assessment. The fire growth and spread is modeled in a state transition model (STM).
The fire barriers are presented as part of the fire protection model (FPM), which is based
on a portion of the National Fire Prevention Association (NFPA) decision tree. An emergency equivalent network is introduced to couple the fire growth model (FGM) and the
FPM so that the spread of fire beyond the room of origin can be computed. An example is
presented in which a specific building floor plan is analyzed to obtain the shortest expected time for fire to spread between two points. To obtain the probability and time for
each link in the network, data from the results of fire tests were used. These results were
found to be lacking, and new standards giving better data are advocated.
KEY WORDS: fire protection, fires, fire resistance, fire tests, fire growth, fire spread,
fire safety, probability analysis, fire risk assessment, and stochastic processes
The quantitative assessment of fire risk requires a rational framework to
measure the "fire safety" of a building or other structure. Harmathy [/]^ has
defined a fire-safe building as one for which there is a high probability that
all occupants will survive a fire without injury and in which property damage
will be confined to the immediate vicinity of the fire area. Under certain circumstances, such as for nuclear reactors, the criteria would need to be changed
slightly, but, in general, the definition can be applied to many buildings or
other structures and systems. Note, however, that this definition is conditional because it assumes that a fire will occur, and it is the impact of such a
fire which is being minimized.
'Ph.D. candidate and professor of engineering science, respectively. Department of Civil Engineering and Lawrence Berkeley Laboratory, University of California, Berkeley, Calif, 94720,
^The italic numbers in brackets refer to the list of references appended to this paper,
38
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
39
Friedman [2] starts his analysis of the threat from a rapidly growing fire at
an early stage, and he defines his framework as the product of two probabilities:
1. The probability of ignition, for a given ignition exposure specified by the
scenario.
2. The probability of being unable to escape or be rescued from the fire before the escape route is blocked by heat, smoke, toxic gases, etc. (This probability clearly depends on detection time and on the degree of mobility of the occupants, as well as on the rate of fire growth. Accordingly, some people will fail
to escape even from a very slowly growing fire.)
Friedman does not develop the probability of ignition, although he notes that
it may be developed in a "straightforward manner." He represents the fire
growth with an exponential equation which yields a range of "fire doubling
times" for a variety of fuel materials. He then develops a ratio of criticalthreat variables that relate to the manner in which the fire is detected and to
the manner in which it reaches the critical condition.
One of the essential steps in establishing a quantitative assessment of fire
risk is to separate the diverse and complex issues of unwanted fire so that
they can be treated individually. Williamson [3] has suggested that every
serious fire has two essential characteristics: (1) it can be considered as a
causal chain of events, and (2) the element of time is of ultimate importance.
"But for the rapid growth of the fire in a particular location" is a typical
phrase that almost always marks the difference between a serious fire with a
tragic ending and an incidental fire which was easily controlled. This sine
qua non (but for the presence) logic and the element of time are important
components of a universal fire model. Therefore, in the following section a
state transition model (STM) of fires will be introduced which represents the
growth of the fire in the room of fire origin and its subsequent spread beyond.
The fire resistance tests of walls and floor or ceiling assemblies, as well as
of the elements interrupting the integrity of those assemblies, such as doors,
windows, and pipe and cable penetrations, are introduced in the risk assessment as quantitative data concerning the ability of those components to contain fire. The fire tests for flame spread and the full-scale fire growth experiments are used to quantify the growth of fire in the room of origin and then,
once the containment elements are breached, its spread to other spaces. The
quantitative assessment of fire risk is strongly dependent on the quality and
completeness of the information provided by the fire tests.
State Transition Model
Williamson [3] presents a method for combining both deterministic and
stochastic modeling techniques for unwanted fire which is based on an STM.
This STM is flexible enough to accept theoretical and experimental, as well
as subjective, information. Williamson's formulation of the problem is based
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
40
FIRE RISK ASSESSMENT
on the fact that, though there are a number of distinct aspects to any serious
fire, they all have a common time coordinate, and certain events can be identified as being particularly significant. Any complete mode! of unwanted fire
development must be able to answer the following four questions:
1. What is the extent of fire growth with time?
2. What is the extent of the spread of combustion products (smoke) from
the fire as a function of time?
3. What are the roles of people involved with the fire in causing its ignition, in fighting it, and possibly in becoming its victims?
4. What fire protection techniques can be expected to prevent the spread
of fire and smoke?
Each one of these questions can be looked upon as a separate issue and can
be utilized as a subsection of a complete model. One way to visualize this process is to compare it with the "separation of variables" common to solving
many determinisiic problems utilizing differential equations. The complete
model thus consists of submodels which are defined by the four questions just
listed.
The first of these questions can be answered by a fire growth model (FGM),
the second by a smoke spread model (SSM), the third by a human response
model (HRM), and, finally, the fourth by a fire protection model (FPM).
Each of these subsections of the overall model can be formulated as state
transition models which are divided into separate states or realms that represent the conditions of the fire appropriate to each subsection of the complete
model. The use of state transition models in representing fires was given an
important boost by the National Fire Protection Association/Department of
Housing and Urban Development (NFPA/HUD) research project in the
period 1974-76. In particular, the works of Connelly [4,5] and Berlin [6]
were important in spreading these ideas.
The FGM can be considered to give a running account of the fire until it is
contained by barriers or has been extinguished by manual or automatic suppression. Williamson [7] presents an FGM that is based on dividing the fire
growth process as if it were a full-scale experiment in which a given product is
exposed to an ignition source, and the growth of the fire is recorded as a function of time.
If the initial state before the fire is called "I," the subsequent states can
bear the letters that follow in the alphabet, J, K, L, M, and N. If the unwanted fire is inside a building or similar structure, it will proceed in a manner similar to that shown schematically in Fig. 1 for a living room fire. Figure
la shows the room before the ignition events which start the fire scenario. A
fire could have started in a number of locations, but we have chosen the wastepaper basket, and Fig. lb shows a plume of flame and smoke rising out of
that location. This depicts State J, since only the waste container and no
other object has been ignited. It is important to note that the flame does not
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
41
(a)
FIG. 1—Schematic diagram of flame spread in the room of origin from ignition to flashover.
reach the ceiling, although there must be convective heat transfer to the ceiling since heated air and combustion products rise to the ceiling. The black
"smoke cloud" above the wastebasket can be considered to represent schematically the buoyant heat plume as well as the smoke per se, and this heat
and smoke will "pool" in the upper portion of the room. This pool of hot gases
and smoke will be at least as deep as the soffits of the highest door or other
opening in the room.
In his full-scale experiment, which would correspond in a more general
household fire scenario to the ignition of the wall or furniture, Williamson
[7] defined the end of State J and the beginning of State K as the ignition of a
second object. There may be multiple ignitions in the vicinity of the original
ignition source, but if a fire is to become serious, the actual number of burning objects is not as important as the conditions in the zone of hot gases filling the upper portion of the room. At this state of the fire growth, the instant
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
42
FIRE RISK ASSESSMENT
(b)
FIG. 1—Continued.
the flames reach the ceiling constitutes one of the most important events,
since combustible gases, or the ceiling itself if it is combustible, can ignite
and contribute to the burning in the room. The flames shown in Fig. Ic have
reached the ceiling and are spreading along its surface. These flames add to
the heated zone at the ceiling and reinforce the radiation which is bathing the
room. One of the important consequences of this radiation is that other combustible materials can be ignited, as is shown schematically in Fig. Ic by the
ignition of papers on the desk and the couch near the windows. Following ignition of the wall or furniture, the moment when the flames first reach the
ceiling thus stands out as the next most critical point in the chain of events of
fire growth. Williamson chose to denote this moment as the end of State K
and the beginning of State L.
As more and more heat builds up in the room, there is a point in time, known
as flashover, when all of the combustible materials in the room can ignite and
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
43
(c)
FIG. 1—Continued.
begin to contribute to the fire. From that moment, the fire in such a room is
considered "fully involved." The flashover state is schematically shovi^n in
Fig. Id, where the windows are broken due to the thermal shock, and flames
are shown spreading out through both the windows and the doorways. Fires
will rarely cause injury to very many people or cause large property damage
unless they reach flashover in at least the room of origin, and, in general,
such fires are also characterized by rapid fire spread beyond the room of origin.
Williamson shows the flashover event as the end of State L and the beginning
of State M, which is the initial state of the fully involved room fire. State M is
characterized by an air supply rate which is insufficient to bum completely
all of the fuel released by the contents within the boundaries of the room.
This ventilation-controlled portion of the fire has ended when enough fuel
has been expended or more openings have been supplied to allow most of the
burning to take place in the room. That event ends State M and begins State
N which, unless new fuel is supplied or the air supply is diminished, will be
the final state of the fire.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
44
FIRE RISK ASSESSMENT
Flashover
(d)
FIG. 1—Continued.
The definitions of these states of fire growth in a room are as follows:
I—the initial prefire period of time, which ends with ignition of the
"source";
J—the period of time from the ignition of the source until the ignition of
the wall or furniture;
K—the period of time from ignition of the wall or furniture until flames
touch the ceiling;
L—the period of time from the moment the flames first touch the ceiling
until full involvement (flashover) occurs;
M—the period of time from flashover until sufficient air enters to bum the
fuel in the compartment; and
N—the period from the time that enough air enters to bum the fuel in the
compartment until the fire has burned out.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
45
An event diagram, such as the one in Fig. 2, can be constructed to represent this fire growth model. The fire can terminate in any state, and generally, until flashover occurs, the fire can be considered confined to the original
compartment. It is during State M that the fire is most likely to spread through
openings in the walls or ceiling. In the adjacent space, the fire growth process
then starts from a new State J in which the flames emerging from the room of
origin can be considered the ignition source. This is schematically represented
by the symbol (R) in Fig. 2.
This paper will concentrate on the FGM and the FPM, and it will not explicitly develop the SSM or the HRM. The FPM consists of three distinct elements: controlling the combustion process; suppressing the fire; and controlling construction, as described in the well-known National Fire Protection
Association (NFPA) decision tree under the heading, "Manage the Fire" [8].
That portion of the NFPA decision tree is shown in Fig. 3 for illustrative purposes. This decision tree is a success tree, with the top event denoting the satisfactory management of the fire to prevent injury or loss of property. This
type of success tree is the mirror image of a fault tree, in-which the top event
would be the loss of fire control leading to fire injury or property loss. The
FPM is more appropriately described by a success tree since it is designed to
manage the fire rather than to allow its uncontrolled growth. While the FGM
focuses on fire growth, the FPM essentially opposes it.
c
0
STATE
•1_
i'!initi
sourc
rce
burning
wall or
furniture
burning
STATE•
J
fire fully
involved
STATE ,
K
r
iqnition
of source
fire goes through
^
walls and/or c e i l i n q \ j /
flames at
ceiling
T
ignition
of wall or
furniture
STATE
L
Process terminates
/TN
\V
fire goes through
window
-0
f i r e contained in
t-0
original
conipartmen
t
flames at
ceiling
0
©
©
fire goes through
•?oors
STATE
M
r
FLASHOVER
Fire process starts
in room of origin
1_STATE
t
' '
switch from
ventilation to
fuel control
Process starts from J
in adjacent space
FIG. 2—Fire growth model for fire in a room.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
46
FIRE RISK ASSESSMENT
i iij
rm r
^
^
L
®-
i
s
p k^H
^S^
m
in
c
i
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
47
Applying Fiie Tests to Quantify the State Transition Model (STM)
Both the FGM and FPM can be quantified by the use of fire test information. The FGM can be quantified by the use of full-scale fire growth tests
[3,7] and tunnel test information [9]. One of the major NFPA decision tree
branches, that involving controlling by construction, can be quantified by
using fire-resistance test results. It is assumed here that flashover has occurred
in the room of origin. It should also be remembered that the containment of
a fire depends on the walls, the doors, the ceiling, and perhaps other building
elements such as windows or cable and pipe penetrations.
The quantitative development of both the FGM and FPM will be presented
through the use of an example. Let us suppose that a certain building is being
considered for rehabilitation. Figure 4 shows the floor plan of a high-rise of-
i _ ^ elevator rv—^
\~7|J lobby L i W
ii i
, stairway
FIG. A—Twenty-third floor of a high-rise office building built in the 1930s.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
48
FIRE RISK ASSESSMENT
fice building constructed in the late 1930s. The arrows point to the only stairways which provide escape routes from each floor. The many offices on both
sides of the dead-end corridor labeled A-B could be isolated if a fire started
in an office such as Room 1, and then blocked the corridor between Rooms 1
and 3. Moreover, the corridor doors in the actual building are 90 percent plain
glass (as opposed to wire glass), and once a fire has fully involved an office,
such as Room 1 in Fig. 4, it would break the glass in the door and spread
smoke and fire into the corridor. Not only would this prevent the escape of
everybody in other offices served by this corridor, but it would also cause the
plain glass in the door of the opposite office. Room 3, to shatter and allow the
fire to spread into that space. The fire would then proceed down the corridor
from A to B, entering the remaining offices by way of the plain glass doors.
More than 30 floors of the building have the same basic floor plan.
At the start of our fire risk assessment we can see that the configuration of
Rooms 1 through 4 shown in Fig. 4 can be represented schematically by the
elementary floor plan in Fig. 5 (left). The rooms are simplified to squares,
and the corridor is represented by Segments C) and C2 which are opposite
Rooms 1 and 2, respectively. Then, in order to enable us to consider the spread
of fire beyond the room of origin, the floor plan is transformed into a network, as shown in Fig. 5 {right). This is similar to the process described by
Busing et al [10]. Each link in the network represents a possible route of fire
spread for the FGM, and some of the links are exit paths for the HRM.
The method [11] to be used in this paper for calculating the expected
shortest time for the fire to spread from one compartment to another is based
on the premise that the duration of fire resistance of each barrier element,
and the time needed for the fire to develop fully in each compartment, have a
given independent discrete probability distribution. We shall begin by showing how to construct a discrete probability distribution from fire test data.
First, partition the time axis into equal spacing intervals. Then construct a
,
•
Rm 3
Rm 1
^1
Rm 3
I.
- -13 . .
0
Rm 2
Rm 4
Rm 2
Rm 4
•
FIG. 5—Illustration of the transforming of a floor plan (left) into a network (right).
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
49
histogram for the duration of the fire resistance of a particular material and
element. From this histogram, a discrete distribution can be made by collapsing everything within each interval into the midpoint of that interval. For
example, the fire resistance of various 10.16-cm (4-in.) clay tile masonry
walls is reported to vary between 10 and 35 min [72]. There are eleven different fire tests on exposed clay tile assemblies which can be used to produce
the histogram shown in Fig. 6 {left). This histogram can then be replaced by
the discrete distribution shown in Fig. 6 (right). The discrete distribution for
this 10.16-cm (4-in.) clay tile wall is
Pb(0
Pi
= 1/11 = 0.091 at f = 5
P2
= 3/11 = 0.273 at ^ = 15
Pi
= 6/11 = 0.545 at r = 25
PA
= 1/11 = 0.091 at t = 35
If we assume that Room 1 is the room of origin and that, once a compartment has been burned out, fire cannot return to that compartment, then the
network shown in Fig. 5 can be modified to represent the spread of fire and
the passage of time. It now becomes the FGM network, and is schematically
arranged in Fig. 7 to show a flow of time to the right. The nodes denoted by
primes each represent a fully developed fire in the compartment. Three different types of links are identified. The first corresponds to the fire growth in
a compartment, the second to the fire breaching a barrier element, and the
third to the fire spread along the corridors. To each link i, a pair of num-
4 -
10
20
30
TIME (MINUTES)
40
10
20
30
40
TIME (MINUTES)
FIG. 6—Histogram (left) and discrete distribution (right) of 10.16-cm (4-in.) clay tile walls.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
50
FIRE RISK ASSESSMENT
Rni 1
Rni 3
Cl
--*•-
(Pftf)
\
(Pt,.t„)
»» • ' »
\JPf.tf)
Rin 4
Rm 3 '
Rm 4 '
»
>^
(p^.tj,)
) \
\
(Pf.tf)
(P^,t^)
(Pf.tA,
't-b'tb'
(P , t
\
(n , t )
\
- • i ^ — • «» ««
Rm 2
—H
fire growth within compartment
fire breaches barrier elements
fire spread along corridor
F I G . 7—Probabilistic network
of fire spread of Room I to Corridor
C2
bers, (pi,/i), is assigned. In this paperpj represents the distributed probability that a fire will go through Link i, and fj represents the time distribution
that it will take for such a fire to go through Link i.
The number pair, (pf,^t)< represents the probability and time for the fire
in the room to progress from State J through State L, that is, to become fully
involved. The flashover time, ?f, in a room varies according to the type of wall
and ceiling lining material used, as well as the room's contents. For this example the FGM in each space is simplified by taking the approach of Lie and
giving a probability distribution based on flame spread or reaction to fire
tests [9]. Figure 8 shows Lie's estimates for the time to flashover of three levels
of performance represented by a flame spread classification of 25, 75, and
150 as measured by the tunnel test [ASTM Test for Surface Burning Characteristics of Building Materials (E 84-80)] or the Dutch "flashover" test. It
should be noted that the authors doubt the validity of this behavior for plastic
lining materials but consider it to be realistic for cellulosic materials [13].
The section of the corridor, C,, opposite Room 1 is treated as a separate
fire compartment and is assigned a (pf,ff) for the link from C] to C i ' . The
number pair ips^h) represents the probability and time for the preflashover
spread of fire along the corridor from C, to Ci. As a first approximation, the
probability of preflashover spread of fire in the corridor from C] to C2 could
be considered to be governed by the flame spread classification of the corridor's finish materials on the walls and ceiling, as measured by a test
method such as ASTM Test E 84-80 (tunnel test). This, again, would be limited to cellulosic materials, although a detailed review of the test results could
yield information for plastic lining materials.
Once full involvement occurs in Section C1 of the corridor outside Room 1
(that is, Node C\' is reached) the fire spread in the corridor is influenced
more by the ventilation in the corridor and by the contribution of Room 1
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
ilRONG
CONTRIBUTING
150 TUNNEL IE SI [3.8)
._„1 ,
0.05 c
FUVSMOVEB IE5I [3.21]
sO)
MODE RATE
CONTRIBUTING
75 TUNNEL TEST [3.8]
0.2 col.'cm' !
FLASHOVER TEST [3.21]
51
NON
CONTRIBUTING
25 TUNNEL TEST [3.8]
0.4 col/cm2 t
FLASHOVER TEST [3.2l]
^XD
10
FLASHOVER
IS
TIME,
20
MINUTES
FIG. 8—Frequency offires with a certain flashover lime for rooms lined with materials which
are strong contributing, moderately contributing, and noncontrihuting factors to the fire
growth.
than by the material properties of the corridor itself [14]. Thus, there is a
separate link, C , ' to C2, which has its own (p^, t^).
The number pair (p^,, t^,) represents the probability of failure of the barrier
element, with t^, representing the endurance of the barrier element as measured by a standard fire-resistance test such as the ASTM Fire Tests of Building Construction and Materials (E 119-80). Barriers may consist of such
components as walls, ceilings, and doors. Penetrations such as piping or
cables through walls will affect the fire resistance of barrier elements. If not
properly protected, the penetration tends to increase Pb and decrease t^,.
With regard to doors, an open door would offer no resistance to fire spread,
and ?b would equal zero. On the other hand, if the door is closed, t^ would
depend on the type of materials that the door is made of. Figure 9 shows
some distribution of t^ for doors, based on the authors' subjective judgment.
The probability of doors being open would depend on the presence of selfclosing devices and the absence of door stops.
Coupling the Fire Growtli Model (FGM) and the Fire Protection Model (FPM)
As shown earlier, the FGM and FPM can be quantified by the use of fire
test data, but to solve the problem of determining the risk of fire spread it is
necessary to couple these models together. We have chosen the use of an
emergency equivalent network. This is an abstraction of the given network in
which (a) each link has a Bernoulli probability of success, (b) the link delay
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
52
FIRE RISK ASSESSMENT
OPENED
DOORS
aASS/IIOLLOW
DOORS
20 MINUTES
RATED DOORS
20
ONE HOUR
RATED DOORS
—r—
—r-
40
30
TIME
—r-
60
50
(MINUTES)
FIG. 9—Frequency distributions for fire containment by three types of doors.
time is deterministic, and (c) the expected shortest time between two points
in the emergency equivalent network is equal to the expected shortest time in
the original network. If one link in the original network, say the link between
Nodes j and k, has discrete probability distribution
'Pr^iort = <m
m = 1,2, . . . r,
P\M(') =
(1)
Po„ =
1 —
I ! Pi
for
t - CO
then the transformation into an emergency equivalent network can be accomplished by replacing the (j,k) link by r parallel links (j,m,k), m = 1,2,.. .r,
where each link (j,m,k) has an independent probability of success, /?„, and
associated time, t^, where
Pi
P\
2,3,
...r
(2)
If we are interested in the expected shortest time for the fire spread between
two nodes in the network—call it the source node and the sink node—we
have to list all the possible paths between these two given nodes. Many different algorithms are available to find all the paths [15]. Let Pi,P2,.. .Pn be
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
53
all the paths going from the source to the sink node. For each Path Pj, i =
1,2, ... n, find the travel time, Tj, and the probability, P\, where
T,=
L
linkjePi
u
Pi=
n
Pj
(3)
linkjePi
Number the n paths so that T^ < Ti < • • • < T„. Let E^ denote the event
that Path Pj is connected; that is, with probability greater than zero, fire will
spread through each node in Pj. Let Q^ denote the probability that the fire
will spread from the source to the sink node by way of one or more of the
paths Pi,P2, .. • Pm- That is mathematically expressed
Q „ = prob [ £ , U £2 U . . . U E„\
(4)
Let the conditional probability that Paths P| or P2 or . . . or Pn,_| are connected, given that Path Pn, is connected, be denoted by Puim- That is
Pu|„ = prob[£, U r 2 U . . .
U£„_,|rj
Let us define Pu|o ~ Puji ~ Qo ~ 0. We then obtain /"uim from Q^-x using
Pu|m = [ ( 2 m - l ] / ' . - ,
(5)
where the subscript Pm-i indicates that all the link probabilities associated
with P„ are set equal to unity in the subscript expression. Using basic probability theoty, we can then show that
<2m =
Qn,-.+/'n,[l-Pu|m]
(6)
Where P^ is the probability that Path Pn, is connected. Equations 4 and 5
can be used recursively to determine expressions f o r P u | i , P u | 2 . ••• PuinThe probability of connectivity, R, is then given by
/? = P, + ^2 (1 - Pu|2) + • • • + ^n (1 - PU|„)
(7)
and the expected shortest time for the fire to spread from the source node to
the sink node, given that the two nodes are connected, T. is given by
r = [r,p,+ 7^2^2(1-Pu 12) + ••• + r„/'„(i-Pu|„)]/i?
(8)
Nnmerical Sample Solution
Let us go back to the example stated previously involving unrated doors on
the corridor. Take the floor plan shown in Fig. 5a and suppose that we want
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
54
FIRE RISK ASSESSMENT
to find the expected shortest time for a fire to spread from Room 1 to Corridor C2, given that Room 1 has ignited. This would indicate that the fire
growth is such as would make the corridor unavailable for either an exitway
or route of access for fire suppression in the rooms along the corridor. The
following assumptions are made for this illustration:
1. All the paths involving Rooms 3 and 4 are ignored.
2. The room of origin (Room 1), as well as the other rooms, is to have contents and lining materials that are characterized by Lie [9] to be moderately
contributing to fire growth and, thus, according to Fig. 8, has a tf of 10 min.
3. The probability that a fire in the room of origin will grow to flashover is
0.50, and in the other rooms it is 1.
4. The initial assumption will be that all rooms have plain glass doors
without self-closing devices leading to the corridor. This will be modified
later to illustrate a shift to self-closing 20-min doors. It will be assumed that
the probability of a door being closed is 0.50. This gives the following formulas
p, - yi-
0.50 at f - Omin
P2 = '/2 = 0.50 at f = 5 min
Using Eq 2 results in having two links between rooms and corridors with
(Pi,fi) being equal to (0.50, 0) and (1, 5).
5. The walls between the rooms are 10.16-cm (4-in.) clay tile masonry
walls. Using the discrete distribution shown earlier and Eq 2, we have four
links between rooms, with {p\,t-^ being equal to (0.091, 5), (0.30, 15), (0.86,
25), and (1, 35), respectively. Notice that the application of Eq 1 results in
the fire having a probability of 1 of eventually breaching the wall in 35 min.
6. The differentiation of Cj and C | ' will be dropped for this example. An
open door or a plain glass door will cause a very severe fire in the corridor,
and transition to flashover in the corridor will be very rapid. To estimate the
fire propagation from C, to C2, we will rely on National Bureau of Standards
(NBS) corridor test data reported by Fung \16\. Neglecting the inclusion of
the initial fire growth, these data give the following distribution
Pi = '/2 = 0.50 at t = 7.5 min
P2 = '/8 = 0.125 at t = 12.5 min
This results in having two links between C] and C2, with {p\,t-^ becoming
equal to (0.50, 7.5) and (0.25, 12.5), respectively.
An emergency equivalent network to represent the fire growth and the
various barriers to its spread can now be constructed for this example, and it
is shown in Fig. 10.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
55
L5 ( 0 . 2 5 , 1 2 . 5 )
FIG. 10—Emergency equivalent network with 5-min unrated doors.
There are a total of twelve links in the network labeled Xj, L2, L3, ... X12,
from which twelve different paths from Room 1 to C2 can be established.
These paths are listed in Table 1, and the corresponding P, and T^s (as
calculated from Eq 2) are given. The paths are enumerated so that Tj < T2
... < T\i and the component links are identified in Table 1, thereby enabling the reader to identify the path of the fire. Each of these paths can be
described by a fire scenario in words; for instance, Path 1, consisting of Links
£ ] , L2, and X4, would be expressed as "The fire flashes over, escapes from
Room 1 through an open door into the corridor Cj, and then spreads along
the corridor to C2." The probability of that scenario is 0.13 ('/s); it is strongly
dependent on the original subjective assignment of the probability being 0.50
that flashover will occur in Room 1, and of the probability being 0.50 that the
door will be open. The time, J,, of 17.5 min is the combination of the assigned times of 10 min for Room 1 to flashover and 7.5 min for the fire to
spread in the corridor from Cj to C2. If more sophisticated models were incorporated for these links, the final Pj and T^ would be different, but the
reader can see the origin of each value in Table 1. One can also see that as the
paths begin to include the higher-fire-rated wall between Rooms 1 and 2 the
probability increases. For instance. Path 12 consists of Links Zj, X9, L^Q, and
X12, which could be described as follows: "The fire in Room 1 breaches the
clay tile wall and spreads to Room 2. After having reached flashover, the fire
breaches the closed door into the corridor C2." The probability for this
scenario is 0.50 with a time of 60 min.
The emergency equivalent network can be examined further by finding all
the Pu|ni> iTi = 2,3, . . . 12, from Eqs 4 and 5 and then the connectivity, R,
can be found from Eq 6 to be 0.50 and the expected shortest time, T. from
Eq 7 to be 29.6 min. The origin of the connectivity of 0.50 ('/2) is a direct
result of the assumed probability of 0.50 ('/2) for flashover in the room of
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
56
FIRE RISK ASSESSMENT
TABLE 1—Pathways through the example emergency equivalent network assuming 5-min
unrated corridor doors.
Paths
Component Links
1
2
3
4
5
7
8
9
10
11
12
1-2-4
1-2-5
1-3-4
1-6-10-11
1-3-5
1-6-10-12
1-7-10-12
1-8-10-11
1-8-10-12
1-9-10-11
1-9-10-12
Probability, P,
'/8
i/i6
'A
1/44
1/8
'/22
=
=
=
=
=
=
VAO =
3/14 =
3/7 =
'/4 =
1/2 =
0.13
0.06
0.25
0.02
0.13
0.05
0.08
0.21
0.43
0.25
0.50
Time, T,, min
17.5
22.5
22.5
25.0
27.5
30.0
35.0
40.0
50.0
55.0
60.0
origin and the occurrence of unity probabilities in the remaining links which
make up certain paths through the network.
Let us change the starting assumptions slightly and explore how the network changes. Suppose self-closing 20-min fire-rated doors had been installed in the corridor of our example. We will assume that the reliability of
the self-closures is perfect and that door stops had not been allowed. This
changes Assumption 4, described previously, but all other conditions remain
the same. The emergency equivalent network is shown in Fig. 11 and the P,
and Ti for the paths are given in Table 2. Note that the links have been
renumbered for this example. For instance, Lq from Fig. 10 has become Xg in
Fig. 11. The connectivity remains 0.50 in this case, but the expected shortest
time becomes T = 47.1 min. All of the TiS have been increased by the
presence of the fire-rated door, which reflects the rationale of model building
codes that require self-closing 20-min doors on corridors.
Conclusions
The interplay of fire growth and fire protection has been modeled in the
emergency equivalent network. This model facilitates an evaluation of design
changes and affords ready comparison of the different strategies under consideration to effect such changes. The numerical example was necessarily
simplistic, but it illustrated how the original assumptions could be clearly
traced. The individual fire tests on clay tile, for instance, could be seen in the
final pathways of the network. If the original fire in Room 1 had been modeled
more completely, the ventilation-controlled phase of the fire (State M) may
have ended before the 35-min wall had been breached. This more sophisticated step could easily be added, and this would make the connectivity, R.
more meaningful.
Looking at this model in detail, we realize that more information is re-
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
LING AND WILLIAMSON ON QUANTITATIVE RISK ANALYSIS
57
TABLE 2~Pathways through the example emergency equivalent network assuming self-closing
20-min rated corridor doors.
Paths
Component Links
1
2
3
4
5
6
1-2-3
1-2-4
1-5-9-10
1-6-9-10
1-7-9-10
1-8-9-10
Probability, P^
1/4
1/8
1/22
3/20
3/7
1/2
=
=
=
=
=
=
0.25
0.13
0.05
0.15
0.43
0.50
Time, T,, min
37.5
42.5
45
55
65
75
Rni 2
FIG. 11—Emergency equivalent network with self-closing 20-min rated doors.
quired from the fire tests performed on building elements than we now obtain from standard tests. At this time, the furnace fire conditions do not
necessarily represent the positive pressure and excess fuel conditions of the
ventilation-controlled phase of a fire, which thereby decreases the meaningfulness of the test results. The end-point criteria are features of the standard test which are not rationally based, and many of the tests terminate
before these criteria are reached. It is our recommendation that new test
standards be established which are rationally based with regards to the fire
exposure and the criteria, and which at the same time ensure that passing the
new standard would be equivalent to passing the old standard. Such new test
standards are essential to the development of more sophisticated techniques
for risk analysis.
Acknowledgments
The authors gratefully acknowledge the discussion and comments of Professor Richard Barlow, Fred Fisher, Joe Zicherman, and Cecile Grant.
In addition, the authors wish to thank Claire Johnson and Ann Armijo for
their assistance in preparing this manuscript.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
58
FIRE RISK ASSESSMENT
This work was sponsored by the National Bureau of Standards, Center for
Fire Research, and the Nuclear Regulatory Commission through Sandia
Laboratories, Albuquerque, N.M., at the Lawrence Berkeley Laboratory
which is supported by the U.S. Department of Energy under Contract
W-7405-ENG-48.
References
[/] Harmathy, T. Z., Wood and Fiber. Vol. 1, 1977, pp. 127-144, and DRB Paper No. 805,
National Research Council Canada 17038, Ottawa, Canada.
[2] Friedman, R., Fire and Materials. Vol. 2, No. 1, Jan. 1978, pp. 27-33.
[3] Williamson, R. B., Fire Safety Journal. Vol. 3, No. 4, 1981, pp. 243-259.
[4] Connelly, E. M. and Swartz, J. A., "An Approach to Quantitative Measurement of Fire
Safety," presented at the Operations Research Society of America/Institute of Management Science (ORSA/TIMS) Conference, Miami, Fla., Nov. 1976.
[5] Connelly, E. M. and Swartz, J. A., "State Transition Model of Fire Growth Presented in a
Residence," presented at the 81st Annual Meeting of the National Fire Protection Association, May 1977.
[6] Berlin, G. N., Fire Safety Journal, Vol. 2, No. 3, March 1980, pp. 181-189.
[7] Williamson, R. B., Fire Performance Under Full Scale Fire Test Conditions—A State
Transition Model. 16th International Symposium on Combustion, Combustion Institute,
Pittsburgh, Pa., 1976, pp. 1357-1371.
[8] National Fire Protection Association, Decision Tree, Committee on Systems Concepts for
Fire Protection Structures, Boston, Mass., 1980.
[9] Lie, T. T., Fire and Buildings. Applied Science Publishers, Ltd., England, 1972.
[10] Dusing, J. W. A., Buchanan, A. H., and Elms, D. G., "Fire Safety Design Development
of an Analytic Technique," Civil Engineering Department Research Report No. 78-4, University of Canterbury, Christchurch, New Zealand, March 1978.
[//] Mirchandani, P. B., Computations and Operations Research. Vol. 3, Pergamon Press,
New York, 1976, pp. 347-355.
[12] Williamson, R. B., Grant, C , Zicherman, J., Fisher, F., and Hasegawa, H., and National
Institute of Building Sciences, Guideline on Fire Ratings of Archaic Materials & Assemblies, Vol. 8, HUD-PDR-613-8, U.S. Department of Housing and Urban Development,
Oct. 1980.
[13] Van Volkinburg, D., Williamson, R. B., Fisher, F. L., and Hasegawa, H., "Toward a
Standard Ignition Source," Paper No. 78-65, Fall Meeting of the Western States Section,
Combustion Institute, Laguna Beach, Calif., 1978.
[14] Quintiere, J. G., "A Characterization and Analysis of NBS Corridor Fire Experiments in
Order to Evaluate the Behavior and Performance of Floor Covering Materials," NBSIR
75-691-1, National Bureau of Standards, Center for Fire Research, Washington, D.C.,
June 1975.
[15] Sloan, N. J., Bell System Technical Journal, Vol. 51, No. 2, Feb. 1972, pp. 371-390.
[16] Fung, F. C , Suchomel, M. R., and Oglesby, P. L., Fire Journal. Vol. 67, No. 3, May
1973, pp. 41-48.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
V. M. Brannigan} and Rachel Dardis^
Legal and Economic Criteria for
Test-Based Fire Risk Assessnnent
REFERENCE: Brannigan, V. M. and Dardis, Rachel, "Legal and Economic Criteria for
Test-Based Fire Risk Assessment," Fire Risk Assessment. ASTM STP 762, G. T.
Castino and T. Z. Harmathy, Eds., American Society for Testing and Materials, 1982,
pp. 59-74.
ABSTRACT: This paper defines test-based fire risk assessment as it is used for economic
and legal analysis. Its economic use is illustrated by cost-benefit and risk-benefit techniques. Its legal use is discussed in the context of tort liability, contracts, and regulation.
In both areas, there is a need for definition of the risks being measured, the confidence in
the result, and the type of assessment and data being produced.
KEY WORDS: legal, economic, risk-benefit analysis, cost-benefit analysis, tort liability, contracts, regulation, fire risk assessment
The purpose of this paper is to give the test designer in test-based fire risk
assessment an understanding of the legal and economic use of his output, so
that he can incorporate criteria based on those used in economic and legal
analysis into the test design.
It is the hypothesis of this paper that test-based risk assessment follows a
general methodology. For our purpose, the elements may be defined as
follows:
Test—A technique for measuring a physical phenomenon.
Standaid—The acceptance criteria for a physical phenomenon.
Risk assessment—A determination of the probability of an occurrence and
the consequence (technical, economic, legal) of the occurrences.
Risk response strategy—The determination of choices available to an individual faced with risks.
It should be noted that these are independent elements of the methodology. Each term may, to a certain extent, be developed independently of the
'Assistant professor and professor, respectively. Department of Textiles and Consumer
Economics, University of Maryland, College Park, Md. 20742.
59
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
60
FIRE RISK ASSESSMENT
others. Economics and law play a significant part in risk assessment, because
the legal system defines the rights of those affected by a risk, and economics
predicts and defines their response to the risk and its consequences.
We shall assume that each individual faced with a risk assesses the risk in
terms of the consequences to him, or to those to whom he owes a legal obligation, either by contract, tort liability, or his responsibility as a public official.
This "actor's" view must be distinguished from a "global" view, which
assesses risk from the perspective of society as a whole. This is a more
"realistic" view than a global perspective, in which the total loss, as opposed
to the allocation of loss, is the only critical factor. This concept of an actor's
view emphasizes the role of legal and economic analysis. One of the obligations of the legal system is to apportion property rights which define whose
interest is at stake in a situation involving risks. In the real world, actors are
assumed to maximize their own perceived "best interest." Economics
describes how this is done.
As an example, consider the problem of a laundry room fire in a garden
apartment. Assume that a tenant negligently starts a fire, and, though the
fire is contained to the room, the smoke damages the contents of several
apartments. If the landlord has no responsibility for that damage, then it is
not part of his "risk" and therefore not part of the risk response strategy of
that actor. A test-based risk assessment which ignores smoke damage would
be perfectly acceptable to the landlord.
This direct impact on the test design of the legal and economic environment of the actor forces us to classify the consequences of a risk into factors
affecting the test designer.
Primary factors are those which have legal or economic effects and are intimately connected with technical factors which the test designer must either
include or exclude from his test. These factors include smoke, heat, toxic
gases, and structural stability. Exclusion of a primary factor from a test
makes it difficult or impossible for the risk assessor to determine the impact
of that factor on the risk.
Secondary factors are independent of the test design. They include insurance, risk aversion, cost sharing, litigation, and so on. They focus on the
existing social response to loss and are independent of specific physical
phenomena. They include a "standard," which, as an accepted or imposed
acceptance criterion, represents the chosen risk response strategy and allocation of responsibility.
The purpose of this division of factor type is to underscore the problem of
risk assessment with partial knowledge. If the test designer, for whatever
reason, has ignored or downgraded a primary factor, then each actor is faced
with a different confidence in the test result, since he has no way of knowing
whether the omitted factor seriously affects him. The secondary factors are
outside the realm of the test designer and can be used as part of each actor's
own risk response strategy.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
61
There is often no clear-cut, broadly accepted classification of primary and
secondary factors, since they depend on each actor's interpretation of his own
risk.
In this paper we are assuming that risk assessment is being done for a purpose, that is, to decide what, if anything, to do in response to the risk. Risk
response strategies are of many types; for our purposes it is sufficient to note
that they involve a mix of technical, economic, and legal considerations.
The legal and economic concepts are useful in specifying the type of output
produced by the test-based risk assessment. The economic analyst uses the
output in making quantitative judgments that involve the risk response
strategy.This paper will focus on cost-benefit analysis as an example of the
economist's use of test-based risk assessment.
The legal system, on the other hand, uses qualitative concepts in assigning
rights and responsibilities. Legal concepts of negligence and assumption of
risk and reasonable risk rarely turn on accurate numerical assessment of
risk.
Economic Analysis
This section of the paper will discuss the role of economic analysis with
regard to the assessment of risk and risk response strategies. Data requirements for such analyses are presented in order to demonstrate the role
of test-based fire risk assessment in the implementation and evaluation of
risk response strategies.
Risk Assessment
The contribution of economic analysis to risk assessment is the measurement of consequences in dollar terms. The use of a common unit of measurement has two advantages: (1) it permits comparison of risks for different
systems and structures, and (2) it permits comparison of the benefits from
risk reduction, since risk assessment serves as a basis for estimating the
potential benefits from risk reduction.
The economic consequences of an occurrence may be classified as direct or
indirect. Direct losses include structural damage and medical costs associated with injuries and deaths. Indirect losses include output losses due to
disability, losses incurred by business or industry due to business interruptions, and losses to the community from the interruption of services. In
general, direct losses are easier to measure than indirect losses. In particular,
there has been considerable discussion concerning the measurement of the
loss from death. This debate has been tempered in recent years by the
realization that all lives must eventually be lost—we cannot postpone death
indefinitely. Thus, the loss of life years as opposed to the loss of life is really
the outcome which should be measured.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
62
FIRE RISK ASSESSMENT
Regardless of whether all loss components are assigned dollar values, the
resulting risk assessment should provide information to the policy maker
concerning the need for a risk response strategy. This is important since
strategies cannot be undertaken for all hazards because of resource constraints.
Assessment of Risk Response Strategies
A variety of strategies may be considered, including education, expanded
use of fire services, and standards or codes based on fire tests. In all instances, economic analyses may be used to evaluate the impact of the various
strategies.
Two major techniques which may be used to assess risk response strategies
are cost-effectiveness analysis and cost-benefit analysis. These techniques
have been discussed extensively in the literature [1-6].^ In cost-effectiveness
analysis, the desired goal is first established, for example, a 50 percent
reduction in fire losses. The costs of various strategies for achieving this
reduction are then compared. Costs are expressed in dollar terms, while
goals or objectives are expressed in nondollar terms, such as the number of
fire incidents or the number of fire-related injuries or deaths.
In cost-benefit analysis, the costs of implementing a particular strategy
are compared with the ensuing benefits. In contrast to cost-effectiveness
analysis, benefits are expressed in dollar terms, thus ensuring that both program costs and benefits are measured in the same units. This is important in
the case of strategies with multiple outcomes, such as the lives saved, injuries
avoided, structural damage averted, because some method for aggregating
the various outcomes must be used. The degree of protection provided by a
particular strategy plays a crucial role in both cost-effectiveness analysis and
cost-benefit analysis. It is a function of costs and also affects benefits, which
are based on the percentage reduction in existing losses.
Cost-effectiveness analysis provides the decision maker with less information than cost-benefit analysis. While it is useful in ranking strategies according to their effectiveness, it does not address the issue of whether or not a particular strategy should be attempted. For example, a comparison of Strategy
X with Strategy Y may result in the finding that the former provides the same
protection at a lower cost. However, Strategy X may be undesirable when
both costs and benefits are considered.
A comparison of the two evaluation techniques is shown in Fig. 1. The
degree of protection (or risk reduction) provided by a strategy is given on the
horizontal axis, while the additional costs and benefits of the strategy are
given on the vertical axis. There are two strategies, X and Y, and the two
curves, MC(X) and MC(Y), correspond to these strategies. The shapes of the
^The italic numbers in brackets refer to the list of references appended to this paper.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
MC(Y)
Marginal
Costs
63
KC(X)
Marginal
Benefits
0
0.45
0.5
Degree of
FIG. 1—Comparison of cost-effectiveness
1.0
Protection
and cost-benefit
analysis.
marginal cost curves reflect decreasing returns to risk reduction, that is, it
becomes increasingly costly to achieve higher degrees of protection because
of technological and other constraints. In contrast, the marginal benefit
curve is relatively unchanged since it is assumed that the benefits from risk
reduction are in proportion to the degree of risk reduction.
In cost-effectiveness analysis, consideration is only given to achieving the
desired objective at the lowest possible cost. According to the diagram,
Strategy Y is the preferred strategy as long as the desired degree of protection
is less than 50 percent. Beyond this point, Strategy X should be used since its
marginal costs are lower. Cost-effectiveness analysis thus permits the attainment of any degree of protection, provided the lowest cost method is used. In
contrast, cost-benefit analysis requires a comparison of costs and benefits, so
that Point E on the diagram represents the optimal solution. The decision to
increase the degree of protection beyond 45 percent results in a net loss since
the additional costs are greater than the additional benefits. For similar
reasons, a degree of protection less than 45 percent, is also nonoptimal. The
determination of a given objective (in this case, the degree of protection)
without considering both costs and benefits is a major weakness in costeffectiveness analysis and limits its usefulness.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
64
FIRE RISK ASSESSMENT
Major issues of importance in cost-benefit analysis are the following:
(a) identification of relevant costs and benefits,
(b) measurement of such costs and benefits,
(c) selection of a discount rate to measure future costs and benefits, and
id) determination of the appropriate criteria for comparing costs and
benefits.
Benefits have already been discussed under risk assessment, where it was
noted that there are problems in connection with the measurement of the loss
from death—that is, the value of a life. This problem may be resolved in
cost-benefit analysis by estimating the value of a life implicit in the decision
to adopt various risk-reduction strategies. In this procedure the potential
benefits are assigned dollar values, with the exception of the lives saved. The
difference between the total benefits and the total costs divided by the
number of lives saved provides an implicit value of a life. The strategies may
then be compared on the basis of implicit life values, with the preferred
strategy being the one that possesses the lowest value of a life. Thus, in the
case of a proposed standard for upholstered furniture, implicit life values
ranging from $611 000 to $1 528 000 were obtained. Since considerably
lower values were obtained for smoke detectors, the latter strategy was
preferred.
The costs of any strategy are basically the resources entailed in implementing the strategy. They may include costs of labor, raw materials, and other
inputs needed to implement the strategy. These direct costs are easier to
measure than the indirect costs, which include such factors as the reduction
in consumer choice (that is, a particular product may no longer be available
to the consumer) and reduced competition (that is, some firms may not be in
a position to comply with the proposed strategy and may exit). The fact that
the strategy may be a voluntary as opposed to mandatory standard does not
necessarily prevent this situation, since widespread adoption of a voluntary
standard by most firms and product users could reduce the potential sales for
a noncomplying firm. A final indirect cost is the impact on innovation; that
is, is the proposed strategy likely to encourage or stimulate the adoption of
new technology or is it likely to perpetuate the status quo?
Identification of the costs and benefits also serves to indicate whether they
are borne by the same parties. For example, the costs of the proposed standard for upholstered furniture would be incurred by all purchasing
households, whereas the primary benefits would accrue to those households
with careless smokers. Manufacturers also complained that their business
would be adversely affected. As a result, the proposed standard has still not
been implemented. It is important, therefore, to consider the differential
distribution of costs and benefits in determining its acceptance. If some
groups are adversely affected, then a particular strategy may be protested or
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
65
may become ineffective from lack of support. Cost-benefit analysis can play
a role in providing information concerning the economic impact of a strategy
on various groups and hence its potential acceptance.
The two remaining items are the discount rate and the appropriate
cost-benefit criteria. In general, the present value of net benefits is used to
evaluate different strategies. This is given by PVB minus PVC where
PVB = fi,/(l -I- i) + ^2/(1 + iy + ••• BJ{1
+ iT
PVC = C,/(l + i) + C2/(l + i? + ... C„/(l + /)"
and
5,
Q
m
n
/
=
=
=
=
=
benefits accrued in Year t, t = I, 2, ... m,
costs incurred in Year t, t = I, 2, ... n,
time period over which benefits accrue,
time period over which costs accrue, and
discount rate.
As these terms indicate, the importance of the discount rate pertains to the
manner in which benefits and costs accrue. If they are distributed equally
over the life of the project, then the particular discount rate selected will have
little impact on the resulting evaluation. However, most benefits and costs
occur at different points in time. If the costs of constructing a fire-resistant
system or structure are incurred in the first year, whereas the benefits are
distributed over a 20-year period, then the discount rate can have a significant effect. The higher the discount rate, the lower the present value of
future benefits and the greater the possibility that the project will be rejected.
Data Requirements
Major data requirements for the analysis of any risk response strategy include the following:
(a) expected losses due to an existing hazard,
(jb) degree of risk reduction achievable by various risk response strategies,
and
(c) costs of various risk-reduction strategies.
Information concerning the first component may be obtained from past
data, including frequency of occurrence and severity of fire incidents.
Economic analysis may then be applied to the different outcomes, such as
structural damage, deaths, and injuries, to estimate the expected dollar
losses from the hazard.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
66
FIRE RISK ASSESSMENT
The second and third components are dependent on the particular risk
response strategy. In the case of codes and standards, test-based fire risk
assessment may be used to obtain estimates concerning the degree of risk
reduction and the associated costs from the implementation of codes and
standards. A primary consideration in the case of test data is the ability to
translate the results of laboratory tests into "real-life" situations. The decision maker needs to know the consequences of various risk response
strategies in order to select the optimal strategy. Unfortunately, test data
which may be satisfactory to the engineer are not necessarily of the type
useful to the policy maker. This, in turn, limits the use of test-based fire risk
assessment in developing appropriate risk response strategies.
In the following discussion, consideration is given to the different types of
information which may be used by the analyst in evaluating various risk
response strategies. Strategies in this instance are confined to codes and standards, and it is assumed that cost data may be estimated based on test requirements and existing technology. Information concerning the degree of
protection for the various strategies is then the significant factor. Strategies
may be classified as follows: (1) strategies with an unknown degree of protection, (2) strategies with similar degrees of protection, (3) strategies with
qualitative differences in the degree of protection, and (4) strategies with
quantitative differences in the degree of protection. Each possibility is
discussed in the following paragraphs.
1. Strategies with unknown degrees of protection. This information is insufficient to compare strategies. However, it is possible to use estimates of existing losses to determine the maximum benefits from a strategy. If the costs
of the strategy exceed the maximum benefits, then the strategy is inefficient.
This approach requires quantification of benefits in dollar terms.
2. Strategies with similar degrees of protection. This information may be
combined with cost data to compare the cost-effectiveness of various
strategies. However, as noted earlier, cost-effectiveness analysis does not address the issue of whether a strategy should be adopted. This decision requires a comparison of costs and benefits which may be estimated from existing losses and the degree of protection provided by the strategy.
3. Strategies with qualitative differences in the degree of protection. In
this instance, strategies are ranked according to the degree of protection. For
example. Strategy X is more effective than Strategy Y, which is, in turn,
more effective than Strategy Z. The costs of the strategies increase with their
effectiveness. However, the differences between the strategies are not quantified. This presents a problem, since we do not know if the additional costs
from the adoption of Strategy X are justified. One possible solution is to
estimate the additional benefits, such as the degree of risk reduction which
would be required to justify the additional costs of Strategy X. The required
risk reduction could then be discussed with experts in the field to determine
whether such a reduction were possible. It should be noted that the decision
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
67
to adopt the superior strategy (Strategy X) is not justified unless there is
reason to believe that the additional benefits will exceed the additional costs.
4. Strategies with quantitative differences in the degree of protection. Cardinal, as opposed to ordinal, measures are provided in this instance for the
degree of protection. Determination of the optimal strategy is a simple matter since it is only necessary to compare the net benefits from the different
strategies. Provision for uncertainty concerning the degree of protection provided by the various strategies may also be made. Thus a range of values for
the degree of protection (such as 10 to 20 percent) may be used in place of a
single-point estimate.
Discussion
The preceding section indicates the need for qualitative and quantitative
data in assessing risk response strategies. Economic assessment will, in turn,
facilitate decisions concerning risk response strategies. However, economic
analysis is dependent on the availability of data, and test-based fire risk
assessment may not always provide the required information. It is important
for test designers to consider the needs of potential users and to recognize the
role of test data in increasing the effectiveness of the entire fire safety effort.
Legal Analysis
The legal use of test-based risk assessment includes data requirements for
the legal process and use of risk assessment in the legal process. These are
discussed in the following sections.
Data Requirements for the Legal Process
The type of data produced by a test or a risk assessment can be classified
as nominal, ordinal, interval, or ratio.
Nominal measures of risk assign a name to a particular risk. They simplify
a regulatory program since they require no further analysis. Such a test
method or standard gives us no way of knowing whether they are "high" or
"low" risks, only that the risk exists. Such measures are particularly popular
for regulatory purposes since they make determinations very easy. The
Delaney Clause of the Food and Drug Act [21USC348(c)(3)(A)] is of this
type. It gives the Food and Drug Administration no discretion: if a substance
causes cancer, it must be banned. In the "real" world such an insistence
would lead to strange results, such as the nineteenth-century case where a
buyer was allowed to reject a house because the pipe used was identical to
that specified by another brand. Such a result is no longer acceptable in most
contract law.
Ordinal measures of risk rank multiple risks in terms of various factors,
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
68
FIRE RISK ASSESSMENT
including their likelihood of occurrence and the consequences. They are
useful from a regulatory perspective. The Consumer Product Safety Commission (CPSC) was faced with a choice of whether swimming pool slides should
be located at the shallow end of the pool, with a risk of accidents resulting in
paralysis, or at the deep end, with risk of drowning [7]. An ordinal ranking of
the risk would be of great help to the regulator, even if no one else can use the
information. We commonly accept ordinal assessment in areas such as meat
and produce grading. Pass/fail flammability tests are ordinal. The pill test
for carpet (FF 1-70) is of this type. A carpet sample either passes or fails. The
test method gives no other information. The standard based on the test cannot reconcile two test results, one passing and one failing, on the same roll of
carpet. In fact, this standard excludes some of the information available in
the test, because it does not define a "passing" carpet but only a "failing"
carpet. The CPSC enforcement personnel claimed the right to continue
testing the carpet until they achieve a failing result. The same problem would
occur in any risk assessment based on the test. How do you cope with
conflicting test results [8]?
Cardinal (interval) and ratio data can be manipulated mathematically.
Clearly, ratio results are more useful than nominal or ordinal results.
However, providing such data is both difficult and expensive, and it may not
be possible.
It is crucial that the test results indicate what type of data are produced.
A second data requirement pertains to confidence in the assessment conclusions. How confident are we that the result represents the true state of the
world? The risk assessment may be expressed in terms of a statistical confidence interval; a point plus or minus a certain amount; a good estimate
provided assumptions A, B, and C are true; or the best judgment of a consensus of experts.
Particularly in the regulatory use of risk assessment, an honest explanation
of the level of confidence in the risk assessment and the test methodology is
crucial.
A test is robust when we can alter some of the test conditions and still have
some confidence in the test result. It is up to the test designer to determine
how robust the test is and to communicate that to the users. For example,
can the pill test for carpet be performed on a used carpet with any confidence
that it is still viable [8]? The test method does not answer the question. This
is particularly important in the regulatory use of test-based risk assessment.
Finally, the type of risk assessment must be considered. Fire is a probabilistic phenomenon. However, risk assessment deals with two separate
types of assessment. They can be defined as actuarial or projection risk
assessments. In actuarial assessments, the key measurement is the rate of occurrence of an event in the past. This rate is then used to estimate the future,
with the assumption that the future is similar to the past. Actuarial probabilities are often estimated from test data. Projection is used where there is
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
69
no historical record from which to calculate occurrences. In this assessment,
probabilities are based on defined sets of assumptions and data, which are
then manipulated according to a theory to produce the projection. Tests provide the data for manipulation.
Use of Risk Assessment in the Legal Process
Risk assessment may be used as "evidence" either in court proceedings or
in administrative hearings.
It should be noted that each type of proceeding has "rules" of evidence
which define whether the proceeding will accept the evidence. In an area
such as risk assessments based on tests, the usual problem is that courts need
to be convinced that the evidence is more than speculation. While actuarial
methods are accepted rather easily since they are based on past history, projection methods may prove troublesome. The court may insist on both substantial documentation of the risk assessment methodology and assurance
that there is a scientific discipline of test-based risk assessment.
These legal requirements clearly pose some problems for test-based risk
assessment. The determination of what factors are to be measured, the confidence in the test result, and the robustness of the test will all require
substantial documentation. Presumably, in the case of consensus standards
the documentation must be approved in the same way as the assessment. The
three major legal areas which involve risk assessment are tort liability, contracts, and regulation.
Tort Liability—The use of test-based risk assessment in tort liability is normally retrospective. Therefore, we are not dealing with a "risk," as such, but
an event, usually an injury. The role of tort liability is to determine who will
bear the assigned financial loss from injury. The ability of that party to pass
on the cost through the marketplace is beyond the scope of this paper. In
making this determination, courts find "facts" and apply the appropriate
legal standard. The court is often faced with a situation of a "reverse" crystal
ball. It must look into the past and determine what the various parties knew
or should have known at some past time. The court then determines whether
the response to the risk, that is, the risk assessment strategy, was that of a
"reasonable man."
The tort liability system intends to provide a remedy for all legal wrongs.
Therefore, the court will make its conclusions concerning the risk response
strategy even if no formal risk assessment has been made. Obviously, those
trying to convince a court that their actions are those of a "reasonable man"
would be aided by a proper risk assessment. However, an inadequate risk
assessment, particularly one which fails to cover a primary factor, can be a
major element in assessing liabilities. For example, in the Aretz case, there
was a disaster in a Thiokol Chemical Corp. plant which assembles illuminants into night flares [9]. Twenty-nine people were killed and 150 injured by
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
70
FIRE RISK ASSESSMENT
the rapid spread of fire in the plant and an ensuing explosion. The risk involved was assessed by a test method and standard which defined the illuminants as subject to "deflagration" rather than "detonation." Dissatisfaction with the test method led to a new risk assessment by the government,
which determined that the illuminant should be treated as a detonation
hazard. Unfortunately, this information was not transmitted to the company. In the ensuing litigation, the government was held liable, since it had
not acted as a "reasonable man" in conveying risk assessment and response
strategy to the contractor. What is particularly interesting about this case is
that the primary question was, "Does the test method accurately assess the
risk?" Before 1967, the illuminant was a Class 7 detonation hazard. It was
reclassified by the "card gap test" as a Class 2 deflagration hazard. In 1968,
two independent researchers reported that the card gap test was inadequate.
The U.S. Army confirmed those results and on 16 Nov. 1970 reclassified
the illuminant as a Class 7 hazard. The information should have been immediately transmitted to Thiokol, but it was not. The explosion occurred on
3 Feb. 1971. In particular, a Class 2 hazard was defined as
. . . items which bum vigorously with little or no possibility of extinguishment
in storage situations. . . . Explosions normally will be confined to pressure
ruptures of containers and will not produce propagating shock waves or damaging blast pressure beyond the magazine distances specified by this class. There
may be a severe hazard of the spread of fire from tossing about container
materials. Toxic effects normally will not exist beyond the inhibited building
distances specified for this class.
This classification is distinguished for the purposes of this case from Class 7
hazards, which include
items . . . most of the entire quantity of which will explode virtually instantaneously when a small portion is subject to fire, to severe concussion or impact,
to the impulse of an initiating agent, or to the effect of a considerable discharge
of energy from without. Such an explosion normally will cause severe structural
damage to adjacent objects and the simultaneous explosion of other separated
explosives and ammunition placed sufficiently close to the initially-exploding
pile.
Since the government decided that the illuminant was a Class 7 hazard and
had been negligent in transmitting that information, and since the victims
were injured outside the building, liability followed almost automatically.-'
Certain peculiarities in the tort liability of the federal government made it
crucial that the government actually had accepted the charge in risk assessment. A private defendant would be liable if a "reasonable man" would have
reclassified the material.
^This opinion was vacated on 15 April 1980, 616 F.2d 254 (5th Cir. 1980) (en Banc) and certified to the Supreme Court of Georgia for determination of the duty of the federal government
to classify the material correctly. However, all findings of fact were accepted as supported by the
evidence. Aretz v. U.S. 635 F.2d 485 (1981) (5th Circuit) (en Banc).
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
71
For tort purposes, the difference between the primary and secondary factors is vital. A decision that it is cheaper to insure against a risk than to prevent it can virtually guarantee liability. Failure to include a primary factor in
the test can make the producer liable if the inclusion of the factor would have
resulted in rejection of the product.
Certain developing tort theories impose liability for defective products even
if the manufacturer could not have prevented the defect. The term "defect"
is so flexible that injured parties can now introduce their own test-based risk
assessment to illustrate the "defects" in a product.
To sum up for tort purposes, the issues are (1) whether a formal test-based
assessment should be done, (2) what happens if it is done badly, (3) what
happens if it alerts those responsible to a hazard but there is no response, (4)
what happens if the risk response strategy concentrates on secondary factors
rather than primary factors, and (5) possible use of the assessment by injured
plaintiffs.
Contracts—Tort law is public law, that is, everyone is bound by the same
principles of law. Contract law is mostly private law, particularly between
businesses. The use of tests and standards in contract law is far greater than
the use of risk assessment. There are exceptions, however, including (1)
where the risk assessment is the subject of the contract, (2) where a contract
is for the reduction of control of the risk, and the risk assessment is the tool
for measurement of success, and (3) where the contract is one of insurance
and the risk assessment provides the information necessary for the contract.
In these types of contracts, the question is usually "Does the risk assessment
promise a result, or is it merely a tool, to be accepted with its limitations?" In
the first case, where the assessment is the subject of the contract, the accuracy, confidence, and so on of the methodology are of little importance.
The robustness of the test is significant because it can help us determine
when the contract has been substantially performed. Documentation of the
methodology is also vital. In the second case, where the risk is being
measured, the confidence interval of the test is vital. In addition, the type of
risk assessment, that is, the actuarial assessment risk or projection assessment, is important since actuarial risk assessment requires substantial time
before it can effectively measure a change in risk. Significant liability may attach before the change in risk is known. Risk assessment for insurance purposes (the third case) is normally an internal rate-making device. Here the
question is usually "Are there any primary factors affecting our risk which we
have not included in the assessment?" In insurance, as in tort law, a severely
flawed risk assessment may still be valuable, if it is sufficiently robust; that
is, its effectiveness at measuring the risk may be poor, but as long as it is better than nothing the system will use it.
Regulation—The regulatory use of risk assessment poses several special
problems. The government is rarely under a legal requirement to regulate.
When it does regulate, it has the choice of using legislation or delegating the
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
72
FIRE RISK ASSESSMENT
authority to an administrative agency. Administrative agencies often have a
problem. In theory, they should represent the "public interest" in their
assessment, but in practice they may feel a responsibility to protect one portion of the public in preference to others. In addition, they include their own
institutional problems in the risk response, that is, their ability to enforce the
regulation.
These agencies may also lack jurisdiction over a primary factor, so they exclude that factor from their risk assessment and response strategy. The CPSC
might prefer to regulate cigarettes rather than upholstered furniture, but
while they have jurisdiction over furniture, they have none over cigarettes.
An agency may make a risk assessment and decide on a risk response
strategy, but its primary method of regulating is to promulgate a standard.
Obviously, it would be most convenient if the test used in the risk assessment
was suitable for enforcement purposes. That may not be the case. For example, if the test method has a significant variation in results due to test variation, it may be suitable for risk assessment if a large number of tests can be
performed, but for standard enforcement, such a variation would make use
of the test either difficult or irrational if a single test failure was considered
indicative of the hazard proscribed by the standard. For example, in the
carpet standard, eight specimens are tested and a 7 to 1 result is a pass, a 6 to
2 result a failure. The purpose of permitting occasional failures is to allow for
test variation. Imposing substantial penalties based on a test which has such
variation raises due process questions. If the "false negative" occurs with
some regularity, what is the legal impact if the government conducts a
number of tests, until it gets a 6 to 2 result by chance? The robustness of a
test must also be documented if it is to be used in enforcement, since the
agency may need to test under adverse conditions, such as after a fire.
In a hearing on a mattress standard, a respondent argued that the enforcement practice of lighting all 15 cigarettes at the same time failed to test the
mattress in accordance with the standard [10]. The CPSC argued that the
provision of the standard requiring one cigarette at a time was to protect the
operator and did not affect the test. Such a failure to document the test protocol can render the standard useless.
Some agencies have argued that risk assessments are so imprecise that they
should not have to do them. The Supreme Court decided in the Benzene case
[11] that the Occupational Safety and Health Administration must find as a
"fact," supported by substantial evidence, that it is more likely than not that
10 ppm of benzene will cause injury. This requirement of at least a
"nominal" assessment supported by evidence will put new burdens on risk
assessment to meet standards defined by the courts rather than by peers.
The Ford Motor Company was criminally prosecuted for an inadequate
response to a test-based fire risk assessment. While this is the first such prosecution, it indicates the depth and seriousness of the legal system's interest in
risk assessment.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
BRANNIGAN AND DARDIS ON LEGAL AND ECONOMIC CRITERIA
73
Conclusion
If test-based fire risk assessment is to mature as a tool for industrial and
public application, it is vital that the test designers cope with the user's need
for confidence, robustness, and thorough documentation of the test. In addition, its use requires an acceptance on the part of test designers that fire is a
special design criterion. All unfriendly fires should be considered hazards.
Tests which have no value other than as laboratory research tools can divert
test designers from the problem of fire risk. There is no reason to measure
fire performance except in terms of risk. Statements such as
This standard should be used solely to measure and describe the properties of
materials, products, or systems in response to heat and flame under controlled
laboratory conditions and should not be considered or used for the description,
appraisal, or regulation of the fire hazard of materials, products, or systems
under actual fire conditions [12].
indicate that test designers prefer to do what they have always done rather
than accept the difficult challenge of risk assessment. Even the somewhat
less cautious statement.
This standard should be used to measure and describe the properties of
materials, products, or assemblies in response to heat and flame under controlled laboratory conditions and should not be used to describe or appraise the
fire hazard to materials, products, or assemblies under actual fire conditions.
However, results of this test may be used as elements of a fire risk assessment
which takes into account all of the factors which are pertinent to an assessment
of the fire hazard of a particular end use [13].
does not give us enough useful information since it does not explain what
"factors," if any, are assessed by the test itself. It is difficult to include the
"other factors" unless we know what this test tells us. These issues must be
addressed by involving users in the test design, not ignoring them with
disclaimers. It reminds one of the joke where the boy is searching the
sidewalk for his lost quarter rather than in the grass where he lost it, because
"I could see it on the sidewalk, I can't see anything in the grass."
References
[I] Anderson, L. G. and Settle, R. F., Benefit-Cost Analysis: A Practical Guide, Heath, Lexington, Mass., 1977.
12] Chase, S. B., Ed., Problems in Public Expenditure Analysis, The Brookings Institution,
Washington, D.C., 1968.
[3] Mishan, E. J., Cost-Benefit Analysis, 2nd ed., Praeger, New York, 1976.
[4] Prest, A. R. and Turvey, R., Economic Journal, VoL 75, Dec. 1965, pp. 683-735.
[5] Sassone, P. G. and Schaffer, W. A., Cost-Benefit Analysis: A Handbook, Academic Press,
New York, 1978.
[6] Thompson, M., Benefit-Cost Analysis for Program Evaluation, Sage Publications, Beverly
Hills, Calif., 1980.
[7] Aqua Slide & Dive vs. CPSC, 569 F.2d 831 (1978).
[8] CPSC vs. Barrett Carpet. Inc., Docket 75-5.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
74
FIRE RISK ASSESSMENT
[9] Aretz vs. United States, 604 F.2d 417 (1979).
[10] CPSC vs. Bay Area Mattress, Docket 75-2.
[11] Industrial Union Department AFL-CIO vs. American Petroleum Institute et al, U.S.
Supreme Court No. 79-911, 2 July 1980.
[12] ASTM Policy Defining Fire Hazard Standards, Limiting the Scope of Properties—Description Standards and Establishing a Committee on Fire Hazard Standards, adopted 18 Sept.
1973.
[13] ASTM Test for Flammability of Clothing Textiles [ANSI/ASTM D 1230—61 (1972)] 1979
Annual Book of ASTM Standards, Part 32 Textiles— Yarns. Fabrics, and General Test
Methods. American Society for Testing and Materials, Philadelphia, 1979.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
R. G. Gewain^
Assessment of Fire Risk for Exterior
Structural Members
REFERENCE: Gewain, R. G., "Asaessment of Fhe Risk for Exterior Structural Memben," Fire Risk Assessment. ASTM STP 762. G. T. Castino and T. Z. Harmathy, Eds.,
American Society for Testing and Materials, 1982, pp. 75-94.
ABSTRACT: The traditional fire exposure in fire tests used to evaluate the performance
of construction materials is the standard time-temperature curve specified in ASTM Fire
Tests of Building Construction and Materials (E 119-80). There is growing concern that
the severity of a fire defined by ASTM Method E 119-80 may not be representative of
most fires which occur in buildings such as offices, schools, hotels, apartments, and other
similar occupancies. Extensive research on postflashover conditions in building compartments supports this concern. However, the heating conditions in ASTM Method E
119-80 are not appropriate for exterior structural members. Therefore, research described in this paper provides a simple method for describing fire exposure of structural
members located outside the enclosing walls of buildings.
The following procedure establishes the intensity of the fire in the compartment under
ventilation-controlled or free-burning conditions, the shape and size of the flame emerging from openings in exterior walls of the fire compartment, and the temperature of the
flame, all of which are needed to calculate the heat transfer to structural members outlined by the American Iron and Steel Institute.
KEY WORDS: radiation, convection, emissivity, flame shape, fire load density, fire
severity, burning rate, fire temperatures, fire compartment geometry, ventilation, forced
draft, flame plume, buoyancy, horizontal flame projection, effective flame boundary,
natural draft, flame axis, temperature distribution, maximum temperature fire duration,
fire risk assessment
Fire exposure arises from interior and exterior origins. The evaluation of the exterior exposure can be done only with difficulty in quantitative terms, and the
gradual accumulation of data from actual fires will probably continue as the
main guidance in providing the proper protection.
A s this quote from S. H . Ingberg, t h e n senior engineer of t h e National
Bureau of Standards [1]^ implies, t h e d e v e l o p m e n t of fire test and fire protection requirements in building c o d e s , now d e t e r m i n e d by A S T M Fire T e s t s
'Chief fire protection engineer, American Iron and Steel Institute, Washington, D.C. 20036.
^The italic numbers in brackets refer to the list of references appended to this paper.
75
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
76
FIRE RISK ASSESSMENT
of Building Construction and Materials (E 119-80), focused attention on
structural elements located inside a building exposed to a fire inside the
building. This is still true today; however, engineering methods for both interior and exterior fire exposure are being developed which will place structural fire resistance on the same basis as seismic design and the design for
structural performance when exposed to external loads such as wind and
snow.
Building codes require fire-resistance ratings for specific portions of a
building. These ratings are based on a specific fire exposure period in accordance with ASTM Method E 119-80. Studies in the United States and
abroad clearly indicate that the fire exposure in building fires (Fig. 1) may
not be the same as the fire exposure in this standard. Little has been done to
update the codes except to modify the required fire endurance time for
special applications. Exterior fire exposure is not recognized, nor are the
variations in fire severity.
The conditions of the fire test raise serious questions as to whether proper
fire protection measures are being used in buildings. Structural building
members located in the interior of a building are surrounded by flames and
by heated surfaces (walls, ceiling, and floor) of the fire compartment much as
in the standard fire tests of ASTM Method E 119-80. These heating conditions are not representative of the exposure to exterior structural members.
Exterior structural members receive heat from two sources in a fire: radiation from nearby windows of the fire compartment and radiation and convection from flames and hot gases exiting from nearby windows. The value of
the intensity of the radiation received by the exterior member varies with the
relative position of the window and flame (radiator) and the surface of the
structural member (receiver). However, structural members may also lose
heat by convection and radiation to surroundings at normal ambient temperature. Therefore, depending on the size and location of the structural
member and the behavior of the fully developed fire, exterior structural
members may not need any fire protection.
Early attempts to assess the external fire exposure involved experiments
with ASTM Method E 119 furnace tests in which there was a window in the
furnace wall [2]. This effort showed that no matter how long the ASTM
Method E 119 fire test was conducted, the exterior exposure was generally
less severe than that from a more realistic building fire. As a result, there
have been full-scale experimental fire tests with exterior structural elements
exposed to flames and radiation from "real" fires to obtain code approvals
for specific buildings (Fig. 2) [3].
Recently, a design approach has been developed which will permit analysis
of the heat transfer to exterior structural members and will enable engineers
to calculate the amount of fire protection (if any) that may be needed.
For structural steel elements, the design is based on a limiting steel tern-
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
77
2400
2000
-
1600
1200
800
20
10
30
40
Time,Mtnute$
FIG. 1—Effect
Test
Number
Fire Load
(PSF)
Forced Air
(CFM)
1
2
3
4
10
10
10
10
0
0
0
0
Windows
Number Size, Ft X Ft Areo.Sq Ft
1
1
2
2
2X6
2X8
2X6
2X8
12
16
24
32
of window area on fire temperatures during burnout tests (natural airflow).
perature that will result when heat is transferred from the fire to exterior
structural members in specific locations. The calculation procedure is well
documented {4\.
The main problem is to define the external heat transfer. A large body of
data on building fire and flame behavior exists. It has been collected and
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
78
FIRE RISK ASSESSMENT
FIG. 2—One Liberty Plaza. New York City; the exterior surfaces of steel spandrel beams have
the web bare and the flanges flame shielded.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
79
analyzed by Margaret Law at Ove Arup and Partners, London, England, and
a designers' manual has been prepared to estimate external heat transfer for
practical designs in buildings [5]. Correlations have been derived, not only
from scale model experiments but also from measurements for a wide range
of experimental fires in large-scale building compartments.
The principal elements needed to make this design approach useful are the
following:
L Flame shapes are based on parameters which can be readily identified
by the designer.
2. Temperatures of the room fire and the external flame plume are determined by relatively simple equations.
3. Heat transfer equations are simplified to facilitate calculation.
Some of the symbols that will be used in this paper are defined in Table L
Research
It has been generally accepted that most structural damage from fire exposure occurs in the postflashover period. Therefore, this phase of the fire
has been studied by many researchers. Dr. Philip Thomas [6] has summarized this work, and important parameters involving the way a fully developed
fire bums have been defined by Margaret Law, who has used the large
amount of test data available to show how these parameters interact.
Ingberg was the first to study the relationships between fire load density
and fire severity, and these studies have formed the basis for building code
requirements for the fire resistance of structures. Later, Fujita [7] quantified
the effect of ventilation in terms of the area and height of ventilation openings (usually windows). Later work on models was carried out cooperatively
by the Conseil International du Batiment (CIB) [8], which modified the Fujita relationship on ventilation by introducing the size and shape of the fire
compartment (Fig. 3). In addition to using scale models in the CIB program,
the various factors were studied in full-scale studies at Borehamwood, England [9], Maisieres-les-Metz, France [10], and Carteret, N.J. [77].
Interior Fire Behavior
A number of full-scale experiments have been carried out, using wood
cribs as fuel, in still or lightly moving air, although some tests were conducted with additional air supply.
Two of the most important factors affecting exterior fire exposure from a
fully developed interior fire are the rate of burning, which affects the flame
size and the fire duration, and the fire temperature, which affects the rate of
heat transfer.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
80
FIRE RISK ASSESSMENT
TABLE 1—Explanation of symbols.
Units
Symbol
Aj
^w
D
Meaning
total room surface area = lA^ + 2H(D X W)
window area
depth of compartment
E
<^»^"-J WH
e
H
h
base of natural logs
height of compartment
height of window
J
I
I
Q
R
T
u
W
w
X
X
1
P
T
'•p
f-o.z
English
Metric
ft2
ft2
ft
m2
m2
m
Ib/min ft^'^
kg/s m'^^
ft
ft
m
m
j ^
ft/min^
m/s^
h
ft
m
lb
ft
kg
m
Ib/min/ft
kg/s/m
Ib/min
R
ft/min
ft
ft
ft
kg/s
K
m/s
m
m
m
ft
ft
Ib/ft^
min
min
m
m
kg/m^
s
s
fite load
distance along flame centerline from window
R
rate of burning
absolute temperature (0°F = 460°R)
velocity
width of compartment
width of window
centerline distance of flame tip from window
horizontal distance of center of flame tip from
window
vertical distance of flame above top of window
density
fire duration
free-burning fire duration
these subscripts denote fire, window plane, and
flame, respectively
Rate of Burning
In the case of natural draft conditions where the fire behavior and compartment dimensions control the air flow, continuous weighing of the fire
load during tests has shown that the rate of weight loss is fairly steady over
the fully developed fire period when the weight of the fire load falls from 80 to
30 percent of its initial value. This rate of burning, R, is defined as
R =
(1)
where T is the effective fire duration and L is the total fire load in the fire
compartment.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
R
81
[DT'"
tb/min.H'l'
+n
Symbol
•
W
U
10
IS
20
w
H
D
H
1
2
2
4
2
2
1
4
2S
n - A,/A,hV> - tr»»
Lani«-«cal« data
Q Borahamwood
-t- Mall
C Cartarat
FIG. 3—Variations of R/A„h'''^ with compartment size and ventilation, as given by Thomas
for CIB data.
For a free-burning condition, the value of T is Tp and is determined by the
characteristics of the fire load—thin fuels with large surface areas give a
smaller value of TpWhere ventilation is restricted, as is generally the case, there is an upper
limit for R regardless of how large the fire load may be. It has been shown
that the important parameters are the window area, A„, its height, h, the
area of the heat-absorbing surfaces of the fire compartment, Aj, and the
ratio of the depth, D, to width, W, of the fire compartment (Fig. 4).
Thomas has drawn a line through the points in Fig. 3 which include measurements for R which have been reported for large-scale, ventilation-restricted fires. The values for R are in reasonable agreement with the CIB
data.
The following equation for R results from the Thomas line
1 22
where the ventilation factor is
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
(2)
82
FIRE RISK ASSESSMENT
(3)
and the fire compartment geometry is
D
hW
E = (AJ-
(4)
For a particular fire load and compartment, R should be calculated for both
Eqs 1 and 2 and the lower value used.
Fire Temperature
There is a maximum temperature developed within the fire compartment,
depending on the fire load and compartment dimensions. Thomas gives the
correlation of the CIB measurements of the average rise in fire temperature,
Sf, over the fully developed fire period as a function of ry, as shown in Fig. 5.
The significance of this datum is that the fire temperature rises to a maximum for Tj = 5 to 10 and then declines. The average fire temperature also
depends on the fire load, as shown in Fig. 6 for plots of large-scale tests with
low fire load densities which fall well below the Thomas curve. There is justification for the assumption that the following equation for a no-draft or
natural-draft condition gives a maximum or upper-limit fire temperature (0f
or Tf) for a given value of rj
Tf = 8025 (Function of yj/) (Function of r;) + 520
(5)
where the fire load and fire compartment is
L
(6)
and the function of ^ and TJ are
fW
= 1 - e-o-^si
FIG. 4—Simple fire compartment.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
(7)
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
83
and
-0.18:)
/w
(8)
V^
There is limited information on the rate of burning under/orced draft conditions. The results of fire tests using excess air at Underwriters Laboratories show no significant variation of temperature with rj or the air supply,
but the fire temperature (6f or Tf) can be related to ^i/, as shown in Fig. 7.
The curve shown has the following equation
Tf = 2160 (Function of i/-) + 520
(9)
where the function of i/' is
/(^) =
1 - e-0-20^f
(10)
Exterior Flame Behavior
Yokoi [12] made the first comprehensive study of flame projection from
windows to evaluate the risk of vertical fire spread. He demonstrated several
influences on the trajectory of the flame plume and temperature of the flame
plume. The wall above the window absorbs heat but restricts air from enter-
1
1
2000
/n
1—
—h—
—1
— t
O + <
D
D
•~o^-
—1
Mas. e, Eq.iS)
—.
1500
'
Thomas
'-^*.^,^
c
1000
500
1
1
L a r g s - t e a l * data
*°*
*
+ +t °
V Yokoi
A Tranton
lncr*aaing 4 «r
O Dianay World
0 Borahamwood
B
OUndarwritart
+ Matz
C Carta rat
0-
1
1—
1
15
—1
20
—I
— t
1
25
FIG. 5—Variation of the average fire temperature rise with the compartment size and window
area, in a natural draft.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
84
FIRE RISK ASSESSMENT
Max.e,
0-2
FIG. 6—Variation of the average fire temperature rise with the fire load, compartment size,
and window area, in a natural draft.
2000
»F
10
12
• -L/(A,A,)V-lb/n'
FIG. 7—Variation of the average fire temperature rise with the fire load, compartment size,
and window area, in a forced draft (Underwriters' Laboratories data).
ing at the flame edges, which will force the flame plume away from the wall;
the wider the window, the closer the flame plume remains to the wall. Yokoi
defined the shape of the window as the ratio of the width to the height of the
upper half of the window and derived a series of plume shapes for different
window shapes. In checking the results of his experiments against large-scale
tests using wood cribs, he obtained good agreement. He also observed that
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
85
where ventilation is restricted, the emerging gases will continue to bum after
leaving the window and will affect correlation.
Thomas [13] correlated results of model tests by Webster [14-16] using
dimensional analysis similar to that used by Yokoi. By assigning a flame-tip
temperature of 540°C (1000°F or 1460°R), reasonable agreement is obtained
between these data and those of Yokoi.
Additional work was done by Seigel [17] on tests conducted at Underwriters Laboratories in which he treated the flames as forced horizontal jets,
defined their projection by a temperature of 540°C (1000°F) at the flame tip,
and recorded the flame temperature in the exterior flame plume.
Other work was conducted at Borehamwood, England, and full-scale tests
were carried out in the United States. All of this work was used to establish
the relationships defining the flame-plume dimensions outside of windows.
These relationships are summarized in Table 2.
Flame-Plume Dimensions
No-Draft Condition—The flame dimensions are an important factor in
calculating radiation from the flame to the receiver or exposed structural
member. Thomas and Law [18] analyzed the data of Yokoi and Webster et
al, which showed correlations that take into account the dominant role of
buoyancy and the turbulent mixing of hot gases in a flame emerging from a
window. The recommended correlation shown in Fig. 8 is expressed as follows
z + h
—r~ = 16
2/3
R
(11)
This may be written for the flame height, z, as follows
= 3.55
R
2/3
-h
(12)
The distance of the center of the flame tip away from the exterior face of the
building, x, depends on the shape of the window and on whether there is a
wall above the window. A wall is defined as a vertical surface which retains its
integrity, exceeding two thirds of the flame height. A flame will be projected
away from the wall surface if air can get behind it. In situations where there
is either a narrow window or a wide window with no wall above, air can move
more easily between the flame and the wall, thereby deflecting the flame or
flames outward. The work of Yokoi was studied, and a regression analysis for
flame-tip projection gives
^
= - 5 ^
(13)
where n is the ratio of the width to the height of the upper half of the windows.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
86
FIRE RISK ASSESSMENT
Q
3
.
^H« =
+
i
_^
-a.
'% 2
x
^^ f1
.i^
i
d
1 Q
\ X
V
f N
• ^ ^
if
it
1
I!
i «
1
• " ^
4»
CN
2
1
•5
<^
" • ^
I
eg
E
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
THoMM snd LBW
(z + h)/h
A-'
Sq.(ll)
X
y-
y
X'
xD
/
-^
/-"
.'r«x o
/^
y
y
X
.
IX
1
1
X
y
D
O
^
v'<P
0-02
O'OS
o / ^
^k—\
87
1
1\
FIG. 8—Flame heights for large-scale tests with a natural draft. (Note: range of n — 0.5 to
18.7.)
Data plotted in Fig. 9 indicate the flame decreases with n and is less than
half the window height for values of n exceeding unity, which includes most
situations. Therefore, the value of x is independent of n and may be represented by the following equation where there is no wall above
X =
0.6A
[i]
1/3
(14)
Where there is a wall above the window, and where h (window height) is less
than 1.25 times the window width
A
(15)
3
Where there is a wall above the window, h is greater than 1.25 times the window width, and the distance to any other window on the same floor exceeds
four widths of an individual window, the following equation can be used
0.54
X =
0.3h
Lw, J
(16)
Forced Draft
The effect of a forced draft on a fire is to increase the rate of burning of a
ventilation-controlled fire. Also, for a given rate of burning, wind or a throughdraft that blows the flames out a window may also affect the flame size and
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
88
FIRE RISK ASSESSMENT
0-75 •
VVokoi
A Tr*nton
0[><«n«y World
DBorahMTMOod
OUnd«fwnt«rs
XTranM
C C*rttr«l
K Kordina
040
Eq. (13)
0-25
V
\ .
x/h
V
»
0-
FIG. 9—Horizontal projection of flume tip for large-scale tests with a natural draft and a wall
above.
direction. Experimental data from tests at Underwriters Laboratories, Inc.
are plotted in Fig. 10 and indicate that the correlation proposed by Seigel,
with /?/v4„y^ raised to the power of unity, is reasonable provided the wind effect, u. is included. The following equation solves for the height of the flame
above the bottom of the window
0.43
+ h = 17.7
Q
(17)
where
Q =
R
A^.
(18)
Analysis of experimental data shown in Fig. 11 shows a correlation between
horizontal projection, flame height, window height, and wind speed. Comparison with this correlation shows that, as wind speed increases, flame
height decreases, but the horizontal projection increases. The horizontal projection, AT, of the flame tip away from the building is given thus
X = 0.077/0-22(2 + h)
(19)
J = Jil
h
(20)
where
The maximum width, w^, of a flame emerging from a window usually exceeds
the window width. The angle made by the emerging flame, shown in Fig. 12,
averages 11 deg giving
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
2x
=
89
0.194
(21)
0.4A:
(22)
therefore
= w +
Effective Flame Boundary
In order to estimate the heat transfer from a flame plume, the boundaries
must be defined. The temperature distribution across a flame section, as
shown in Fig. 13, shows that one approach is to define the flame boundary by
the 540°C (1000°F) contour. Margaret Law proposed that since radiative
transfer will be an integrated effect, which can be made equivalent to a uniform effective temperature, an equivalent step function distribution should
be used. Since radiant transfer is so sensitive to the value of temperature, she
suggested the adoption of the axial temperature (maximum) for the step
function with a defined effective thickness. The problem is to define this effective thickness, bearing in mind an assumption of maximum temperature
across the flame thickness.
Natural Draft
In Fig. 14 the flame emerges above the neutral plane from the upper two
thirds of the window. The flame width varies little with the distance from the
window plane. Therefore, it is reasonable to assume that the step function remains the same size throughout the trajectory (w X 2h/3). This assumption
is consistent with estimated values of emissivity. Wind may deflect the flame
sideways, and it is assumed that the deflection will not exceed 45 deg.
400 J
1
,
Suggested corralation ^
Q y*
^
eq.(17)
Or/ -^
200 J
u»"(z+h)
tt'7iiiin»'
3
Q ^
100
O
OO
"•A'"'""'
0 °
10
20
R / A „ V 4 - lb/f1,min
FIG. 10—Flame heights for large-scale tests with a forced draft (Underwriters Laboratories
data). [Note: range of u = 0.5 to 1.9 m/s (100 to 367ft/min).]
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
90
FIRE RISK ASSESSMENT
100
( u ' / h l " " («h)
200
lt'"/min°"
FIG. 11—Horizontal projection of a flame tip for large-scale tests with a forced draft (Underwriters Laboratories data). (Note: range of u^/h = 1000 to 22 500 ft/min^J
FIG. 12—Plan view of the emerging flames with a forced draft.
Forced Draft
The flame shown in Fig. 14 can emerge from the entire window opening.
The width of the flame does increase with distance out from the window. It is
also reasonable to assume that the vertical dimension increases; however, the
upper vertical increase is already contained in the value for z. It is proposed
that the size of the flame at the window opening be A X w, increasing to A X
(w + 0.4x) at the flame tip, as shown in Fig. 14.
Temperature at Flame Axis
Correlation of test data to analyze the temperature distribution in flames
for natural-draft or no-draft condition is illustrated in Fig. 15. The line in
Fig. 15 has the equation
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
1¥fnp«r«tur«
~lSt«0lunction
FIG. 13—Temperature distribution across aflame section.
2h/3
2h/3
2h/3
Jz»-^(x-5)'Natural draught
w > O-Sh wall abova
Natural draught
w < 0-8h wall atxtva
or no wall abova
Forcad draught
wall or no wall abova
1
1
Natural draught
Plan
IF I
*''<
w, - w+ 0*4K
Natural draught
Plan
Forcad draught
Plan
FIG. 14—Assumed trajectories of the emerging flames.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
91
92
FIRE RISK ASSESSMENT
eo
=
1 - 0.33
R
(23)
A similar correlation for the forced-draft data is shown in Fig. 16, and the
line has the following equation
IA 1/2
(24)
R
eo
Note in Eqs 23 and 24 that the decrease in flame temperature is in direct proportion to the distance along the center line of the flame. By substituting Q^ or
T^ = 940°F (1000°F - 60°F) andl = X in Eqs 23 and 24, the value of GQ or TQ
may be derived. For fires with natural draft, this may give values of Go or Jo (at
the window) greater than the fire temperature, Gf or Tf, in the fire compartment. This results from ignition of unbumed gases outside the fire compartment. Where/orcec? draft exists, the opposite can be expected.
= 1 - 0.33
Conclusion
In order to be able to calculate heat transfer to a structural member from
flames, it is necessary to establish the shape and size of the flame emerging
from the window and the temperature distribution within the flame. The
flame height and its horizontal projection from the building can be calculated, as can the temperature at any point on the flame axis. From this information, an "idealized" flame is determined so that heat transfer calculations can be made.
The temperature of the fire within the building is determined. Although
the fire temperature in a building fire will vary with time, the maximum temperature reached by the fire is used, since this will result in the highest
temperature in the material exposed.
The assessment of fire exposure using the procedures given in this paper
Vvokoi
A Tr«nton
OoiWMy World
K Kordina
9./e.
DinMiutontMS
p-5
10
l.w/R-M'min/lb
FIG. 15—Flame temperature distribution for large-scale tests with a natural draft.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
GEWAIN ON EXTERIOR STRUCTURAL MEMBERS RISK ASSESSMENT
1
—(
Dim«nsionl«st
Sugg«it«d corralation
Eq.(24)
o
1
^ < «Lu
93
1
o s.i(.- -y
O'S
1
1
H
0-5
1-0
I.A,**/R-(l'.min/lb
1
50
FIG. 16—Flame temperature distribution for large-scale tests with a forced draft (Underwriters Laboratories data).
will provide a conservative basis for calculating temperatures of exterior
structural members. In some building fires, the duration of the fire may be
sufficiently long that bare or unprotected structural members will be heated
to "steady-state conditions." In other words, the material may reach an
equilibrium temperature so that the heat falling on the material will be balanced by the heat losses to the surroundings. The steady-state condition
assumes the most conservative approach to heat transfer.
References
[/] Ingberg, S. H., NFPA Quarterly. Vol. 22, No. 1, July 1928, pp. 43-61.
[2] Pryor, A. J., Fire Exposure of Exterior Structural Members. Final Report, Southwest Research Institute, San Antonio, Tex., July 1965.
[3] Seigel, L. G., Materials Research and Standards. Vol. 10, No. 2, 1970, pp. 10-13.
[4] Design Guide for Fire Safe Structural Steel. American Iron and Steel Institute, Washington, D.C., March 1979.
[5] Ove Arup and Partners, Design Guide for Fire Safety of Bare Exterior Structural Steel—
Technical Reports and Designers Manual. Report for American Iron and Steel Institute,
Ove Arup and Partners, London, England, Jan. 1977.
[6] Thomas, P. H., Fires in Model Rooms: CIB Research Programmes. Building Research
Establishment Current Paper CP 32/74, Building Research Establishment, Borehamwood, England, 1974.
[7\ Fujita, K., Characteristics of Fire Inside a Non-combustible Room and Prevention of Fire
Damage. Report 2(2), Japanese Ministry of Construction, Building Research Institute,
Tokyo.
|5] Heselden, A. J. M. and Thomas, P. J., Fully Developed Fires in Single Compartments.
CIB Report No. 20. Fire Research Note 923/1972, Joint Fire Research Organization,
Borehamwood, England, 1972.
[9] Heselden, A. J. M., Smith, P. G., and Theobald, C. R.,Fires inaLarge Compartment Containing Structural Steelwork—Detailed Measurements of Fire Behavior, Fire Research
Note 646/1966, Joint Fire Research Organization, Borehamwood, England, 1966.
[10] Arnault, P., Ehm, H., and Kruppa, J., Rapport Experimental sur les Essais avec de Feux
Naturels Executes dans la Petite Installation. CECM 3/73-11-FCTICM. Puteaux, France,
June 1973.
[//] Gross, D., Field Burnout Tests of Apartment Dwelling Units. Building Science Series 10,
National Bureau of Standards, Washington, D.C., 29 Sept. 1967.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
94
FIRE RISK ASSESSMENT
[12] Yokoi, S., Study on the Prevention of Fire-spread Caused by Hot Upward Current. Report No. 34, Japanese Building Research Institute, Tokyo, 1960.
[13] Thomas, P. H., On the Heights of Buoyant Flames. Fire Research Note 489/1961, Joint
Fire Research Organization, Borehamwood, England, 1961.
[14] Webster, C. T., Raftery, Monica M., and Smith, P. G., The Burning of Fires in Rooms.
Part U. Fire Research Note 401/1959, Joint Fire Research Organization, Borehamwood, England, 1959.
[15] Webster, C. T., Raftery, Monica M., and Smith, P. G., The Burning of Fires in Rooms.
Part HI, Fire Research Note 474/1961, Joint Fire Research Organization, Borehamwood, England, 1%1.
[16] Webster, C. T. and Smith, P. G., The Burning of Fires in Rooms. Part IV. Fire Research
Note 574/1964, Joint Fire Research Organization, Borehamwood, England, 1964.
[771 Seigel, L. G.,Fire Technology. Vol. 5, No. 1, 1969, pp. 43-51.
[18] Thomas, P. H. and Law, Margaret, Fire Prevention Science and Technology. No. 10, Dec.
1974, pp. 19-26.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
STP762-EB/Mar. 1982
Summary
The compilation of papers presented in this publication treats the assessment of fire risk through the coverage of two broad subject areas: (1) defining
fire risk assessment traditionally and systemically and (2) using analytical
methods and test-based quantification to formulate acceptable levels of risk.
The first subject area is examined by Rowe and by Roux in papers that attempt to clearly define and simplify fire risk assessment—the terminology
and concepts as a whole. The second subject area is treated in papers by
Cohn, Ling and Williamson, Brannigan and Dardis, and Gewain, employing
various qualitative, quantitative, and test-based analyses.
Rowe's paper examines the nature of fire risk assessment and develops
bases for assessing these risks systemically. Rowe advises that risks from fire
can never be eliminated completely, but, by setting realistic objectives to reduce the probability and consequences of fire occurrences to acceptable
levels, control of risk from fire can be realized. Such control requires examination of fire occurrences on a systemic level, including identification of
causes; consideration of prevention, mitigation, and cost-effectiveness; and,
ultimately, a re-ordering of priorities aimed at establishing specific control
action plans. The author's use of a systems approach to assessing and controlling fire risk represents an effective means of examining the interaction
among causes, resources, and the strengths and weaknesses of risk-reduction
methods.
An understanding of traditional and current perceptions of fire risk assessment is provided in Roux's definition, characterization, and conceptual analysis of fire risk. In this paper, various definitions of fire risk are examined in
terms of scalar quantification, qualitative levels of acceptable risk, and probabilistic analysis. This treatment of the subject complements Rowe's systemic
approach in that both recognize that acceptable levels of risk relate to a
variety of factors and that the control of risk can be achieved through the reduction of the probability and consequences of fire occurrences.
Cohn's paper presents a differing view of the assessment of fire risk, which
is based on the practical recognition that methods for accurately measuring
degrees of risk have not been used, and, therefore, the formulation of acceptable levels of risk has lacked both consistency and quantitative substantiation. Rowe and Cohn concur that systematic methods must be applied if effective fire risk measurement techniques are to be developed. Rowe's systemic
95
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright'
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
96
FIRE RISK ASSESSMENT
approach involves the estimation of risk through the whole range of risk relations, from causative events, through exposure probability, and ending in
evaluating consequences. On the other hand, Cohn's systematic critical path
approach traces alternatives for fire development in a building emanating
from the incipient point of fire occurrence and sets relative probabilities for
each alternative; that is, it is aimed at preventing fire development at the
ignition stage, rather than reducing the consequences of fires.
The role of fire testing and relevant fire test methods in the assessment of
fire risk are analyzed by Ling and Williamson. The authors draw upon fire
modeling techniques which utilize elements of a "decision tree." A specific
building floor plan is studied with respect to fire growth and occupant-egress
times. The merits and importance of test-based fire risk analyses are demonstrated by way of probability and time determinations, using data from fire
tests on the building materials involved.
The overall accuracy and significance of the test-based fire modeling analysis described by Ling and Williamson is largely dependent upon the relevance and end-point criteria associated with the standard fire tests now in
use. Some of these fire test methods are not based upon rationally developed
exposure conditions and end-point criteria and, therefore, may not qualify as
elements of a fire risk assessment standard or protocol. The need for fire test
exposures and criteria is indicated by the Ling and Williamson paper, if the
challenge of the complexities of the assessment of fire risk are to be successfully met.
By way of balance in the treatment of the assessment of fire risk, Brannigan
and Dardis describe the output of test-based analyses as input for economic
and legal analyses. Economic considerations are related to cost-benefit and
risk-benefit techniques, whereas legal considerations are presented in the
context of tort liability, contracts, and regulations. These authors reflect the
general conclusions of Ling and Williamson—namely, that if test-based fire
analysis is to develop as a means for assessing fire risks on the part of building designers, insurance interests, and regulatory authorities, the test methods must: (1) deal with the fire conditions of severity under examination; (2)
relate to material, product, and system fire performance in terms of risk; and
(3) not serve only as laboratory research tools.
Gewain provides a specific example of the assessment of fire risk based
upon the results of fire tests in his paper on the full-scale testing of exterior
structural steel members. This paper examines the practical application of
the test-based analysis of a specific fire risk situation—the identification of
the fire exposure conditions, examination of the results of full-scale fire experiments, and the development of mathematical treatments to permit the
calculation of heat transfer to structural members during building fire conditions, including the fire temperature and the flame plume shape and size.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
SUMMARY
97
Gewain's paper represents an extension of the conclusions of Ling and
Williamson and Brannigan and Dardis to the extent that it presents a basis
for assessing a fire risk by using test methods that have exposures commensurate with the severity of the fire conditions under examination and by incorporating risk-related fire performance measurements.
This summary has identified the primary messages of six papers covering a
broad range of subjects relating to fire risk assessment. It is clear that more
comprehensive analysis is needed. Useful information has been produced,
but questions remain. What are the unanswered questions; what reviews and
studies must yet be carried out? Rowe perhaps has captured the essence of
the unanswered questions associated with fire risk assessment in the Summary of his paper, wherein he states: "The major question left unanswered is
whether a substantial analysis at the systems level would be useful and warranted. If one expects final solutions, the answer is 'no.' If one is looking for
better insight to focus resources more effectively, locate weak points for corrective action, and make the 'players' involved aware of their responsibilities,
limits of action, and ability to respond to the system, then it might be of considerable merit."
It becomes apparent in summarizing these papers that the difficulties
associated with defining and quantifying fire risk assessment and acceptable
levels of fire risk serve as testimony to the need for a comprehensive multifaceted approach to the development of a fire risk assessment protocol describing all the techniques treated herein, including systemic, systematic,
qualitative, and quantitative test-based analyses. This protocol would find
extensive use by a wide variety of interests in the fire community and would
represent a major contribution to fire safety. The development of such a protocol should be the highest future priority of ASTM Committee E-5 on Fire
Standards and other members of the fire community to improve fire standards development, based upon the meaningful fire risk assessment techniques described by Rowe and the other authors in this volume.
Test-based assessments require data from accurate, reproducible, and relevant fire test methods that qualify as elements of fire risk assessment standards. Such test-based assessments, in combination with representative fire
models, and appropriate probability theory, can, according to these authors,
yield prediction of the causal factors in the growth and spread of unwanted
fires in buildings.
The papers presented in this publication were intended to provide an updated review of techniques, methods, and analytical approaches for quantitatively assessing the risks associated with unwanted fires, while at the same
time identifying qualitative and subjective public perceptions of fire assessment which establish acceptable levels of risk. This compilation of papers
can serve as an element in a sound foundation for building a comprehensive
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
98
FIRE RISK ASSESSMENT
protocol for fire risk assessment standards and stimulate the development of
standards capable of yielding meaningful risk assessment information.
G. T. Castino
Underwriters Laboratories Inc., Northbrook,
111. 60062; symposium chairman and editor.
T. Z.
Harmathy
National Research Council of Canada, Ottawa,
Ontario, Canada KIA 0R6; symposium
chairman and editor.
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
STP762-EB/Mar. 1982
Index
British Standards Institution, 19
Building codes {see Codes, safety)
Building design
For fire resistance, 10
For fire safety, 38
Building Firesafety Model, 24
Buildings
Construction types, 29
Content of, 25
High-rise, 28
Safety of, 29
Burning, rate of, 79, 80-82, 87
Accidents {see also Three Mile
Island nuclear accident), 7
Class 9, 19
Consequences of, 23
Probability of, 23
Actuarial assessment, 68, 69, 71
Aretz case, 69-70
Arson, intentional, 7
ASTM Committee E-5 on Fire Standards, 30
ASTM Fire Tests of Building Construction
and
Materials
(E 119-80), 51, 75-76
ASTM Policy on Fire Standards, 16,
20
ASTM Subcommittee EOS. 15 on
Building Content, 25
ASTM Subcommittee E05.91 on
Planning and Review, 16, 17,
20, 24
Carbon monoxide generation, 36
Carpet pill test, 68
Casualties of fire, 23
Catastrophes, 7
Systems prone to, 8
Codes, safety, 4, 10, 29, 66
Building, 29, 30, 56, 75, 76, 79
Federal and state, 11
Combustible gases, 42
Combustibles, 4, 7
Combustion, type of, 4
Conseil International du Batiment
(CIB), 79, 82
Consumer Product Safety Commission (CPSC), 19, 68, 72
Containment, fire, 4, 10, 47
Contracts, 69, 71
Cost-benefit analysis, 18, 61, 62, 63,
64,65
B
Benefits (see also Cost-benefit
analysis)
And risks, 5
From risk reduction, 61, 63
"Benefits of Environmental, Health,
and Safety Regulation," 20
Benzene case, 72
Beverly Hills Supper Club fire, 18
British Fire Research Station, 23
99
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
Copyright
1982 b y A S T M International
www.astm.org
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
100
FIRE RISK ASSESSMENT
Cost-effectiveness analysis, 12, 13,
14, 62, 63, 66
Council of Environmental Quality,
18, 19
D
Data, types of, 67-69
Delaney Clause, Food and Drug Act,
67
Design capacity, 7
Detection, fire, 4
Detection time, 39
Disasters, natural, 7
Early warning containment systems,
10
Economic consequences of fire, 6167
Emergency equivalent network, 51,
52, 55, 56, 57
Energy Research and Development
Administration, U.S. (ERDA),
23
Escape, 33, 39
Event, fire
Frequency of, 17, 18, 19, 23, 24,
65
Probability of, 36, 37
Exposure
"Attractive," 7
Degree of, 17, 18, 19, 23, 24
Exterior structural members, 75-93
Extinguishing systems, 4, 10, 26
Fatalities, 21, 22, 31
Probabilities of, 33
Federal Emergency Management Administration/Department of
Health, Education, and Welfare/National Bureau of Stanards symposium, 31
Fire
Cause of, 3
Classes of, 4
Consequences of, 3
Death due to, 21, 22
Growth {see Fire Growth)
Prevention, 4, 8, 10
Probability of, 3
Products of, 17, 24, 25
Protection (see Fire protection)
Resistance (see Fire resistance)
Risk (see Fire risk)
Risk assessment {see Fire risk
assessment)
Safety {see Fire safety)
Spread of, 48, 49, 50, 51-53 {see
also Fire growth)
Suppression of, 45, 54
Fire behavior
Exterior, 83-88
Interior, 79-83
Fire development, tracing, 31
Critical path method, 32, 33,
34-35
Fire doubling times, 39
Fire event {see Event, fire)
Fire exposure
Exterior, 76-93
Interior, 76
Fire extinguishers, 11
Fire fighting, 10
Fire growth, 40, 41, 42, 43-57
Experiments, full-scale, 39
Model (FGM), 40, 45, 47, 48, 49,
51-53
Rate of, 39
States of, 44, 45
"Fire Hazard and the Design and
Use of Fire Tests," 20
Fire load, 79, 80, 81, 82, 84
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
INDEX
Fire management, 45
Systems, computer-controlled, 29
Fire modeling, mathematical, 25
Fire prevention programs, 10
Fireproof, 28
Fire protection, 28
Model (FPM), 40, 45, 47, 51-53
Systems, 36
Techniques, 40
Fire resistance, 4
Duration of, 49
Ratings, 76
Of structures, 79
Tests, 47
Fire resistive, 28
Fire risk
Control of, 3
Formulating acceptable levels of,
28-37
Probabilistic analysis of, 32, 33
Fire risk assessment
Components of, 6
Definition of, 59
Economic criteria, 61-67
For exterior structural members,
75-93
Legal criteria, 67-72
Methodology, 4
Process of, 4-11
Systemic, 3-15
Fire safety, 3, 10
Analysis, 29
Control system, 12
Goals for, 30
Success of, 24
From systems viewpoint, 4
Firesafety Concepts (Decision) Tree,
NFPA, 23, 24
Fire stations, location of, 30
Fire temperatures, 77, 79, 82-83,
89, 90-93
Flame-plume dimensions, 85-88
Flame retardant materials, 12
101
Flame spread {see also Fire growth),
39, 41
Flammability tests, 68
Flashover, 41, 42, 43, 44, 45, 47, 50,
51, 54, 55
Food and Drug Administration,
U.S., 67
Forced draft conditions, 83, 84, 8788, 89, 90, 92, 93
Ford Motor Co., 72
Fuels, transportation of, 7
Furniture, upholstered, standards
for, 25, 64
H
Handicapped, safety for, 31
Harm, potential for, 17, 18, 19,
24-25, 26
Hazard, 7 {see also Risk)
Definition of, 20-21, 25
Liability for, 70-71
Quantified, 32
Versus "risk," 20
Hazardous substances, 21
Heat transfer, 77, 79, 89, 92, 93
Convective, 41, 76
Radiation, 76, 89
Heated zone, 41, 42
Human response model (HRM), 40,
45,48
I
Ignition
Ease of, 26
Event, 40
Of combustible materials, 42, 44
Of unburned gas, 92
Multiple, 41
Probability of, 39
Insurance, fire, 10, 11, 71
International Standards Organization (ISO), 19
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
102
FIRE RISK ASSESSMENT
Legal consequences of fire, 67-72
Liability
Protection against, 11
Tort, 69-71
Life safety systems, 28, 29, 30
Liquefied natural gas (LNG), 5
Loss
Measurement of, 64, 65
Per fire exposure, 24, 61
Postflashover, 79
Prevention, fire {see Fire, prevention)
Projection assessments, 68, 69, 71
Protection
Degree of, 66-67
Design standards, 13
Requirements of, 76
Public Health, Massachusetts Department of, 21
M
Material, duration of fire resistance
of, 50, 51, 52, 54, 55, 56
Materials, construction, 4
N
National Bureau of Standards
(NBS), 54, 75
National Fire Protection Association
(NFPA), 19
Decision tree, 45, 46, 47
Systems Concepts Committee, 23,
24
National Fire Protection Association/Department of Housing
and Urban
Development
(NFPA/HUD), 40
Natural draft conditions, 80, 82,
84, 85, 87, 88, 89, 90, 92
Natural gas, storage of, 7
Negligence, 61
Nuclear accident, 17, 18, 21
Nuclear reactors, 7, 38
Nuclear Regulatory Commission, 18
Nuclear war, 7
O
Occupational Safety and Health Administration, 72
Outcomes, probabilities of, 10
Quantitative risk analysis, 38-57
R
Rasmussen report, 17
Regulation, 69, 71-72
Regulatory organizations, 12, 67
Rescue, 4, 10
Response systems (see also Risk,
response strategy), 10
Risk
Acceptable levels of, 5, 10, 11
Defining, 11-12
Determining, 12, 28-37
Analyst, 4
Assessment (see Fire risk assessment)
Aversion, 10
Consequences of, 60
Primary factors, 60, 71
Secondary factors, 60, 71
Control of, 26
Definitions of, 4, 17-20, 23
Dimensions of, 21-23
Elements of, 17, 18, 20, 23, 26
Environmental, 19
Estimation, 5, 8, 9, 10
Exposure, consequences of, 8, 10,
23
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
INDEX
Identification, 5
New, 5
Public concern with, 5
Public's response to, 10
Quantification of, 23
Reduction, 13, 14, 61, 63, 65, 66
Response strategy, 59, 60, 61,
62-67, 70, 71
Voluntary versus involuntary, 21,
22
Risk Management Guide (ERDA),
23
Safety (see also Fire safety)
Approaches to, 4
Definition of, 20, 25, 28
Senate Committee on Governmental
Affairs, U.S., 20
Smoke detectors, 11, 26, 64
Smoke spread model (SSM), 40, 45
Standards, fire, 11, 29, 36, 66, 72
ASTM, 16, 17, 20, 30, 37
Definition of, 59
Design, 13, 14
In system process, 12-14
Performance, 13
Voluntary versus mandatory, 64
State transition model (STM), 39-51
Structural damage, 79
Structural fire resistance, 76
103
Technology and fire risk, 5, 7
Test-based quantification, 59-73
Testing methods
Fire performance, 25
Standardizing of, 37
Tests, fire, 38-57, 75-93
ASTM, 51, 75-76
Condition of, 76
Definition of, 59
Design of, 73
Thiokol Chemical Corp., 69-70
Three Mile Island nuclear accident,
17, 18, 21
Tort liability, 69-71
Toxic materials, potential release of,
7
Tunnel test, 47, 50
U
Underwriters' Laboratory tests, 83,
85, 88, 89
Ventilation, effect of on fire, 43, 50,
81, 85
W
Warning labels, use of, 21
Warning systems, 10
Window area, effect of on fire, 77,
83, 84, 85, 88
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG
&RS\ULJKWE\$670,QW O DOOULJKWVUHVHUYHG :HG0D\*07
'RZQORDGHGSULQWHGE\
0LFKDHO)XUPDQVNL )HGHUDO$YLDWLRQ$GPLQLVWUDWLRQ SXUVXDQWWR/LFHQVH$JUHHPHQW1RIXUWKHUUHSURGXFWLRQVDXWKRUL]HG