Fire hazard and fire risk assessem

INTRODUCTION TO FIRE HAZARD AND FIRE RISK ASSESSMENT ~ Importance of Carbon Monoxide in the Toxicity of Fire USE OF FIRE TESTS FOR FIRE HAZARD ASSESSMENT £qluating the Hazards of Large P

Fire Hazard and Fire Risk Assessment

Marcelo M Hirschler, editor

ASTM Publication Code Number (PCN)

04-011500-31

ASTM

1916 Race Street

Philadelphia, PA 19103

Fire hazard and fire risk assessment/Marcelo M Hirschler, editor.

(STP; 1150) Based on papers presented at a symposium held in San Antonio,Tex., on Dec 3, 1990

"ASTM publication code number (PCN) 04-011500-31."

Includes bibliographical references

ISBN 0-8031-1447-8

1 Fire hazard assessment-Congresses 2 Fire risk

assessment-Congresses I Hirschler, M M II Series: ASTM special

is paid directly to CCC, 27 Congress St., Salem, MA 01970; (508) 744-3350 For thoseorganizations that have been granted a photocopy license by CCC, a separate system ofpayment has been arranged The fee code for users of the Transactional Reporting Service

is 0-8031-1447-8/92 $2.50 + 50

Peer Review PolicyEach paper published in this volume was evaluated by three peer reviewers The authorsaddressed all of the reviewers' comments to the satisfaction of both the technical editor(s)and the ASTM Committee on Publications

The quality of the papers in this publication reflects not only the obvious efforts of theauthors and the technical editor(s), but also the work of these peer reviewers The ASTMCommittee on Publications acknowledges with appreciation their dedication and contribution

to time and effort on behalf of ASTM

Printed in Philadelphia, PA

This publication, Fire Hazard and Fire Risk Assessment, contains papers presented at the

symposium of the same name, held in San Antonio, TX on 3 Dec 1990 The symposiumwas sponsored by ASTM Committee E-5 on Fire Standards Marcelo M Hirschler of SafetyEngineering Labs in Rocky River, OH presided as symposium chairman and is the editor

of the resulting publication

INTRODUCTION TO FIRE HAZARD AND FIRE RISK ASSESSMENT

~ Importance of Carbon Monoxide in the Toxicity of Fire

USE OF FIRE TESTS FOR FIRE HAZARD ASSESSMENT

£qluating the Hazards of Large Petrochemical Fires-N R KELTNER,

adncal Cable Fire Hazard Assessment with the Cone

Current Hazard Analysis Methodology-p J DINENNO AND C L BEYLER 87

Fire Engulfment of Pressure-Liquefied Gas Tanks: Experiments and

Modeling-u K SUMATHIPALA, G V HADJISOPHOCLEOUS, N U AYDEMIR, C M YU,

Development of a Benefit Analysis for an Onboard Aircraft Cabin Water Spray

FIRE RISK ASSESSMENTHow to Tell Whether What You Have is a Fire Risk Analysis Model-1 R HALL, JR 131

Predicting Product Fire Risk: A Review of Four Case Studies-R W BUKOWSKI,

A Reliability Methodology Applied to Halon 1301 Extinguishing Systems in

Fire has been a scourge of society for a very long time now, both in terms of its humanand economic impact: fire fatalities, fire injuries, and direct and indirect losses from fire.North America, in particular, has the dubious distinction of hosting the highest fire fatalityrate per capita in the industrialized world

The traditional way in which fire studies have been made is by using fire tests, of variousdegrees of usefulness, which measure a particular fire property (or fire-test-response char-acteristic, in ASTM fire parlance) The results have then been used to rank materials based

on a single fire property Unfortunately, fire performance (response of materials or products

in fires rather than fire tests) is often poorly predicted by many tests which have usuallynot been designed based on sound engineering principles

It has now become clear that there needs to be better predictive ways to make fire safetydecisions These predictive tools are fire models, which are used to analyze (or assess) thedanger (fire hazard) associated with burning a particular material, product, or assembly in

a specified situation (fire scenario) Thus, ASTM has defined fire hazard as "the potentialfor harm to people, property, or operations" (ASTM Terminology Relating to Fire Stan-dards, E 176-91d) However, fire hazard presupposes that a fire will take place Fire risk

is a measure of fire loss (life, health, animals, or property) that combines (a) the potentialfor harm in the various fire scenarios that can occur and (b) the probabilities of occurrence

of those scenarios, within a specified period, in a defined occupancy or situation As such,fire risk does not assume that a fire will take place, but it incorporates the probability ofthe fire occurring Thus, whereas fire hazard measures the potential for harm with respect

to one single scenario, fire risk measures the potential for harm in the full range of allpossible scenarios, using the probabilities of each one of those scenarios to measure therelative importance of each of them Therefore, a fire risk measure is a statistic derivedfrom an underlying probability distribution on a measure of fire hazard It is important tostress, however, that by its nature, a fire risk measure is not applicable to the prediction ofthe occurrence or of the potential for harm of an individual fire

With the expansion of the capability of large computers and the increased use of thepersonal computer, it has become possible for many people to manipulate large amounts

of information, and to use it in order to predict fire performance Among the consequences

of this has been the appearance of a number of fire models that can predict fire hazard orfire risk

The ASTM board has adopted a policy on fire standards This policy acknowledges theexistence of three kinds of fire standards: fire-test response standards, fire hazard assessmentstandards, and fire risk assessment standards The board gave committee E-S on Fire Stan-dards the exclusive authority to write fire hazard or fire risk assessment standards In order

to better understand what this involves, Subcommittee E-S.3S on Fire Hazard and Fire RiskAssessment is working on standard guides for the development of fire hazard and fire riskassessment standards Several other subcommittees are also working, and have made variousdegrees of progress on a number of fire hazard assessment standards

In order to aid in the understanding of fire hazard and fire risk assessment models, ASTMCommittee E-S has organized this international symposium, conceived within subcommitteeE-S.32 on Research, held in San Antonio, TX, on 3 Dec 1990

The 16 papers published herein can be divided into 5 broad categories: (1) Introduction

1

to Fire Hazard and Fire Risk Assessment, (2) Use of Fire Tests for Fire Hazard Assessment,(3) Fire Hazard Assessment, (4) Fire Risk Assessment, and (5) Fire Risk Assessment andBuilding Codes.

Introduction to Fire Hazard and Fire Risk Assessment

This section includes papers that deal with important aspects that need to be considered

in order to carry out a fire hazard or a fire risk assessment

A common misconception in the public view of fire hazard is that fire hazard is primarily

or exclusively a matter of smoke toxicity The paper by Debanne et al summarizes the twomost important sets of human forensic studies carried out to investigate the issue of smoketoxicity and its implications for fire hazard The studies showed that the victim populationdistributions are very different for fires and non-fire carbon monoxide fatalities, but thatthe blood carboxyhemoglobin distributions, once equal populations are compared, are verysimilar Furthermore, the importance of carbon monoxide in fire atmospheres has notchanged between the 1940s and the 1980s The authors conclude that death in fires, bysmoke inhalation, appears to be overwhelmingly associated with carbon monoxide poisoning.Moreover, their work shows that carbon monoxide can kill human beings (rather than testanimals) at blood carboxyhemoglobin levels much lower than was previously thought pos-sible The combination of this finding and the fact that most small scale smoke toxicity testscannot predict carbon monoxide levels adequately means that such tests, and smoke toxicity

in general, must playa small role in fire hazard assessment

It is often thought that fire hazard and fire risk assessment is necessarily the result ofcomplex mathematical models Watts shows that heuristic models of fire safety, which hecalls fire risk rating schedules, can be used as indicators of fire safety He presents in hispaper, three examples of fire risk rating schedules which have varying degrees of sophisti-cation The first one is the prediction of heat release rates of upholstered furniture by using

a model that combines laboratory scale heat release measurements with various empiricalparameters The next one is the basis for ASTM Practice for Assessment of Fire Risk byOccupancy Classification (Commentary), E 931, which developed an occupancy classificationbased on a Delphi approach and assigned various weighting values to a number of elements.Although this is no longer accepted as a form of fire risk assessment, it is a very usefulsimple means to give numerical results to common sense The final example is a trade-offmodel, again derived from a Delphi approach, to trade off various fire safety alternatives,such as active (smoke detectors, sprinklers) and passive (products with better fire perfor-mance) fire protection measures

Use of Fire Tests for Fire Hazard Assessment

The papers in this section deal with means by which fire tests can be used to predict firehazard in a variety of fire scenarios

One of the types of fire which has the most serious potential is the case of the high intensityfires which can occur in petrochemical facilities, or when liquid fuels are transported InASTM Subcommittee E-5.11 there is work in progress to develop a standard test method

to address such fires Keltner et al addresses one aspect of this problem, when dealing withpetrochemical plant fires They show that it is essential in such cases to adequately char-acterize the fire environment Moreover, in that connection they discuss those cases wherefire temperature is the dominant issue to be addressed and those other cases where fire heatflux is the more important parameter, since there is no univocal correlation between thetwo They conclude that for good fire testing practice both temperature and heat flux should

be taken into consideration before making a fire safety decision

The papers in this section address specific fire hazard assessment problems, by usingcomputer models, statistics, a combination of models and engineering judgment, or a com-bination of experiments and analyses.

DiNenno and Beyler address the way in which new, unconventional materials (composites)might be used to replace traditional materials in naval applications The new materials hadnot been expected to offer the same degree of fire protection, but were known to yieldvarious other advantages Thus, the authors indicate that the results of various fire responsetests can then be used, in conjunction with a "fire hazard analysis package" to determine

an acceptable level of fire hazard Unfortunately, the authors argue that the failings of theexisting "packages" make it necessary to use sound engineering judgment to overcome theirlimitations, and to combine different approaches Once that has been done, progress can

be made in evaluating fire hazard and in weighing the results against the other advantagesand disadvantages

Pressure liquefied gas tanks, especially when used in transportation, need to offer ticular protection because of the potential intense energy of any resulting fire The bestknown examples of such fires are the Boiling Liquid Expanding Vapor Explosions (BLEVEs)which have caused many serious accidents Sumathipala et al have developed a zone firemodel, which they call PLGS-l, together with computational rules, PLGS-2D, to describethe behavior of such tanks when there is an external pool fire in their proximity Resultsfrom experiments involving both midsize (40 L) and large externally-heated partially-filledhorizontal cylindrical vessels have been compared with the predictions of the computer

par-models This work has focussed on heat transfer and pressure response parameters, in order

to improve the understanding of the physical phenomena and develop fire protection strategies.Fatal fires in aircraft seldom occur However, when such a fire occurs, often following asurvivable crash, the results can be catastrophic Hill et al have initiated a study of thebenefits and disbenefits of installing an on board aircraft cabin water spray system Thestudy, which is currently underway, involves aviation authorities of various countries, in-cluding the United States, United Kingdom, France, and Canada This system has beenshown by the authors in full-scale aircraft fire tests, to decrease fire temperatures, and thus,lower burning rates, heat release rates, and smoke emission rates On the other hand, such

a system is generally incapable of extinguishing the fire and can cause a series of unwelcomeconsequences in case of false discharges The final result of the study will likely determinewhether such systems will be installed in commercial aircraft

Fire Risk Assessment

This section contains papers which address the problem of fire risk, either by explainingwhat a fire risk model is, or by using one to apply to a specific fire scenario

After lengthy discussions and disagreements, both based on fundamental concepts andterminology, the paper by Hall is an attempt to clarify what a comprehensive fire riskassessment is and is not The paper describes the most common misconceptions about firerisk analysis The paper also describes key concepts in fire risk analysis and shows how firerisk is simply one facet of overall risk In particular, it explains the types of fires and types

of human behavior that should not be excluded The paper poses questions to the reader

in order to ascertain whether the model in question is a fire risk assessment model orsomething else

The concepts discussed by Hall are illustrated in the work of a research team put together

by the National Fire Protection Research Foundation (NFPRF) to develop a comprehensivefire risk assessment methodology that could be applied to a large number of fire scenariosand a large number of products The paper by Bukowski et al is one result of that program.The authors start by explaining the methodology developed, which is based on the use ofthe fire hazard model HAZARD I, followed by an 8-step procedure In this approach,after the product/occupancy set has been selected, representative characteristics are chosenand incorporated into the fire model The model is then run for a base case product, whichcan represent the average of what is presently being used, or a particular product of specificinterest, for whatever reason The fire risk assessment is then carried out for the base case.The product characteristics are then changed to those of a new product and the fire riskassessment carried out again The process ends with the two results being compared Thispaper describes four cases studied: (1) upholstered furniture in residences (the single firescenario associated with the largest number of fire deaths), (2) carpets in offices (a very lowfire risk scenario), (3) concealed combustibles (electrical cables) in hotels (a low fire riskscenario, but one which has been associated with public controversy), and (4) interior finish

in restaurants (a case which would address heavily regulated products and would introducevertical flame spread into the model) In every case, the results are compared with the fireexperience The work succeeded in developing a methodology and applying it satisfactorily

to a variety of scenarios

A different kind of fire risk assessment methodology is applied by Steciak and Zalosh tothe use of gaseous (Halon 1301) extinguishing systems in computer rooms The methodologyuses occurrence probability data applied to the different failure scenarios for its calculations.Finally, the effects of various measures, such as human intervention, inspection intervals,

dential dwellings governed by the Ontario Building Code.

The final paper is an application of the concepts of fire risk assessment at the communitylevel Harvey, a fire chief, has been instrumental in using the concepts encompassed in afire risk assessment model to evaluate fire safety in his community of Boulder, CO Localplanners, building and fire officials, owners, designers, and builders worked as a team touse new engineering methods for enhancing fire safety in the community while minimizingcost increases This involved not only fire protection measures in the buildings themselves(sprinklers were eventually mandated) but also in the number of fire stations and theirequipment Entry into the system was voluntary at first but became mandatory when it wasfound that overall cost savings were attained, with no opposition from the public who couldsee the benefits

Conclusions

The papers summarized above should provide the reader with a broad understanding ofthe issues involved in fire hazard and fire risk assessment and with an overview of some ofthe most interesting techniques available today The diversity of papers is probably sufficient

to offer different perspectives and tools both for workers in the field and for other readersconcerned with fire safety This is an area where advancements are occurring in leaps andbounds; however, the papers presented here should serve as an excellent literature database The symposium chairman thanks the other members of his committee, in particular

Dr Harry K Hasegawa (Lawrence Livermore National Laboratory) and Dr James R.Mehaffey (Forintek) for their invaluable assistance to make this publication possible Healso acknowledges the efforts of the authors and of the ASTM personnel who made thishappen

Marcelo M Hirschler,

Safety Engineering Laboratories,Rocky River, OH, 44116,symposium chairman and editor

Introduction to Fire Hazard and Fire Risk

Assessment

ASTM STP 1150, Marcelo M Hirschler, Ed., American Society for Testing and Materials, Philadelphia, 1992, pp 9-23.

ABSTRACT: This paper addresses the issue of determining toxic fire hazard in a manner that

is relevant to the large-scale fire scenarios that cause most lethalities, ventilation controlled Oashover fires.

The work involves (a) a literature study of background information on the toxicity of carbon monoxide (CO) to humans, (b) two very extensive forensic studies (with 2241 and 2673 cases, respectively) on human lethality involving CO in fires and non-fires, (c) statistical analyses of the forensic data, to ensure full separation of variables, and (d) an analysis of the literature

011 the effects of different parameters in fire atmospheres and in small-scale tests.

This work addresses five issues essential to fire hazard assessment:

(I) Relative role of toxicants other than CO in causing fire fatalities and critical hemoglobin (COHb) values representing lethality.

carboxy-(2) Differences in populations between fire victims and those dying in other CO-containing atmospheres.

(3) Comparison of modern fire atmospheres, containing smoke from man-made materials, with traditional fire atmospheres.

(.t) CO yields in flashover fires and effects of fuel chemistry.

(5) CO yields in small-scale toxicity tests and fuel effects.

The work has shown that:

(I) The toxicity of fire atmospheres is determined almost exclusively by the amount of CO The lethal CO threshold level depends on the physical condition of the victim, but COHb values> 20% can produce lethality with no other apparent cause.

(2) Fire and non-fire CO victim populations are very different: fire victims are much older

or much younger and more infirm, and thus more sensitive to CO than those in fire exposures.

non-(3) Replacing large amounts of natural and traditionally used materials by man-made terials has made no difference to fire atmosphere toxicity.

ma-(4) CO concentrations in large-scale flashover fire atmospheres are determined by oxygen availability and such variables, but are little affected by chemical composition of fuels (6) Small-scale tests give excessively low CO yields so that they cannot be used to predict toxic fire hazard for ventilation controlled flashover fires, although post computational

CO concentration corrections can make these tests useful as part of fire hazard ment calculations.

assess-KEY WORDS: carbon monoxide, carboxyhemoglobin, fire fatalities, fire gases, fire hazard, fire scenario, flashover, smoke toxicity, toxicity test

'Associate professor, Department of Epidemiology and Biostatistics, Case Western Reserve werntv, Cleveland, OH 44106.

Uni-0Safety Engineering Laboratories, 38 Oak Road, Rocky River, OH 44116.

'Dean, College of Science and Liberal Arts, Florida Institute of Technology, Melbourne, FL

32901-It has been discussed in the National Fire Protection Association (NFPA) Quarterly, in

1933, that the main direct cause of death in fires was combustion product toxicity [1] ForASTM use, smoke is interpreted as the sum total of the gaseous, liquid, and solid airborneproducts of combustion (ASTM Terminology Relating to Fire Standards, E 176) Smoke

is, thus, not a uniform material, so that its composition depends on the exact conditionsunder which it was generated However, smoke always contains two principal types of toxicgases: asphyxiants and irritants

All organic (that is, carbon-containing) materials give off carbon monoxide (CO), carbondioxide, and water on combustion [2] Many other combustion gases are released fromburning materials, but most of them are characteristic of particular classes of fuels Carbonmonoxide is a highly toxic asphyxiant gas, which is odorless and tasteless It is the individualtoxic gas associated with the largest fire hazard Many other components of smoke are muchmore toxic, but those are usually present in much smaller concentrations On the otherhand, some gases are present at higher concentrations (typically carbon dioxide and water)but they are of much lower toxicity Two studies which sent fire fighters into actual buildings

on fire, while equipped with combustion product monitors, [3,4] had the same conclusions;the overwhelming hazardous toxicant in fire is CO

The other principal toxicants in fires are acrolein that has the second highest ratio, after

CO, of peak level found to lethal level and is emitted by polyolefins and cellulosic materials),hydrogen cyanide (emitted by N-containing materials), and hydrogen chloride (emitted byCl-containing materials) [5]

The mechanism by which CO acts on mammals is by competing with oxygen for thehemoglobin in blood and tying it up as COHb, rather than as the normal oxyhemoglobin.The hemoglobin (Hb) fraction tied up as COHb is normally expressed as percent COHb(which means the percentage of the total hemoglobin present as COHb rather than asoxyhemoglobin) Carbon monoxide reacts ~21O times faster with hemoglobin than doesoxygen, so that it can lead to an oxyhemoglobin deficiency, even at low CO concentrations.This lack of oxyhemoglobin then leads to hypoxia, which can cause cerebral damage andeventual death by asphyxiation However, the reaction of CO with hemoglobin to yieldCOHb is reversible, and COHb levels will decrease when the CO exposure ceases.Traditional wisdom has set a value of 50% COHb as the threshold level for human lethality[5,6] This has been taken to mean that blood COHb levels of 2:50% COHb inevitablyleads to death, and that if a fatality is autopsied and its COHb level was <50%, CO poisoningcould not be the sole cause of death However, it is known that 25% COHb already causesmyocardial damage [1]

Carbon monoxide is also present in non-fire atmospheres Two typical examples are thoseresulting from malfunctioning unvented gas or charcoal heaters, or from automobile exhaust

In these two cases, CO is widely recognized as being virtually the only toxicant of anyconsequence present This is an important difference from fire atmospheres, which containseveral other toxicants, some of which may be present at high enough concentrations tocause senous concern

Several fire hazard issues will be addressed in this paper, which includes a description ofsome of the major results from a series of studies sponsored by the Society of the PlasticIndustry, Inc [8]

(1) The presence of these other toxicants and their importance in the toxicity of fireatmospheres is an issue with important implications for fire hazard assessment because

of the way in which smoke toxicity is normally determined One of the issues thatneeds to be understood is the relative role of toxicants other than CO is causing firefatalities and whether there is any critical value of COHb that represents lethality

(2) Fire fatalities are most commonly unwitting victims of external events that overtakethem, while many individuals who succumb to the intoxicating effects of automobileexhaust do so at their own initiative Thus, it is worth investigating whether thepopulations that lose their lives in fire atmospheres and in other CO-containing at-mospheres are identical If it is found that the two populations are significantly dif-ferent, the follow up query is whether the effects of CO can be separated from othereffects, inherent in the characteristics of the population involved Fire hazard or firerisk assessment models require this answer since they have to concentrate on thepopulation most at risk in fires.

(3) The modern world in which we live contains a large proportion of products madefrom synthetic materials, for example, plastics The combustion of these materialscan generate toxicants which are different from those traditionally generated by naturalmaterials This has caused a great deal of speculation that the toxicity of modern firesmay be significantly different from that of fires of another era Therefore, anotherissue of importance to fire hazard is an understanding of the relative toxicity of firesinvolving new and traditional materials

(4) The vast majority of the fire fatalities that occur in the United States are found awayfrom the room of fire origin That is a clear indication that those fires which lead tofatalities tend to be big fires In other words, they are flaming fires, which haveinvolved the entire room of origin and gone beyond it These types of fires are oftendescribed as flashover or post-flashover fires [9] Any smoke toxicity assessment thatwould attempt to address fire risk should, of course, address such high intensity flamingfires Therefore, it is of interest to fire safety to understand what types of atmospheresare to be expected in these fires, particularly as regards their CO content

(5) Small-scale toxicity tests are the most common means of assessing smoke toxicity ofmaterials or products The final question to be asked in this study regards the COyields in such tests and how they compare to those in the real full-scale fires thatcause the majority of fire fatalities

Literature Search

Dr Gordon L Nelson, at the Department of Polymer Science at the University of ern Mississippi, carried out a literature search on information available regarding humanexposure to CO and fatalities This study unearthed more than 100 references, and foundseveral major forensic studies, both of fire victims and of victims of non-fire CO exposure.Full details of this study will be given elsewhere, together with those of other associatedstudies, all sponsored by the Society of the Plastics Industry [8] However, the 12 mostnotable studies are described in Table 1 Only two of these studies have been very widelypublicized, namely, those in Maryland and Glasgow Interestingly, none of the referencedstudies involved over 1 000 cases This means that several of them would have had greatdifficulties making any statistically valid separate analyses of individual variables

South-It is worth discussing some of the most interesting results of these studies They havefound that when various individuals are exposed to a particular atmosphere, their COHblevels can be very different Furthermore, some people may even survive the exposures,and others, for no apparent reason, succumb These studies have also found both survivors

of CO exposure who have COHb levels of well over 50% and fatalities from pure COexposure with COHb levels in the range of 20 to 40% In fact, around the turn of thecentury, John Scott Haldane poisoned himself with CO and measured COHb levels of over50%, without succumbing to the experience

The age distribution of fire victims was found to have a bimodal distribution, both in the

Maryland and Glasgow studies [19-22] (Fig 1) In contrast, the age distribution for fire suicide victims, from automobile exhaust in the Oslo study [15], is a simple distributionwith a single peak (Fig 2).

non-A histogram of victim frequency versus COHb levels for non-fires shows that there is astrong difference between accidental victims (CPSC study of gas heaters [17]) and thosecausing their own deaths (Oslo study of automobile exhaust [16]) Figure 3 shows that suicidevictims end up with very high COHb levels, typically in the 65 to 85% levels On the otherhand, accidental victims of CO poisoning end up with much lower levels, between 26 and81%, and with a median of 45% This is a very important finding, because it indicates thataccidental victims of CO poisoning can die at very low levels, without any additional factorbeing involved

Figure 4 compares the frequency distribution of COHb levels for fire fatality studies withthat of an accidental non-fire CO exposure forensic study It is clear that the distributionsare very similar and that both have significant proportions of fatalities at levels well under50% COHb, the traditional lethal level

This portion of the study shows that 50% COHb can not be a magical number such that

it is the inevitable pass/fail lethality criterion for pure CO poisoning

Forensic Investigation on CO Victims Across the United States

The Department of Polymer Science at the University of Southern Mississippi (USM)carried out an extensive investigation, on a countrywide basis, of fatalities associated with

CO The investigators were Gordon L Nelson, Dennis V Canfield, and James B Larsen.This investigation prepared a data base of 2 241 cases, originating in 37 laboratories acrossthe entire United States The vast majority of the victims died over a relatively short timeframe in the 1980s Table 2 shows the variables investigated, and Table 3 presents someinformation on the data base itself It is worth pointing out that the data base, which is

CARBON MONOXIDE AND FATALITIES

much more extensive than any of the earlier ones, contains roughly twice as many victims

of fires as of non-fire CO poisoning More extensive details of this study will be givenelsewhere [8]

The cases that were retained for further investigation where those where the COHb levelwas ~ 20%, because it was decided that it was very unlikely that victims with lower levels

of COHb would have died exclusively of CO poisoning In this connection it is worthmentioning that smokers can exhibit COHb levels of up to 14% (for example, Ref 25).Table 3 shows that only -8% of all the cases in the data base had a known COHb bloodlevel of under 20% saturation, evenly distributed among fire and non-fire cases

Over 10% of the victims in either data base (fire or non-fire) had a blood COHb level

TABLE 2- Victim variables included in the USM database.

1 Data about the Victim Itself:

TABLE 3-Details of USM database.

of the victims being either very young or elderly On the other hand, non-fire victims have

a unimodal age distribution, which peaks between 30 and 45, similar to the one shown inFig 2 The data discussed earlier have indicated that there is an effect of population char-acteristics on susceptibility to CO poisoning Consequently, as expected, the COHb distri-butions for both sets of data are not identical, with the fire one being centered at a somewhat

lDwer COHb level than the non-fire one (Fig 6) The difference in age distributions,

'Ibis will be analyzed in the next section

This portion of the study reinforced the suggestion that CO can kill at COHb saturationlI:vels of well under 50% without requiring an additional toxicant This can now be considered

a firm finding

llatistical Analysis of Forensic Investigation

A team led by Sara M Debanne and Douglas Y Rowland, in the Department of llatistics and Epidemiology at Case Western Reserve University (CWRU) analyzed theUniversity of Southern Mississippi data base, by means of a separation of variables The

Bio- unvariables analyzed separately were: age, alcohol level, presence of drugs and preexistingJlhysical condition, as well as the source of CO Again, the full set of results will be publishedelsewhere [8], but a summary will be discussed here The results were very interesting Theyshowed that the COHb frequency distributions are very similar, once these other factorsDve been accounted for Figure 7 shows the distributions for two ethanol-free populations

which are very comparable, all fire victims ages 6 to 20 and all non-fire victims above age

6 The similarity of the two curves indicates, or at least suggests very strongly, that CO isdearly, the principal, if not the overwhelming cause of death in fires

This study suggests that there is generally no need to look for an additional source oflethality in fire atmospheres if the COHb level ranges between 20 and 50% Since CO alonecan account for all deaths, this indicates that the effect of other toxicants in fire atmospheres

• invariably small, if not negligible, in the majority of cases This gives a clear answer now

10the first issue to be addressed

This study also gives a fairly definitive answer to the second issue to be addressed; thepopulations of fire and non-fire victims are generally very different Moreover, the fireW;tim population contains more individuals that are at higher risk of succumbing to carbon

FIG 7-COHb distribution for selected victims (comparable populations) in the study carried out at the University of Southern Mississippi (USM) on fire (ethanol-free 6 to 20 years old) and non-fire (ethanol- free> 6 years old) CO fatalities.

monoxide poisoning than does the population as a whole or than does the population thatdies from CO poisoning in non-fire cases

Forensic Investigation on CO victims across time in Cleveland, OH

The same team at Case Western Reserve University also created a new data base of COvictims of fires and non-fires This data base contains all victims of CO poisoning investigated

by the Cuyahoga County Coroner's office between the years of 1938 and 1979 CuyahogaCounty, in Ohio, is the county surrounding the city of Cleveland The reason for the choice

of period was that this coroner's department was headed by the same man (Dr SamuelGerber) over the entire period concerned The Cuyahoga coroner's office investigated andcarried out autopsies on the victims of every fire in the county as well as of every case ofaccidental death and was respected for its high scientific integrity and accuracy This database, containing 2 637 cases, was somewhat different from the USM database It containedroughly twice as many non-fire victims as fire victims This makes it the largest data baseever put together on CO and fatalities Table 4 shows some of the main features of the database, with more details to be shown elsewhere [8J

There were two reasons to carry out this study The first was to discover whether theCOHb levels at which fire victims died have changed over the years and the second was tocompare, once again, fire and non-fire data

As discussed in the introduction, the modern world, particularly in the United States,provides the population with a large number of products made with synthetic materials.Such materials were first developed in the 1930s and 1940s, and did not generate a largefraction of the overall combustible products in use until the 1960s or 1970s Their relativeproportion of use in the United States is probably still rising

In Fig 8, the median level of COHb found each year for the cases in the study is shown.The data are presented as differences with the 1941 value and two curves are presented,

issue that was to be addressed; the effect of new materials on the toxicity of fire atmospheresappears to be negligible.

The separation of variables shows some remarkable similarities between this study andthe previous one That is, the populations of fire and non-fire victims are very different,and this difference can be seen to affect their sensitivity to CO poisoning

Study of Fire Retarded and Non-Fire Retarded Products

Effect of Study on Usefulness of Small-Scale Tests

A study was carried out by the Center for Fire Research at the National Institute forStandards and Technology (NIST) for the Fire Retardant Chemicals Association [26] Thisinvolved burning 5 sets of products in fire retarded and non-fire retarded versions: TVcabinets, business machines, upholstered chairs, circuit boards, and electrical cables.The small-scale techniques used were the cone calorimeter and the NBS cup furnacesmoke toxic potency test In those apparatuses the individual materials were tested as mock-

up combinations The medium-scale technique used was the furniture calorimeter, wherethe entire products were tested individually The full-scale test used involved combiningall five products together in a single room

The results indicated that the fire retarded products were safer than the non-fire retardedproducts, indicating that improved fire performance (by whatever means it is achieved) leads

to lower fire hazard The results showed that, in the full-scale tests:

• FR materials lost less than half the mass of non-fire retarded ones

• FR materials released Y4 the heat of non-fire retarded ones.

• FR materials generated Y3 of the toxic gases of non-fire retarded ones.

• FR and non-fire retarded materials generated similar smoke obscuration

The results also indicated that the cone calorimeter could be used to predict full-scaleignitability, heat release, and flame spread It could not be used, however, to predict release

of CO The authors found that the bench-scale tests were expected to give adequate yields

of gases other than CO, but will almost inevitably give too low yields of CO Thus, scale tests would be expected to be biased in favor of non-CO species, meaning that theyexaggerate the relative importance of such non-CO species

small-The production of combustion gases in small-scale tests is, thus, heavily influenced bythe chemistry of the materials being burned However, the production of CO in large-scalefire tests, and thus presumably in real fires, is only somewhat influenced by the chemicalproperties of the substance being burned

Moreover, the CO production in large-scale fires appears to be much more influenced

by the availability of oxygen in the fire This, in turn, is affected by variables such asgeometry, ventilation, configuration, turbulence, and mixing Therefore, the authors con-clude that the use of any less-than-room-sized tests for making CO predictions has to bedeferred until these oxygen supply variables have been sorted out

This indicates that the toxicity of the atmosphere in real fires is likely to be little affected

by the type of materials causing the fire The CO levels in large-scale fires are governed

by the number and size of openings, the mass loading, and the burning characteristics (that

is, ignitability, flame spread, and heat release) of the fuels

Rough algorithms now exist for assessing the CO yields for various large-scale fire narios These are probably not yet adequate for widespread use, but they indicate the wayforward to assess the toxic hazard associated with any large-scale fire scenarios Furtherstudies are needed before these can be considered well documented [27]

sce-Experience also shows that in full-scale tests where full flaming room involvement isachieved (those fires that cause the majority of fire fatalities) there is ventilation controland the oxygen levels get very close to zero Small-scale tests are rarely carried out underconditions of low oxygen and high heat flux, because flaming combustion does not occurunder such conditions.

Consequently, it is now possible to give answers to the last two issues to be investigated

in this study The atmospheres in those large-scale fires that cause most lethality have atoxicity determined mostly by the concentration of CO and by a low oxygen content Theyield of this CO is almost independent of the type of materials being burned

Toxic potency of smoke is a quantitative expression which relates the concentration ofsmoke and the exposure time to the achievement of an adverse effect, usually lethality, on

a test animal (E 176) It has to be stressed that the toxic potency of smoke is heavily dependent

00 the conditions under which the smoke is generated, which affect both the quality andthe quantity of smoke

During the 1970s and early 1980s, many small-scale tests for smoke toxic potency weredeveloped [6] These tests differ in many respects, including: fire model, being static ordynamic, use of animals or analytical tools, animal model for bioassay, and end point Partlydue to all these differences, the tests lead to tremendous ranking variations for the smoke

of various materials This was illustrated in a study of the toxic potency of 14 materials by

2 methods which ranked one material most toxic by one protocol and least toxic by theother protocol Although neither of those protocols is in common use now, this workillustrates some of the inevitable problems of such tests

The toxicity of smoke is a function of its composition, which depends, in turn, both onthe fuel and the combustion mode Thus, the composition of the smoke generated by anyindividual material varies broadly from test to test, and so will its toxic potency In fact,the toxic potency of the smoke of most ordinary materials (whether natural or synthetic) isvery similar [6] Thus, relative smoke toxic potency rankings depend on the exact composition

of the smoke being tested, that is, the small-scale test protocol, and are of little interestfrom the viewpoint of fire hazard assessment

Moreover, small-scale smoke-toxicity tests give inadequately low yields of CO andadequate yields of other toxicants Thus, such tests require post computational correctionfor CO before their results are directly relevant to fire hazard in the real fires that causemost fire deaths This correction for CO yields has been addressed in recent toxicology work[28], where it has been applied to the results of full-scale and small-scale tests using rats

as animal models, with an excellent outcome

From the point of view of fire hazard assessment, the work described here puts intoperspective the importance of one of the elements of fire hazard assessment, namely smoketoxicity It shows that care must be taken when using toxicity data as input for fire hazardassessment

Conclusions

This paper has addressed the five fire hazard issues presented in the introduction, whichare:

(1) (a) The toxicity of fire atmospheres is determined to a very large extent by the amount

of CO; the contribution of other toxicants is usually very small (b) The lethal level

of CO is heavily dependent on the individual concerned, and the 50% COHb thresholdnormally mentioned is not a realistic value Both fire atmospheres and non-fireatmospheres containing mainly CO can cause lethality due exclusively to CO at COHblevels of 20%

(2) The population of individuals who die in fires has a bimodal distribution, with anexcess of very young and very old and infirm people, while that of people who die

in automobile exhaust CO exposures has a unimodal distribution Thus, it appearsthat fire victims are significantly more sensitive to CO poisoning than non-fire victimsand may tend to die at lower COHb levels

(3) The replacement of large amounts of natural and traditionally used materials by made materials among the normal products in everyday use has made no difference

man-to the man-toxicity of fire atmospheres

(4) The CO concentrations in the atmospheres of large-scale fires are determined by theoxygen availability, ventilation, mass loading, and other such variables, but are veryminimally affected by the chemical composition of the fuels

(5) Small-scale tests give excessively low CO yields while they can predict adequatelythe concentrations of other combustion products Thus, such tests cannot be used topredict toxic fire hazard for the fire scenarios of greatest interest, ventilation controlledflashover fires, which cause most of the fire fatalities The use of post computationalcorrection for CO concentration, which is now becoming available, is required tomake these tests relevant to the critical fire scenarios They can only legitimately,however, be used as a part of a fire hazard assessment

Acknowledgment

This work was carried out under the sponsorship of the Society of the Plastics Industry,Inc

(14) Cimbura, G., McGarry, E., and Daigle, J., "Toxicological Data for Fatalities due to Carbon

Monoxide and Barbiturates in Ontario-A Four Year Survey 1965-1968," Journal of Forensic Sciences, 1972, pp 640-642

(15) Kishitani, K., "Study of Injurious Properties of Combustion Products of Building Materials at

Initial Stage of Fire," Journal of Faculty of Engineering University Tokyo (B), Vol 31,1971, pp.1-35

(16) Teige, B., Lundevall, J., and Fleischer, E., "Carboxyhemoglobin Concentrations in Fire Victimsand in Cases of Fatal Carbon Monoxide Poisoning," Zeitschrift Rechtsmedizin, Vol 80, 1977, pp.17-21

117) Consumer Product Safety Commission, Federal Register, Vol 45 No 182, 17 Sept., 1980, p 61880

(IS] Pach, J., Cholewa, L., Marek, Z., Bogusz, M., and Groszek, B., "Analysis of Predictive Factors

in Acute Carbon Monoxide Poisonings," (Toxicological Clinic and Institute of Forensic Medicine,Krakow, Poland,) Folia Medica Cracoviensia, 1978, pp 159-186

119] Berl, W G and Halpin, B M., "Human Fatalities from Unwanted Fires." Fire Journal, Sept

1979, pp 105-123

(20) Halpin, B M., Radford, E P., Fisher, R., and Caplan, Y., "A Fire Fatality Study," Fire Journal,

May 1975, pp 11-\3

(2/) Anderson, R A., Watson, A A., and Harland, W A., "Fire Deaths in the Glasgow Area: I.

General Considerations and Pathology," Medical Science Law, Vol 21, 1981, pp 175-83

(22) Anderson, R A., Watson, A A" and Harland, W A., "Fire Deaths in the Glasgow Area: II.The Role of Carbon Monoxide," Medical Science Law, Vol 21, 1981 pp 288-94

(23) Birky, M M., Malek, D., and Paabo, M., Journal of Analytical Toxicology, Vol 7,1983, p 265

124] Gormsen, H., Jeppesen, N., and Lund, A., 'The Causes of Death in Fire Victims," Forensic Science International, Vol 24, No.2, 1984, pp 107-111

(25) Wald, N., Howard, S., Smith, P G., and Bailey, A" "Use of Carboxyhemoglobin Levels toPredict the Development of Diseases Associated with Cigarette Smoking," Thorax, Vol 30,1975,

pp 133-40

126] Babrauskas, V., Harris, R H., Gann, R G., Levin, B. C, Lee, B T., Peacock, R D., Paabo,M., Twilley, W., Yoklavich, M F., and Clark, H M., "Fire Hazard Comparison of Fire-Retardedand Non-Fire-Retarded Products," NBS Special Publication 749, National Institute of Standardsand Technology, Gaithersburg, MD, July 1988

127] Pitts, W M., "Executive Summary for the Workshop on Developing a Predictive Capability for

CO Formation in Fires," NISTIR 89-4093, National Institute of Standards and Technology, ersburg, MD, 1989

Gaith-(.?S) Babrauskas, V., Harris, R H., Braun, E., Levin, B.C,Paabo, M., and Gann, R G., 'The Role

of Bench-Scale Data in Assessing Real-Scale Fire Toxicity," NIST Technical Note 1284, NationalInstitute of Standards and Technology, Gaithersburg, MD, 1991

sessment, ASTM STP 1150, Marcelo M Hirschler, Ed., American Society for Testing and

KEY WORDS: fire, fire hazard, fire risk, fire risk analysis, fire risk assessment, fire safety,rating schedules

Fire risk rating schedules (FRRSs) are heuristic, quasi-mathematical paradigms of firesafety They constitute various processes of modeling and scoring hazard and safety param-eters to produce a rapid and simple estimate of relative risk FRRSs are useful and powerfultools that can provide valuable information on the risks associated with fire They havebeen applied to a variety of hazard and risk assessment projects to reduce fire safety costs,set priorities, and facilitate use of technical information FRRSs provide an important linkbetween the complex scientific principles of theoretical and empirical models, and the lessthan perfect circumstances of non-laboratory conditions found in real world applications.There is significant diversity in the formation and use of schedule approaches to fire safety,yet there is very little information on methodology or evaluation This leads to a lack ofstructure or consistency in the development process, and a lack of validity or credibility inthe outcomes This paper presents some results from a survey and study to identify char-acteristics and components of, and criteria for, FRRSs The approach of the study is tolocate, review, and evaluate existing FRRSs and their methods of development Critiquingthese methods has formulated concepts to help guide future development and evaluation.Examples described in this paper concentrate on the more technically sound approacheswhich relate most closely to development of fire standards These examples are typicallymore robust than most of the schedules surveyed and the generalized criticisms of FRRSs

do not necessarily apply to the examples cited

Following the examples, general characteristics and common components of FRRSs aredescribed The analysis of these features is intended to suggest a systematic approach to theformulation and validation of such schedules

Examples

FRRSs come in a large variety of formats and a broad spectrum of purposes Examplesused in this paper to illustrate the principles of FRRSs are selected for their relationship tofire standards development

'Director, Fire Safety Institute, PO Box 674, Middlebury, VT 05753

Upholstered Furniture Heat Release Rates

Babrauskas (1] and Babrauskas and Walton [2] present a simplified rule for estimating

the maximum or peak heat release rate of upholstered furniture based on design factors(Eq 1) The most accurate determination of heat release rate involves destructive testing in

a furniture calorimeter The approximation rule uses non-destructive measures and genericdesign categories which are readily determined in the field These factors are multipliedtogether implying interdependent contributions to the magnitude of the heat release rate

Qpeak = 210.0 [fabric factor] x [padding factor] x [mass]

where

[fabric factor] = 1.0 for thermoplastic fabrics (fabrics such as polyolefin, which melt prior

to burning),0.4 for cellulosic fabrics (cotton, rayon), and0.25 for pye or polyurethane film type coverings

(padding factor] = 1.0 for polyurethane foam or latex foam,

0.4 for cotton batting,1.0 for mixed materials (that is, both polyurethane or latex foam andcotton), and

0.4 for neoprene foam

[mass] combustible mass, in kilograms

[frame factor] = 1.66 for non-combustible,

0.58 for melting plastic,0.30 for wood, and0.18 for charring plastic

[style factor] = 1.0 for plain, primarily rectilinear construction, and

1.5 for ornate, convoluted shapes and intermediate values for mediate shapes

inter-_ the constant 210.0 has units kW/kg

The concept of an FRRS is to use readily determinable parameters, combined to yield asingle value which can be associated with fire risk A significant facet of this type of ap-proximation is the mathematical distinction between risk, or the value being approximated,and the result of the approximation There is not a one-to-one mapping of the model output

to the real world In this example, while the rate of heat release for a given mass has aninfinite number of possible values, the rate of heat release for a given mass approximated

byEq 1 has only 40 possible discrete values

Assessing the Fire Risk of Products

In a brief paper, Gross [3] describes a process which illustrates several important concepts

ofFRRSs The suggested approach to assessing fire risk of products involves the followingfive steps:

• Identify appropriate fire response characteristics

• Identify appropriate measure(s) of the fire response characteristics

• Normalize measures to a consistent scale

• Assign weights indicating relative importance

• Combine to yield assessment of fire risk

Steps one and two are largely dependent on experience and expert judgment Step threeuses, as example, a scale of 0 to 100, with the performance limit of a fire response char-acteristic test equated to 100 Step four relies on fire incidence statistics and expert judgment.Step five is not discussed by Gross

As an illustration of this approach, Gross uses the evaluation of fire response characteristics

of curtains and drapes as shown in Table 1 below

The NFP A 701 large-scale test is considered an appropriate test for ignitability and flamespread and is weighted as having 90% of the relative importance A maximum char length

of 250 mm is considered the performance limit in the test; hence, the factor of 0.4 is used

to normalize results to the 0 to 100 scale Similarly, for a performance limit of 800 Dm

(specific optical density) in ASTM E 662, a normalizing factor of 0.125 is used Althoughnot discussed by Gross, one can envision how the products of the weighting factors andnormalized test results could be summed to give a fire risk index

Assessment of Fire Risk by Occupancy Classification

ASTM Standard Practice for Assessment of Fire Risk by Occupancy Classification (E931) presents a method for assessment of fire risk inherent in different occupancies asessential to the development and use of fire standards for furnishings The stated purpose

is twofold and corresponds to basic objectives of FRRSs:

1) provide a uniform procedure for assessment of fire risk; and

2) provide a logical basis for establishing levels of fire performance

The approach is based on consideration of twelve parameters relating to the risk of deathand injury, property loss, and fire potential The rater assigns each parameter a valueaccording to the degree of risk present in the occupancy Each parameter has also beenassigned a fixed weight

The product of the weight and risk rating yields the parameter total Fixed multipliersare provided for fire detection and fire suppression systems where deemed appropriate Thesum of the parameter totals for death and injury and property loss times the total for firepotential, provides a risk assessment index This number is intended as a basis for both firerisk comparisons and for placement of occupancies within risk classes The fire risk assess-ment rating form is illustrated in Fig 1

Decision Logic for Trading Between Fire Safety Measures

Harmathy [5] presents a method of quantitative decision support concerning the alence of various fire safety alternatives This is a comprehensive approach which involvesfour distinct parts

equiv-TABLE 1- Evaluation of fire response characteristics of curtains and drapes.

Fire Response Characteristic Test Method Normalizing Factor Weighting Factor

aStandard Test Method for Specific Optical Density of Smoke Generated by Solid Materials

per year per square meter of floor area.

• A data set to evaluate expected fire loss in a standard reference building

• A set of parameters and conditions which have an impact on expected fire loss

• A data set which quantitatively defines the effect of parameter conditions on expectedfire loss

The probability model (Eq 2) is a straightforward analysis of likelihood that a fire will

pew or spread to a certain stage The fire loss expected in each stage for both property

and life is also analyzed Data for this model is gleaned from North American statistics onfire occurrence, fire spread, and fire losses The parameters and their corresponding quan-titative impacts were derived through a Delphi exercise.

where

L - fire loss (human and property) expectation (dollars per year per m" of floor area);

N = expected number of fire incidents (per year per m" of floor area);

P p = probability that, given ignition, the fire will not reach flashover;

PFN = probability that, given flashover, the fire will not spread to other compartment(s);

P FSD = probability that, given flashover, the fire will spread to other compartment(s) bydestruction;

lp = average loss (human and property) resulting from fires that do not reach flashover

(dollars per incident);

IFN = average loss (human and property) resulting from postflashover fires that do notspread (dollars per incident);

IFsD - average loss (human and property) resulting from postflashover fires that spread

by destruction (dollars per incident);

IFsc = average loss (human and property) resulting from postflashover fires that spread

by convection (dollars per incident)

Decoupling the probability model from the parameter evaluation is an important feature

It illustrates the use of relatively simplistic schedule formats to provide input for morerigorous risk assessment models

Table 2 shows the parameters and their associated quantitative values Column two resents the incremental impact of a parameter on the probability that, given ignition, thefire will not reach flashover Column three is the incremental impact on the probability that,given flashover, the fire will not spread to other compartments For example, if rooms arelarger than 20 m2, the probability of no flashover is increased by 6.4%, and the probability

rep-of no fire spread to another compartment is decreased by 12.4%

These four examples were selected for their relationship to ASTM E 5 activities Theyare generally more technically sound than many other FRRSs presently in use Review ofadditional examples of schedules that have achieved considerable usage may be found inthe SFPE Handbook of Fire Protection Engineering [6].

Generalized Characteristics of Fire Risk Rating Schedules

Fire risk rating schedules have been referred to by various designations such as indexsystems [7] numerical grading [8], and point schemes [9] They originated as insurancerating schedules in the 19th century, but in the last few decades, the basic concepts haveappeared in a wide variety of formats FRRSs have proliferated because of the high utility

in their relative ease of application However, they lack validity due to the unspecifiednature of the selection of variables and their relationships

Characteristics of FRRSs identify their diversity and usefulness They are described here

in terms of their functional relation to other risk evaluation methods and in terms of theirapplications and attributes

The function of an FRRS is, explicitly or implicitly, fire risk analysis By nature of thecircumstances, fire safety decisions often have to be made under conditions where the data

TABLE 2-lncremental effect of various factors on probabilities Pp and PFN•

Average floor area of compartments

Average temperature in January

NOTES: 1 The reference conditions are characterized by 0 increments

2 Positive values indicate improved safety; negative values indicate decreased safety

are sparse and uncertain The technical parameters of fire risk are very complex and normallyinvolve a network of interacting components, the interactions generally being nonlinear andmultidirectional However, complexity and sparseness of data do not preclude useful andvalid approaches Such circumstances are not unusual in decision making in business orother risk venues (the space program illustrates how success can be achieved when there islittle relevant data) Thus, development of approaches to fire risk analysis is a significantongoing process Methods of fire risk analysis may be classified by evolutionary developmentinto four categories: narratives, check lists, schedules, and emulations.

Narratives- The bulk of our present day wisdom on fire safety consists of narrative

descriptions of various hazardous conditions and ways to reduce or eliminate them Theseappear in the form of building codes and various fire safety standards Narratives do notattempt to evaluate the fire risk quantitatively, rather, a risk is judged acceptable if it complieswith published recommendations An obvious limitation to this approach is that such nar-ratives can not hope to be interpreted and applied uniformly for the myriad conditions ofhuman activity While there is much common ground among different fire hazard situations,

in detail there is considerable variation

Check lists- A common accessory of fire safety is a listing of hazards and recommended

practices While narratives take the form of a book, check lists represent a table of contents,identifying complex fire safety concepts with a few words These checklists comprise valuabletools for identifying fire risk factors They do not, however, distinguish among the importance

of these factors For example, the relative value of hydrants, sprinklers, and manual fireextinguishers is not constant, but a function of other features of a building's form andfunction

Schedules- In general, fire risk rating schedules assign values to selected variables based

on professional judgment and past experience The selected variables represent both positiveand negative fire safety features and the assigned values are then operated on by somecombination of arithmetic functions to arrive at a single value This single value can becompared to other similar assessments or to a standard

Emulation-Growing interest in analytical fire risk assessment and an increasing database, has led to use of more sophisticated mathematical techniques Emulation methodsmanipulate fire safety variables according to recognized theoretical principles Among theseapproaches are: computer simulation, linear regression, network analysis, and stochasticmodeling

While both schedules and emulations are numerical approaches to fire risk assessment,they represent different levels of depth and accuracy FRRSs can be considered to be atone end of a continuous quantitative assessment spectrum, where hazard and exposure areestimated through the use of simple models A detailed probabilistic risk analysis would be

at the other end of this spectrum, where hazard and exposure are tested, measured, andassessed as rigorously as possible Choosing the depth of the risk analysis is a critical decisionthat depends on such factors as time and resource committment and the intended use ofthe results

Applications

For many situations where a quantitative risk assessment is desirable, an in-depth theoreticanalysis may not be cost-effective, nor appropriate This could be the fundamental casewhere greater sophistication is not required, where prioritization is the principal objective,

or where it is necessary to institutionalize for a wide base of usage

The level of accuracy demanded for a fire risk assessment is not the same as for otherengineering purposes; often establishing an order of magnitude will suffice Time and re-

source expenditure increases as the depth of analysis is increased In an age where resourcesare scarce, and efficiency is prized, maximizing the utility of the FRRS is clearly desirablefor the many situations where assessing fire risk is fundamental.

FRRSs have gained widespread acceptance as cost-effective prioritization and screeningtools for fire risk assessment Risk assessment can be an expensive and labor-intensiveprocess, and much time and money c,!n be wasted if the products or facilities with the

~atest potential for risk and associated liability are not identified and assessed first Without

a prioritization plan it will not be known whether a risk was worth assessing until the timeand money has been spent In order to be cost-effective, a prioritization system must besimple, rapid, and accurate An appeal of FRRSs is that it can be all of these

Perhaps the most common implicit justification of FRRSs is the need for a simplisticprocess of fire risk analysis In most applications, the target of a FRRS is a broad class ofproducts or facilities for which a detailed fire risk analysis of each individual case is notieasible FRRSs have appeal to administrators charged with risk management decision-making responsibilities, but who may be unfamiliar with the details and mechanics of therisk assessment process Wide spread implementation of a generalized approach to fire risk

is contingent on its appeal to a broad class of users including architects, code officials, andproperty managers

Attributes

From their range of application and relationship to other methods of fire risk assessment,

we abstract the following general attributes of FRRSs:

• Inclusive-FRRSs tend to treat fire phenomena as part of a larger system rather than

in terms of individual elements That is, they tend to address a broad, systemic concept

of fire safety

• Numerical-The objective of FRRSs is to provide a logical and consistent approach

to fire assessment through a mathematically derived ordinal grading

• Heuristic-FRRSs are relatively simple models which do not rely exclusively on onstrated principles of physical or management science (but their credibility is enhanced

dem-to the extent they do employ such principles)

• Experiential-Input and structure is largely based on experienced judgment, supportedwhere possible or appropriate by historical data, fire tests, and more rigorous predictivemodeling

• Surrogate-Output is usually a measure of relative risk with no significant physicalunits (but it may be input to a model of absolute risk)

Components

Components of fire risk rating schedules identify elements they have in common MostFRRSs are found to have three basic components [8]: a list of parameters, procedures forassigning values to the parameters, and relationships which define mathematical operations

on the parameter values to produce an assessment of hazard or risk Various aspects andramifications of these three components are further discussed

Parameters

Parameters of FRRSs, also referred to as elements, factors, variables, and so forth, identifythe ingredients of fire safety Fire safety is a complex system with an inordinately large

number of factors which may affect it These can range from ignitability of personal clothing,

to availability of a heliport for evacuation It is computation ally feasible to deal with only

a relatively small number of variables Therefore, it becomes necessary to reduce the largenumber of variables to an appropriate subset

It is intuitively appealing to postulate that safety from fire is a Paretian phenomenon inthat a relatively small number of factors account for most of the problem This is supported

by general fire loss figures which suggest that a small number of factors are associated with

a large proportion of fire deaths

It is necessary then, to identify some defensible combination of factors which account for

an acceptable portion of the fire risk Such a process is not identified for most FRRSs.Selection of parameters is apparently most often arbitrary, with correspondingly disparateresults Even when some parameters seem comparable, their usage may differ and lack ofoperational definitions leads to confusion in trying to implement the schedule

Where this subject is addressed, the approaches to selection of parameters generally fallinto one of three categories:

• Delphi, or some less formal consensus process which relies on expert judgment

• Fire scenarios, ideally based on loss statistics, but usually employing subjective opinion

• Cut set of a hierarchical, success tree, providing an inclusive list

Values

The second component of FRRSs is the quantitative measures that associate fire risk withqualitative characteristics of the parameters Value selection for the parameters is whereFRRSs show the greatest range of variation Sources for these values range from fire testdata to hearsay

The approach to selection of values can be either objective or subjective The maincriticism of objective estimates is that they may be using past data which are not relevant

to the future conditions being considered For some decisions, the subjective approach may

be superior when it takes into account both historical data and the decision maker's sessment of present and future influences The problem with subjective values is that, likeselection of parameters, they are often arbitrary Although it is possible that much thoughtwent into the determination of parameter values in the FRRSs reviewed, in many cases it

as-is not evident Where a process as-is identified, it takes the form of a modified Delphi exercas-ise

or a less rigorous method of eliciting experienced judgment

In some FRRSs, two sets of values are associated with each parameter, intensity andimportance, for example, Gross [3] and ASTM E 931 Intensity is a measure of the amount

or degree that a parameter is present in a specific application, for example: fuel load, flamespread rating, and so forth Importance is a weight indicating the influence or significance

of the parameter to fire safety This is one of the notable differences between FRRSs andcheck lists, which treat all parameters equally In many FRRSs intensity and importanceare combined implicitly into a single dimensionless value

Since the process for selecting values is undocumented in most FRRSs, there is no way

to update the schedule to account for technological change Where Delphi or similar esses are used, they are generally too cumbersome to allow the schedule to be responsive

proc-to new developments

Relationships

Relationships are the mechanisms by which the parameter values are combined to yield

a measure of fire risk In general, the intensity measure is multiplied by the importance

weight for each parameter and the results are summed, for example, Fig 1 Most often,safety parameters are assigned positive values and hazard parameters are assigned negativevalues This algebraic addition implies the parameters are all independent in their effect onfire risk.

An alternative approach uses the product of all safety parameters divided by the hazardparameters Multiplication of parameter values implicitly assumes that all the parametersare completely interactive That is, the impact on fire risk of any unit change in a singleparameter will be dependent on the values of each of the other parameters

Most of the FRRSs studied do not explicitly address the interaction of parameters Yet

it is intuitive that certain combinations of parameters, for example, smoke detection andautomatic suppression, are not simply additive in their associated effect on fire risk.Linearity is implicitly assumed by the relationships employed in all the methods studied

Summary

Fire risk analysis involves a large number of multifarious factors which are difficult toassess in a uniform and consistent way The analysis of such complex systems is difficult butnot impossible as evidenced by activities in areas such as nuclear safety and environmentalprotection Detailed risk assessment can be an expensive and labor intensive process andthere is considerable scope for improving the presentation of results Fire risk rating sched-ules can provide a cost-effective means of risk evaluation which is sufficient in both utilityand validity

With all their potential value, it is unfortunate that so little attention has been directedtoward the development process of fire risk rating schedules While much effort may havebeen expended, it is not often evident and the results frequently suggest a shortage ofanalytical skill in the development process They typically lack even a statement of underlyingassumptions, much less any other information that would be conducive to a validation It

is paradoxical that while seeking logic and consistency in application, they show little indevelopment

Some notable exceptions to these generalities have been presented here as examples Thispaper also presents generalized descriptions of the characteristics and components of fire

risk rating schedules in an effort to promote a more rational and consistent approach to

their development and use The importance of FRRSs as tools of fire risk analysis ensurestheir continued popularity Hopefully, their general credibility will be improved throughincreased consideration in formulation and evaluation

References

(1] Babrauskas, V., "Upholstered Furniture Heat Release Rates: Measurements and Estimation,"

Journal of Fire Sciences, Vol 1, 1983, pp 9-32.

(2] Babrauskas, V and Walton, W D., "A Simplified Characterization of Upholstered Furniture Heat

Release Rates," Fire Safety Journal, Vol 11,1986, pp 181-192.

(3] Gross, D., "The Use of Fire Statistics in Assessing the Fire Risk of Products," INTERFLAM '85

(4] Methods of Fire Tests for Flame Resistant Textiles and Films, NFPA 701, National Fire Protection

Association, Quincy, MA, 1989

(5] Harmathy, T Z., et aI., "A Decision Logic for Trading between Fire Safety Measures," Fire and

Materials, Vol 14, 1989, pp 1-10

16] Watts, J M., Jr., "Fire Risk Assessment Schedules," SFPE Handbook of Fire Protection

Engi-neering, Section 4, Chapter II, National Fire Protection Association, Quincy, MA, 1988, pp 4-89

to 4-102

(7] Rosenblum, G R and Lapp, S A., "The Use of Risk Index Systems to Evaluate Risk," Risk

Analysis: Setting National Priorities, Proceedings of the Society for Risk Analysis, Houston, TX,

[8] Nelson, H E., "Overview: Numerical Grading Systems," Report from the 1987 Workshop onAnalytical Methods for Designing Buildings for Fire Safety Design, National Academy Press,Washington, DC, 1988.

[9] Rasbash, D J., "Approaches to the Measurement and Evaluation of Fire Safety," presented tothe Third International Fire Protection Engineering Institute, Wageningen, The Netherlands 1980

Use of Fire Tests for Fire Hazard

Assessment

Petrochemical Fires," Fire Hazard and Fire Risk Assessment, ASTM STP 1150, Marcelo M.

Hirschler, Ed., American Society for Testing and Materials, Philadelphia, 1992, pp 37-43.ABSTRACT: Fire testing and fire research would be aided by the standardization of the fireenvironment, test articles, sensors, and procedures for the sake of reproducibility and inter-laboratory comparisons Current and proposed fire standards dictate control by either "firetemperature" or by "fire heat flux" This equivocation can lead to confusion This articlediscusses the circumstances under which temperature may be the more important controlparameter and when heat flux may be the more important control parameter Examples aregiven

KEY WORDS: fire testing, fire research, fire temperature, fire heat flux, fire standards, firecontrol, fire comparisons, hydrocarbon fires, petrochemical fires

Fires that might occur in a transportation accident involving hazardous materials or in apetrochemical industry accident can put workers or the public at risk In an attempt toIUlderstand, and possibly reduce these hazards, there is a strong interest in determining theresponse and/or survivability of a variety of items when subjected to large petrochemicalfires This effort usually involves actual testing as opposed to a purely analytical evaluationdue to complexities of modeling heat transfer in this fire environment

Most of the fire test work done in the past in the United States, and indeed around theworld, has involved fire protection of materials and/or assemblies for buildings, not petro-chemical fires Standard test methods were developed by voluntary standards groups, such

as the American Society for Testing and Materials (ASTM) Committee E-5 on Fire dards, the National Fire Protection Association (NFPA), and the International StandardsOrganization (ISO) Because the testing is part of the process of developing and rating thematerials, large numbers of tests are conducted with relatively simple test assemblies Thetests are typically conducted in a furnace The thermal exposure is usually based on con-trolling the furnace to a time-temperature curve, such as that given in ASTM Method forFire Tests of Building Construction and Materials (E 119) It is understood that this typeattest is not a simulation of a compartment or building fire environment; rather, experienceshows that performance of materials in these tests is a good indicator of performance inactual fires [1]

Stan-For petrochemical fires, the approach has been to develop test specifications that areintended to be reasonable simulations of the fire environment in the use situation, or toproduce damage greater than that expected to occur in a large fraction of the accidents.Organizations around the world have made an effort to develop standardized test methodsfor "hydrocarbon fires"

'Supervisor, senior members of technical staff, respectively, Sandia National Laboratories, ThermalTest and Analysis Division, 2737, Albuquerque, NM 87185

37