mission - c.ymcdn.com in flammable liquid and ... test data accumulated in our laboratory, as ......

35
FIRE PROTECTION FIRE PROTECTION UL 2360: A NEW TEST FOR WET BENCH PLASTICS LESSONS LEARNED FROM A CARBON DIOXIDE SYSTEM ACCIDENT FIRE SAFETY DESIGN OF THE FUNDACIÓN CAIXA GALICIA 43 35 PROVIDING PRACTICE-ORIENTED INFORMATION TO FPEs AND ALLIED PROFESSIONALS FALL 2001 Issue No. 12 ALSO: Mission Critical Fire Protection Mission Critical page 9 FIRE PROTECTION FOR THE OFFSHORE INDUSTRY FIRE PROTECTION FOR THE OFFSHORE INDUSTRY 23 23 30 30

Upload: vandang

Post on 20-Mar-2018

217 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FIRE PROTECTIONFIRE PROTECTION

UL 2360: A NEW TESTFOR WET BENCH PLASTICS

LESSONS LEARNED FROMA CARBON DIOXIDE SYSTEM ACCIDENT

FIRE SAFETY DESIGN OFTHE FUNDACIÓN CAIXAGALICIA

43

35

PROVIDING PRACTICE-ORIENTED INFORMATION TO FPEs AND ALL IED PROFESSIONALS

FALL 2001 Issue No. 12

ALSO:MissionCriticalFire Protection

MissionCritical

page 9

FIRE PROTECTION FOR THE OFFSHORE INDUSTRYFIRE PROTECTION FORTHE OFFSHORE INDUSTRY

2323

3030

Page 2: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

Fire Protection Engineering 1

3 VIEWPOINTLoss Prevention Challenges in the Pharmaceutical IndustryGreg Jakubowski, P.E., CSP, Merck Safety and Environment Engineering

4 LETTERS TO THE EDITOR

7 FLASHPOINTSFire Protection Industry News

23 PROACTIVE VS. PRESCRIPTIVE FIRE PROTECTION FOR THEOFFSHORE INDUSTRYPerformance-based criteria that use the latest modeling programs are begin-ning to grow. This article addresses performance-based criteria for fire protec-tion in the offshore industry.John A. Alderman, P.E., CSP, and Marlon Harding, P.E.

30 UL 2360: A NEW TEST FOR WET BENCH PLASTICSSemiconductor cleanrooms, while they protect highly sensitive chips fromdamaging particles, still have fire protection concerns that, if not addressed,can ruin millions of dollars of property.Jane I. Lataille, P.E.

35 FIRE SAFETY DESIGN OF THE FUNDACIÓN CAIXA GALICIABUILDING IN SPAINA fire protection case study regarding the special requirements of a new cultural center in Spain.George Faller, C.Eng.

43 LESSONS LEARNED FROM A CARBON DIOXIDE SYSTEMACCIDENTSeveral lessons can be learned from an accidental carbon dioxide extinguishing system discharge.Morgan J. Hurley, P.E., and James G. Bisker, P.E.

48 CAREERS/CLASSIFIEDS

50 SFPE RESOURCES

56 FROM THE TECHNICAL DIRECTORA Paradigm ChangeMorgan J. Hurley, P.E.

Cover illustration by Bill Frymire/Masterfile

For more information about the Society of Fire Protection Engineers or Fire Protection Engineering magazine, try us at www.sfpe.org.

Invitation to Submit Articles: For information on article submission to Fire Protection Engineering, go to http://www.sfpe.org/publications/invitation.html.

contents FA L L 2 0 0 1

9COVER STORYMISSION-CRITICAL FACILITIES: TELECOMMUNICATIONS AND E-COMMERCEFire protection systems must ensure continuous operation of computer networks, Websites, and communications in order for businesses to survive in today’s environment.Ray Schmid, P.E.

Subscription and address change correspondence should be sent to: Fire Protection Engineering, Penton Media, Inc., 1300 East 9th Street, Cleveland, OH 44114 USA. Tel: 216.931.9566. Fax: 216.696.7668.E-mail: [email protected].

Copyright © 2001, Society of Fire Protection Engineers. All rights reserved.

FIRE PROTECTIONFIRE PROTECTION

Fire Protection Engineering (ISSN 1524-900X) is published quarterly by the Society of Fire ProtectionEngineers (SFPE). The mission of Fire ProtectionEngineering is to advance the practice of fire protectionengineering and to raise its visibility by providing information to fire protection engineers and allied professionals. The opinions and positions stated arethe authors’ and do not necessarily reflect those of SFPE.

Editorial Advisory Board

Carl F. Baldassarra, P.E., Schirmer Engineering Corporation

Don Bathurst, P.E.

Russell P. Fleming, P.E., National Fire Sprinkler Association

Douglas P. Forsman, Firescope Mid-America

Morgan J. Hurley, P.E., Society of Fire Protection Engineers

William E. Koffel, P.E., Koffel Associates

Jane I. Lataille, P.E.

Margaret Law, M.B.E., Arup Fire

Ronald K. Mengel, Honeywell, Inc.

Warren G. Stocker, Jr., Safeway, Inc.

Beth Tubbs, P.E., International Conference of Building Officials

Regional Editors

U.S. HEARTLAND

John W. McCormick, P.E., Code Consultants, Inc.

U.S. MID-ATLANTIC

Robert F. Gagnon, P.E., Gagnon Engineering, Inc.

U.S. NEW ENGLAND

Thomas L. Caisse, P.E., C.S.P., Robert M. Currey &Associates, Inc.

U.S. SOUTHEAST

Jeffrey Harrington, P.E., The Harrington Group, Inc.

U.S. WEST COAST

Marsha Savin, Gage-Babcock & Associates, Inc.

ASIA

Peter Bressington, P.Eng., Arup Fire

AUSTRALIA

Richard Custer, Arup Fire

CANADA

J. Kenneth Richardson, P.Eng., Ken Richardson FireTechnologies, Inc.

NEW ZEALAND

Carol Caldwell, P.E., Caldwell Consulting

UNITED KINGDOM

Dr. Louise Jackman, Loss Prevention Council

Publishing Advisory Board

Bruce Larcomb, P.E., BOCA International

Douglas J. Rollman, Gage-Babcock & Associates, Inc.

George E. Toth, Rolf Jensen & Associates

Personnel

PUBLISHER

Kathleen H. Almand, P.E., Executive Director, SFPE

TECHNICAL EDITOR

Morgan J. Hurley, P.E., Technical Director, SFPE

ADVERTISING/SALES

Terry Tanker, Penton Media, Inc.

MANAGING EDITOR

Joe Pulizzi, Custom Media Group, Penton Media, Inc.

ART DIRECTOR

Pat Lang, Custom Media Group, Penton Media, Inc.

COVER DESIGN

Dave Bosak, Custom Media Group, Penton Media, Inc.

Page 3: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

By Greg Jakubowski, P.E., CSP

The pharmaceutical industry dis-covers, develops, manufactures,and markets a broad range of

innovative products to improve humanand animal health. Think of where theworld would be today without peni-cillin, without vaccines that prevent amyriad of infectious diseases, and without modern products that controlcholesterol, prevent heart attacks, easepain, and reverse the anaphylacticeffects of something as potentially life-threatening as a bee sting. Pharmaceuti-cals not only save lives but add signifi-cantly to quality of life, and do sowhile controlling overall health-carecosts.

The pharmaceutical business is complex, with years, and sometimesdecades, of research required to discover a beneficial molecule, test itfor efficacy and safety, determine itsvalue in the human body, and produceand distribute it. This complexity, andthe length of time required to bring aproduct to market, engenders not onlya highly competitive business environ-ment, but also one that creates unusualloss prevention challenges. Even asmall incident can prevent a productfrom coming to market for months, orperhaps years. This delay can have adevastating effect for patients that may

have been awaiting a breakthroughtreatment. Because of the complexity ofthis business, pharmaceutical facilitiesencompass a wide range of hazards,from bulk chemical facilities, clean pro-duction facilities, and laboratories, towarehouses and site utilities.

A primary goal of any loss preven-tion program in the pharmaceuticalindustry is to prevent an incident fromoccurring. If prevention is effective,extensive suppression efforts will notbe needed, which is a philosophy notunlike that of many pharmaceuticalproducts. Extensive process safety pro-grams, chemical handling training, andsite master planning assist in prevent-ing incidents. However, there arenumerous fire protection challengesfacing the pharmaceutical industrytoday, including:

1. The ability to understand andapply the codes and standards ofeach country in which facilities arelocated. The business is growingsignificantly in South America,Asia-Pacific, and Africa.

2. Applying sprinklers that must meetextra hazard, quick response, andinsurance company requirementswhile presenting clean surfaces incompliance with good manufactur-ing practices.

3. Effective control of static ignitionsources in flammable liquid anddust operations.

4. Evaluating alternatives to sprinklersthat not only minimize collateraldamage, but also reduce the needfor control of fire water runoff.

5. Minimizing loss potential in high-value research, manufacturing, andstorage areas. This includes protec-tion of susceptible areas fromexternal hazards that may exposeand impact our business.

6. Safe handling of flammable liquidsin research laboratories.

7. Notwithstanding the pharmacologi-

cal benefits of products to patients,avoiding exposure to employeesor the environment of smallamounts of active pharmaceuticalor biological ingredients during anincident which can be a concern,especially in large-scale manufac-turing operations.

8. Conducting safe construction andfacility rehabilitation operations inand around operating plants.

Scientists and manufacturing employ-ees who are used to demanding theapplication of excellent science in theirdaily operations demand the same stan-dard of performance from fire and lossprevention experts when applying solu-tions to these challenges. At the sametime, these challenges must be met in acost-effective manner to meet, if notexceed, the expectations of our cus-tomers.

Effective efforts to control fires andinnovative means of fire prevention arekey challenges to the pharmaceuticalindustry as it strives to respond to soci-ety’s needs for life-saving medicinesand vaccines, improving quality of life,and having a positive impact on overallhealthcare costs.

Greg Jakubowski, P.E., CSP, is withMerck Safety and EnvironmentEngineering.

FALL 2001 Fire Protection Engineering 3

viewpoint

Loss Prevention Challenges in the Pharmaceutical Industry

Page 4: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

We are writing this letter to comment onthe subject matter addressed in recent arti-cles by Edward K. Budnick (“AutomaticSprinkler System Reliability” page 7) andBruce H. Clarke (“MicrobiologicallyInfluenced Corrosion in Fire SprinklerSystems” page 14) in Fire ProtectionEngineering Issue No. 9. These articlesboth address matters associated with thereliability of fire sprinkler systems.

We agree that traditionally fire sprinklershave proven to be highly reliable devices,which serve as a first line of defense toinsure the life and property safety in build-ings in which they are installed. We alsobelieve that by combining emerging newtechnologies and feedback from field ser-vice and testing records, newly developedfire sprinkler systems can even more reli-able. However, we feel it important toshare our observations relating to informa-tion on reliability of specific, existing sys-tems as may relate to development of newfire sprinklers.

As Mr. Budnick points out, existing relia-bility data is mainly based on older stylesof sprinkler heads. Data for newer firesprinkler designs is generally not included.In addition, and of primary importance inhis analysis, reliability of fire sprinklers istreated as a whole regardless of possiblefundamental differences in design andoperating principles on which individualsprinkler designs are based. We believe thatfire sprinkler reliability information for dif-ferent mechanical/design technologiesshould be compiled separately since themore widely accepted designs may workon quite different principles, and individualmodel types themselves can literallyaccount for millions of installations.

If data from such analyses could beapplied readily, “not acceptable designs” ordesigns with potentially low reliabilitycould be isolated more readily and rejectedthan at present. In addition, this approachwould provide an opportunity for the firesafety community cooperatively to developreliability testing protocols for fire sprinklerand associated system components beforenew designs are introduced. Such anapproach is absolutely consistent withassessments needed to better understandenvironmental or aging effects typified by

the MIC problem which Mr. Clarke discuss-es in his article as well as the performanceof the recently recalled Central Omegasprinklers.

Using the Omega case as an example,the reliability of that particular designdegraded as a function of years in service.Test data accumulated in our laboratory, aswell as test data collected by othersinvolved in evaluations of Omega sprinklerreliability, demonstrated a consistent dropin reliability as number of years in serviceincreased. We have seen, for example, reli-ability of Omega sprinklers we have evalu-ated drop below 70 percent after severalyears of service. The inference to be drawnfrom this is that resulting performance willbe significantly less than the 90 percent-plusreliability one would expect an individualfire sprinkler to show to have as describedin the Budnick article.

More recently, the Central GB family ofsprinklers has been the subject of notifica-tions from UL and Factory Mutual and an“Informational Bulletin” from the CaliforniaState Fire Marshall warning of potentialproblems. Test results from evaluations ofalmost 300 GB heads taken from service forvarious periods of time at different loca-tions without fire occurrence show a similarrelationship between failure levels andyears in service to that of the Omega mod-els which were the subject of an earlierCPSC recall. These levels of reliability areunacceptable and suggest problems withproduct design and/or manufacture.

In terms of the subject matter of theBudnick article, it is clearly crucial to pre-vent production of low-reliability “notacceptable” designs in the future so thatsuch sprinklers are not introduced to themarket. Our community needs to learnfrom recent recalls and the informationthey have provided us, and we also needto prevent such failures from happeningagain.

Joseph B. Zicherman, Ph.D.Fire Cause AnalysisPoint Richmond, CA

Editor’s Note: Since the receipt of this let-ter, the manufacturer of the GB series of

sprinklers, in cooperation with the U.S.Consumer Product Safety Commission, hasannounced a replacement program. Fullinformation on the program and the specif-ic models of sprinklers involved is availableat www.sprinklerreplacement.com.

I am writing regarding the recent article“Automatic Sprinkler System Reliability” byEdward K. Budnick. While it is true thatmany of the fire sprinkler reliability studiesare based on the use of older sprinklers,it’s not as if we have no information on theperformance of newer sprinklers.

Residential sprinklers, for example, haveonly been available since 1981. Of 551 firesin sprinklered residential occupanciesreported to Operation Life Safety between1983 and 1995, 90 percent were controlledby a single sprinkler, with another 8 per-cent controlled by two sprinklers operating.Scottsdale, Arizona, which published areport on 10 years of residential sprinkleruse beginning in 1985, found that 41 of 44fires (93 percent) were successfully sup-pressed by only one or two operatingsprinklers. Two of the three fires thatopened more than two sprinklers wereflammable liquid arson fires. In PrinceGeorges County, Maryland, residentialsprinklers have been required since 1992,and were involved in 83 fires by August of1998. In 72 of them (87 percent) only a sin-gle sprinkler activated, with two sprinklershandling another 6 percent of the fires. Thehigh percentage of one-sprinkler and two-sprinkler operations in all of these reportsindicates that sprinkler reliability is notsomething of the past.

Recent fire sprinkler recalls and replace-ment programs attest to the sprinkler indus-try’s commitment to near-zero tolerance ofsprinklers that are not reliable. A substantialeffort is underway to ensure that potentialproblem sprinklers do not remain in placeto eventually damage the historicallyremarkable performance statistics.

Russell P. Fleming, P.E.Vice President of EngineeringNational Fire Sprinkler AssociationPatterson, NY

4 Fire Protection Engineering NUMBER 12

letters to the editor

Page 5: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 7

WASHINGTON, DC — The U.S. Consumer Product Safety Commission (CPSC)and Central Sprinkler Company, an affiliate of Tyco Fire Products LP, of Lansdale,PA, are announcing a voluntary replacement program. The company will providefree parts and labor to replace 35 million Central fire sprinklers with O-ring seals.The program also includes a limited number of O-ring models sold by GemSprinkler Company and Star Sprinkler, Inc., totaling about 167,000 sprinklers.

Central initiated this action because it discovered the performance of these O-ring sprinklers can degrade over time. These sprinkler heads can corrode orminerals, salts, and other contaminants in water can affect the rubber O-ringseals. These factors could cause the sprinklers not to activate in a fire. Central isproviding newer fire sprinklers that do not use O-ring seals and is voluntarilylaunching this program to provide enhanced protection to its sprinkler customers.This is the third-largest replacement program in CPSC history.

Central will provide, free of charge, replacement sprinklers and the labor needed to replace the sprinklers. Central will arrange for the installation by usingeither its own Central Field Service crews or by contracting with professionalsprinkler contractors.

For more information, go tohttp://www.cpsc.gov/cpscpub/prerel/prhtml01/01201.html.

JOHNSTON, RI — New fire protection research by industrial and commercialproperty insurer FM Global and its affiliate Factory Mutual Research has identifiedan uncommon but realistic scenario where K14 Early Suppression-Fast Response(ESFR) sprinklers, used in some warehouse protection applications, may fail toperform as intended. Specifically, the findings revealed certain ignition scenarioscan affect the K14 sprinkler’s ability to suppress a fire in buildings over 40 feet(12.2 m) high.

As a result, FM Global no longer recommends to its policyholders the use ofK14 sprinklers as an alternative to in-rack sprinklers in buildings over 40 feet(12.2 m) high. Rather, the company advises that a combination of in-rack andceiling sprinklers be installed according to FM Global property loss preventionengineering guidelines.

The findings are the result of FM Global’s long-range sprinkler technologyresearch program. Additional research has revealed that this storage configurationissue does not affect K25 ESFR sprinklers installed in 45-foot (13.7 m)-high build-ings nor K14 or K25 sprinklers installed in buildings up to 40 feet (12.2 m) high.

For additional installation guidelines on ESFR sprinklers, visitwww.fmglobal.com or call FM Global Customer Services at 781.255.6681.

fire protection industry news

NewResearchResultsPromote K14GuidelineRevision

O-Ring FireSprinklersRecallAnnounced

flashpoints

Page 6: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 9

By Ray Schmid, P.E.

Many believe that the computeris humanity’s greatestachievement. The past thirty

to forty years have witnessed vastimprovement on this amazing techno-logical achievement and the benefitsof advancements in information pro-cessing. Financial transactions, medicalbreakthroughs, power generation, theconstruction industry, national defense,and the globalization of economies areall due to the power of the computerand the sharing of information.

While the most obvious of theseadvancements may be the Internet,advancements in telecommunicationstechnology have similarly revolution-ized how businesses operate. As thesetechnological breakthroughs continue,society becomes more dependent ontheir performance and reliability. Forexample, consider the importance ofcomputer operations on air traffic con-trol, nuclear power generation, nationaldefense, surgical procedures, and finan-cial transactions.

Maintaining the operability of com-puter networks, Web sites, and commu-nications has become an absolutenecessity, and there are a number ofways that businesses ensure their relia-bility. This article will discuss the wayfire protection systems, both passiveand active, are used to meet the impor-tant goal of continuous operation andmission continuity.

FUNDAMENTALS OF MODERNSYSTEMS

In order to fully understand the wayfire protection systems are used to pro-tect modern data centers and telecom-munications facilities, it is important tounderstand the relationship betweenthe computer hardware and the electri-cal and mechanical systems that servethem.

This relationship is most evident bythe amount of heat that is generated bymodern digital equipment, such ascomputer servers and telephone switch-ing equipment. Older systems, as couldbe expected, were much larger andbulkier than today’s digital equipment.Current digital hardware takes up farless space than earlier-generationequipment occupied and generates alarge amount of heat that must be con-tinuously removed. Depending on thespecific equipment and its physicalconfiguration, failure can occur within amatter of ten or fifteen minutes if con-tinuous cooling is not provided. Typicalmechanical systems that provide thiscooling capacity will be discussed inmore detail later in this article.

This “compression of technology”has obvious benefits, but there isanother more intrinsic price to be paid.For example, telecommunicationsswitching sites can now process calls atmuch higher rates in relatively smallfacilities, serving very large geographi-cal areas as a result. Consequently, the

Providing Fire

Protection for

Mission-Critical

Facilities

Telecommunicationsand e-Commerce

Page 7: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

10 Fire Protection Engineering NUMBER 12

importance of an individual sitebecomes that much more critical, sincea fire at a single facility can have sucha widespread effect.

Another example of this is a facilitythat maintains Web site hosting forbusinesses. Web site addresses havebecome as essential to a business as aphone number, allowing clients andcustomers the opportunity to viewproducts and services, and more impor-tantly, make transactions. As the valueof Web sites increases, the need tohave them continuously availableincreases as well, and many businessesexpect guaranteed operability of theWeb hosting center.

Continuous operation of a telecom-munications facility, Web hosting cen-ter, or data center requires reliable,and in many cases, redundant power,cooling, fire protection, and securitysystems. Building and fire codesaddress some of these issues, but theunique requirements of these facilitiesdemand much more. For a number ofyears, the National Fire ProtectionAssociation (NFPA) has publishedNFPA 75, Protection of ElectronicComputer/Data Processing Equipment,to address fire protection requirementsfor computer rooms. Recently, theNFPA created a technical committee toassist in the development of anew standard, NFPA 76,Protection ofTelecommunicationsFacilities. This standard willspecifically address fire pro-tection criteria for telecom-munications facilities usingboth prescriptive and perfor-mance-based approaches.

These NFPA documents arenot the only sources for deter-mining appropriate fire pro-tection requirements. FMGlobal also has requirementsfor these types of facilities thatapply when the facility isinsured by a Factory Mutualaffiliate. These can also serveas good practice for those thatare not insured by an FM affil-iate. The FederalCommunications Commission(FCC) is also concerned withreliability and operability oftelecommunications facilities.In 1992, The Network

Reliability Council, an organizationchartered by the FCC, issued its Reportto the Nation, which included recom-mended fire prevention and protectionstrategies for major telecommunicationsproviders. The FCC continues theefforts of this organization, which isnow known as the Network Reliabilityand Interoperability Council.

Although these regulatory forces existand provide criteria for these facilities,arguably the most compelling factor dri-ving the need for reliable systems ismarket forces. The competitive natureof the industries, be they telecommuni-cations, e-commerce, the stock market,or financial institutions, simply demandsmission continuity.

Mission continuity is assured for facil-ities through the use of redundantpower supplies, redundant mechanicalsystems, and cutting-edge fire protec-tion systems. Primary and secondaryelectrical switchgear, uninterruptiblepower supplies (UPS) and batteries,and standby generators provide self-suf-ficient electrical services. It is commonfor facilities to be equipped with suffi-cient standby generator capacity andfuel storage for more than a week ofsecondary power generation. Similarly,stored water may be provided to servecertain types of cooling systems should

a failure of the public water supplyoccur.

DETECTION STRATEGIES

Sophisticated fire alarm and detectionsystems, fire prevention practices, com-partmentation, fire suppression, and, insome instances, smoke managementalso serve to provide redundant levelsof fire protection. Perhaps the most crit-ical of these fire protection systems isin the area of detection. Detection sys-tems serve the basic function of alertingbuilding occupants of a fire condition,but are also used routinely to controlthe release of fire suppression systemssuch as preaction sprinkler and cleanagent systems. Normally, these functionsare controlled by standard spot-typeionization and photoelectric smokedetectors, although heat detection,flame detection, and other methodsmay be used where they are moreappropriate for the specific hazard orapplication. Standard spot-type smokedetectors and other devices that detectfire conditions prior to the time atwhich they threaten the building oroccupants are referred to as EarlyWarning Fire Detection (EWFD).

Probably the single most importantfactor affecting the design of EWFD

Returnair

flow

Returnair

flow

Ceiling

Floor pedestals not shown for clarityDischarge

Raisedfloor

Supplyair flow,typical

TypicalCRAH

unit

Computerrack

Slab

Figure 1. Typical raised-floor computer space airflow pattern

Page 8: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 11

systems is the airflow conditions withinthe space. For example, Figure 1 showsa typical raised-floor computer spaceand the airflow patterns that are gener-ated by the computer room air-han-dling (CRAH) units. In this example,the CRAH units draw air from a zoneabove the equipment racks (servers,switches, etc.) into the top of the CRAHunit. The air is conditioned and dis-charged at the bottom of the unit intothe raised floor space. When the air isdischarged into the raised floor, a pres-sure is developed which forces the airup through perforated floor tiles.Supply air in the raised floor can be10°C (50°F) or less. As this air filters upthrough the equipment racks andadjoining aisles, the equipment iscooled. This continuous coolingprocess is essential to the data center ortelecommunications equipment. Failureor shutdown of the cooling systemswill require shutdown of the electricalpower systems within a matter of min-utes, which is literally a doomsday sce-nario for the facility.

While the basic concept is straightfor-ward, the airflow patterns raise a num-ber of important issues that affectsmoke detector location and spacing.First, NFPA 72, The National Fire AlarmCode requires smoke detector spacingin areas of high air movement to bereduced. This reduction in normal spac-ing is dependant upon the rate atwhich air is circulated in the space,expressed as air changes per hour (orminute). It would not be at all uncom-mon to have a required spacing of aslittle as 11.25 m2 (125 ft2) per detector.

When designing EWFD for spacescontaining modern computer and digi-tal switching equipment, it is importantto be aware of the differences betweenphotoelectric and ionization detectors,and their performance in areas of highair movement. Photoelectric detectorsperform better than ionization detectorsin detecting smoldering fires that pro-duce larger smoke particles.1

Photoelectric detectors also tend to beless susceptible to high air movement,although there are a number of ioniza-tion detectors designed for installationin areas of high air velocity. Given thatthe types of fires likely to be generatedin data centers and telecommunicationsequipment spaces are low-energy smol-dering fires, it would follow that photo-

electric detectors are probably moreappropriate. In addition, multicriteriadetectors and complex algorithms pro-vided with some detection software arebecoming more popular in mission-criti-cal applications due their ability tomore completely analyze fire signaturesand screen out false signals.

BEYOND TRADITIONAL FIRE DETECTION

A reduction in spacing may be suffi-cient to accomplish the goal of detect-ing a fire condition and releasing a sup-pression system before the occupantsof the facility are threatened. However,EWFD is generally considered to beincapable of detecting an incipient firecondition prior to the time at which itaffects modern digital computer equip-ment. According to the FCC’s NetworkReliability Council Report to the Nation,as much as 95 percent of all damagecaused to computer and digital switch-ing equipment by fires can be charac-terized as nonthermal damage. Whatthis shows is that the biggest risk tocontinuous operation in these facilitiesfrom fire is the smoke, not the fireitself. Smoldering combustion of one ortwo circuit boards may produce a heatrelease rate of one or two kilowatts. Bycomparison, the heat release rate froma typical trash can fire is on the orderof 15 kW2 or higher. However, relativelysmall amounts of smoke andhydrochloric acid, a common byproductof combustion of PVC cables and digitalcircuit boards, can very effectively dam-age digital servers and switches.

Detecting combustion byproductsfrom these low-energy fires requires amore sophisticated technology. Twocommon methods of detecting fires ofthis magnitude are through the use ofair-sampling smoke-detection systemsor high-sensitivity laser spot detectors.Detectors such as these that can detectproducts of combustion before theysubstantially threaten equipment in thespace are referred to as Very EarlyWarning Fire Detectors (VEWFD).When using these systems, the goal isgenerally to provide notification to facil-ity staff that can then intervene toremove the failed unit from service, dis-connect electrical service to the equip-ment rack, extinguish the fire with ahand-held extinguisher, or take other

appropriate action. Due to their highsensitivity, VEWFD are not generallyused to release fire suppression sys-tems; however, in certain applications itmay be appropriate to use them for thispurpose. When using VEWFD torelease suppression systems, isolationof the space from external smokesources and nuisance alarms must becarefully controlled.

Air-sampling smoke detectors areprobably the most common method ofproviding VEWFD. The detection sys-tems normally consist of a detectionunit, an aspirating air pump, a networkof sampling pipes, and related appurte-nances. The detection unit contains theair pump, the detector, power supply,filters, and electronic interfaces to pro-vide annunciation and connection toexternal fire alarm systems or displays.System piping (typically 20 mm [0.75inch] CPVC) is laid out in a systematicpattern, and small holes called sam-pling ports are drilled into the piping.The sampling ports are spaced accord-ing to the rules of conventional spot-type smoke detectors, unless a reducedspacing is desired based upon the riskfactors associated with the hazard beingprotected. For example, the spacing ofsampling ports may be reduced to 18m2 (200 ft2) if earlier detection isdesired. It should be noted that thesedetectors are more reliable in high airvelocity environments and are not gov-erned by the same rules as spot smokedetectors with regard to high air move-ment. The resulting piping and sam-pling ports form a “zone” of detection,since the detector cannot determinewhich sampling port(s) on a given piperun is drawing in smoke.

Figure 2 shows a typical air-samplingsystem piping network. The detectionsystem operates by drawing air inthrough the sampling ports, throughthe piping, and back to the detectorwhere the sampled air is first filteredand then analyzed. The detector thendetermines whether the air sample iscontaminated with smoke or is “clean.”The detector may also have softwarethat can analyze particle size to screenout unwanted alarms such as dust,insects, etc. The air is then vented backinto the protected area. It is importantto note that as smoke is drawn intoone or more sampling ports, it couldbe mixed with noncontaminated air

Page 9: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

Figure 2. Typical air-sampling piping network

FALL 1999 Fire Protection Engineering 15

that is drawn in through other samplingpoints on the same piping run. Thiscan result in the dilution of smoke as itis transported back to the actual detec-tor for analysis. This dilution effectbecomes more pronounced as thenumber of sampling ports is increasedalong a given pipe run. Therefore, areduction in sampling port spacing (i.e.,more sampling ports) needs to be bal-anced with the potential for diluting thesmoke sample. In some cases, this maynecessitate shorter pipe runs, additionaldetectors, or both.

When smoke is identified, the detec-tor is normally capable of displaying anumber of alarm “thresholds” thatdescribe the level of smoke obscurationat the detector. These thresholds can bedisplayed at the specific detector,annunciated remotely, or transmitted toa conventional fire alarm control panel.When connected to a fire alarm controlpanel, the level of smoke obscurationcan be used to perform different firesafety functions, depending upon thespecific alarm threshold reached. Forexample, upon receiving a “Level 1”alarm from the detector, the fire alarmpanel may initiate a supervisory alarm

to alert facility staff of an abnormalcondition at the detector. A subsequent“Level 2” alarm would indicate that thelevel of smoke sampled by the detectorhas increased, requiring compartmenta-tion of the space. This may initiate clos-ing smoke dampers, sending additionalalarms to facility staff, or shutting downthe air supply from outside the protect-ed area. A “Level 3” alarm may result inactivating the fire alarm system notifica-tion devices, releasing fire suppressionsystems, or transmitting an alarm condi-tion to an off-site monitoring facility.

The functions described above arecertainly not unique to air-samplingdetectors. They can also be performedwith high-sensitivity laser spot detec-tors. These detectors function muchmore like conventional spot smokedetectors; however, they are capable ofidentifying smoke obscuration levelsmuch lower than conventional smokedetectors. For example, one manufac-turer produces a detector that is capa-ble of detecting smoke obscuration lev-els as low as 0.009%/m (0.03%/ft ). Attheir most sensitive calibration, conven-tional smoke detectors generally detectsmoke obscuration levels in the neigh-

borhood of 0.15%/m (0.5 %/ft.). Air-sampling systems by comparison maybe capable of detecting smoke obscura-tion levels as low as 0.005%/m(0.0015%/ft.). It is important to recall,however, that there is the potential fordilution of the smoke sample as it isdrawn back to the detector. Therefore,the effective sensitivity at a given sam-pling port may not be substantiallyhigher than that of a spot-type laserdetector. The overall effectiveness ofthe two technologies is very much afunction of the specific application; inparticular, the ambient conditions andthe desired performance objective.

Compared with air-sampling detec-tors, spot laser detectors have severalkey advantages and disadvantages. Oneof these is that they are point-address-able, allowing the fire alarm controlpanel to directly monitor the status ofeach device, perform sensitivity adjust-ments based upon time of day, adjustmonitoring of adjacent or “shared”detectors, and allow for “cross-zoning”of detectors. However, since the detec-tors are spot-type devices, they aresomewhat passive, relative to air sam-pling systems. In other words, they rely

Air-sampling system detection unit

Note: Vented endcap serves as a sampling port

Typical equipmentcabinets (racks)

CRAH

Typicalsampling port

Page 10: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

16 Fire Protection Engineering NUMBER 12

on sufficient thermal energy to be gen-erated by the fire to transport thesmoke to the detectors. Air-samplingsystems can compensate for this trans-port lag somewhat by actively drawingin air from the space. Therefore, air-sampling systems are not totally depen-dent on thermal energy to transportsmoke to the detector.

Since spot-type laser detectors relyon smoke transport and have the abilityto make logical decisions among multi-ple detectors, in some cases it may beappropriate to design these detectionsystems to release fire suppression sys-tems. In doing so, it would eliminatethe need to have a system of EWFD torelease suppression systems and a sep-arate VEWFD system to alert facilitystaff of incipient fire conditions.

One other important aspect ofVEWFD is detection for the return air-flow to CRAH units and other HVACsystems serving the protected area.Since the rate of smoke generation in alow-energy smoldering fire is relativelysmall and the airflow velocities in theprotected space can be quite high, themovement of smoke within the spacetends to be dominated by the airflowpatterns generated by the mechanicalsystems. For this reason, it is essentialto provide VEWFD at the return side ofthe CRAH units, particularly in spaceswith large clearances between the ceil-ing and the top of the equipment.Since the main concept of the CRAHunits is to cool the space, it is unlikelythat any smoke generated will have suf-ficient buoyancy to reach detectionpoints at the ceiling. This is furthercompounded by the fact that the highairflows tend to pull any smoke back tothe individual CRAH units, diluting itwith clean air in the process. This phe-nomenon will be diminished as theceiling height decreases or one or moreCRAH units in the area of fire origin arenot operating. Providing VEWFD exclu-sively at the ceiling level should bedone only when it has been deter-mined that the mechanical systems willnot adversely impact smoke transportto ceiling-mounted detectors or sam-pling ports.

A variety of configurations, designschemes, and good practices should beobserved when designing fire-detectionsystems for mission-critical facilities.While each method of VEWFD has

advantages over the other, clearly themost appropriate system for a givenapplication will depend on the type offacility being protected, the airflow pat-terns of the space, and the specific riskfactors involved.

FIRE ALARM SYSTEM FEATURES

The overall fire alarm system designalso plays an important role in main-taining continuous operation. For larg-er facilities having many detectors, thenumber of addressable points on asystem can reach the thousands.Displaying alarm information clearlythrough the use of graphic displays,PC-based annunciators, and traditionalLCD annunciators should be consid-ered. In addition, providing multipleannunciators or paging systems canenhance the speed with which thefacility staff can locate and isolate thesource of the fire. Identifying eachdetector (or addressable point) by itsroom designation, column grid, andlocation (above ceiling, below floor,etc.) can also decrease the time need-ed to identify the source of the alarmand correct the problem.

Off-site monitoring is also important,particularly for facilities that may not benormally occupied. In many cases, thebuilding code will require off-site moni-toring of the fire alarm and suppressionsystems. Monitoring the system in accor-dance with Central Station requirementsshould be considered in most cases,unless equivalent reliability and perfor-mance can be provided by some othermethod. In some instances, it may bebeneficial to transmit the alarms to aremote facility monitored by the owner.

MANAGING SMOKE MOVEMENT

The fire alarm and detection systemsform the most critical element in pro-tecting mission-critical facilities. Theynot only serve to detect and alert, butalso control the release of suppressionsystems, initiate compartmentation features, and, in some cases, initiatesmoke-management systems. Smoke-management systems continue to gainsupport in protecting these facilitiesdue to the need to protect against non-thermal damage. One obvious way toprevent collateral damage of equip-ment from fire in a particular piece of

equipment would be to simply exhaustthe smoke. In theory, this is a goodidea, but such a system must be care-fully designed. As with any smoke con-trol system design, there is a gap (oroverlap, depending on how it isviewed) in responsibilities amongmechanical, electrical, and fire protec-tion trades. Programming the systemsto close the correct dampers and initi-ate the proper fan sequences are anabsolute necessity, and this program-ming logic must survive the test oftime as mechanical systems are modi-fied to accommodate changes in thefacility’s operation. Failure, in thisrespect, will likely create a situationthat is worse than no smoke manage-ment at all.

In addition, the required zoning ofthe detection and smoke-managementsystems must also be coordinated. Thisis where the flexibility of addressabledetectors becomes clear. With address-able detection, regardless of how thesmoke zones are configured, the firealarm system can be programmed tocomplement the smoke-managementsystem. With zoned detection systems,this is simply not the case, and indi-vidual detection zones must bedesigned in concert with the smoke-management system. Coordinatingzoning is frequently complicated bychanges in wall locations that occurduring construction.

Another consideration in the designof a smoke-management system is thesize of the zone and the locations ofsupply and exhaust points. Exhaustingsmoke within a particular zone must bedone in such a way that smoke will notbe pulled across equipment racks thatare remote from the source of smoke.This may require multiple exhaust airintakes and an analysis of how thesmoke will move from the source tothe intakes. For example, is it reason-able to assume that the smoke will riseup to the ceiling level and be pulledacross the ceiling to the exhaust intake?This is probably not realistic for spacesthat have airflow patterns dominated byCRAH units and a relatively smallamount of smoke-production. In thiscase, the smoke-management systemrequires almost surgical precision in itsdesign and must be carefully tailored tothe geometry of the space and the air-flow patterns present.

Page 11: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 19

PASSIVE FIRE PROTECTIONSTRATEGIES

Compartmentation, housekeeping,and regulation of equipment also playimportant roles in fire protection designfor these facilities. Smoke and fire barriers, dampers, fire doors, and stafftraining are essential to confining prod-ucts of combustion and minimizing thedamage should even a small fire devel-op. It may also be appropriate to locatethe facility in a building of protectedconstruction, particularly where evacua-tion may be impractical or a multitenantoccupancy exists. Selection of equip-ment may also be an issue when thedata center, particularly a Web hostingor colocation facility, is subject to tran-sitory equipment and multiple clients.Equipment should be evaluated toensure it is listed and is constructed ofmaterials that do not have an unreason-able fire potential. Where the perfor-mance of equipment is questionable,segregation or elimination of the com-ponents should be considered.

Similarly, maintaining critical areasfree of materials that simply have noplace in a data center or telecommuni-cations facility is also important.Contractor staging during renovationsor new equipment installation shouldbe confined to appropriate areas, sothat excessive fuel loads that couldovercome the fire protection systemsdo not exist within critical areas.

SPRINKLER PROTECTION

Should a fire develop due to an equip-ment malfunction, poor housekeepingpractices, or even an act of God, sup-pression systems are relied upon tocontrol or extinguish them. In manyways, these systems are a last line ofdefense in this type of facility. If theyactivate, it means that other efforts atdetection, prevention, compartmentation,and control have at least partially failed.It also means that a significant amountof smoke and heat is being generatedthat will destroy electronic equipmentin that space. The only question becomeshow much damage can be limited.

In many cases, compliance withapplicable building or fire codes willrequire sprinkler protection for thebuilding. This could be triggered byprovisions for unlimited area buildings,

requirements for high-rise buildings,height and area restrictions for thebuilding, local code requirements, orperformance-based design goals. Therealso may be a requirement on the partof the building owner, insurance under-writer, or adjacent tenant that the build-ing be fully sprinklered. While the useof alternative suppression systems maybe an acceptable trade-off for specificareas, the use of these systems as analternative to sprinkler systems general-ly requires the approval of the localcode official. A redundant, directly con-nected complement of reserve agentwould normally be required as well.

Regardless of the individual reasonsfor sprinkler protection of a criticalfacility, wet pipe systems may not bethe preferred system. While the perfor-mance record of wet pipe system relia-bility is good and failures are rare, thepresence of charged sprinkler pipingover critical equipment and processescauses a substantial liability for facilityowners. The added level of protectionagainst accidental discharge or leakagethat a preaction sprinkler system pro-vides can be worth the additional up-front installation and long-term mainte-nance costs. By requiring activation ofthe EWFD system in order to chargethe sprinkler system with water, anadditional level of protection againstunintended discharge is provided in theevent piping is damaged by operationsin the data center. A double interlocksystem also permits the sprinkler systempiping to be monitored for integritythrough the use of low-air pressurealarms connected to the facility fire alarmsystem. However, since this type of pre-action system is more complex, addi-tional failure modes are introduced,making proper inspection, testing, andmaintenance of the systems more critical.

There are also sprinkler systems thatare designed to cycle water flow onand off, depending on the fire condi-tions within the protected area. Thesesystems can limit the collateral damagesprinkler system runoff may cause out-side the protected area, such as anadjacent electrical switchgear room orbattery room. This added benefit comeswith an additional cost, since cyclingsystems generally require separatedetection systems in addition to theEWFD system. In the past, anothermethod of controlling excessive water

damage was to use sprinklers thatwould cycle on and off based upon thetemperature at the individual sprinkler.However, reported problems with leak-age have reduced the popularity of“on-off’ sprinklers, and at least onemajor manufacturer has stopped pro-duction due to low market demand. Asindicated previously, any expected col-lateral damage from a catastrophicevent or component failure should becompared with the potential decreasein reliability associated with a morecomplex system or device.

When designing preaction or cyclingsystems, it is important to consider thesprinkler system zones and locations ofsystem feed mains in the designprocess. Sprinkler zones should begrouped logically based on buildingconfiguration, so that if a fire doesoccur, responding firefighters canquickly determine in which area of thefacility the alarm has occurred. This istrue of any sprinkler system design, notjust mission-critical facilities. The loca-tions of feed mains should also becoordinated with medium- and high-voltage electrical switchgear and unin-terruptible power supply (UPS) mod-ules. In certain instances, NFPA 70, TheNational Electric Code, prohibits theinstallation of sprinkler system pipingfrom passing through rooms containingthis type of electrical equipment. Insuch an instance, the only piping per-mitted would be the piping that wasactually serving sprinklers within theelectrical room or enclosure. It is agood practice to route distributionmains to individual preaction valves orriser rooms around critical electricalequipment rooms, so that if a leakwere to develop, it would occur withina noncritical area, such as a corridor,office, or service area.

Another important consideration insprinkler system design for these facili-ties is the need to pitch the system pip-ing to the main drain. This is recom-mended even if the system piping isnot installed in an area that is subject tofreezing. The removal of water follow-ing a hydrostatic or trip test, or theflushing of the system is importantsince even a small amount of water hasthe potential to damage sensitive elec-tronic equipment. Trapped sectionsrequiring auxiliary drains create anotheropportunity for water to be present in

Page 12: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

20 Fire Protection Engineering NUMBER 12

the system if not completely drainedfollowing a test. In some cases, howev-er, space limitations may require that anauxiliary drain be installed. In these sit-uations, a good practice would be tolimit the number of auxiliary drainsneeded through careful design.Auxiliary drains should also be clearlymarked, even to the extreme, and extraeffort should be made to ensure theyare opened following any testing or tripof the system. Where pendant sprin-klers are required, it may be desirableto install dry pendant sprinklers.Although an expensive option, thiswould eliminate the presence of waterin individual sprinkler drops.

CLEAN AGENT SUPPRESSIONSYSTEMS

In many cases, clean agent suppres-sion systems can provide a level of firesuppression performance that sprinklersdo not, allowing critical systems to con-tinue to operate during system dis-charge. They also require very littlecleanup following a discharge and aregenerally safer for building occupantsthan carbon dioxide or traditional inertsystems. Two of the more commonsuppression agents are FM-200™ andInergen™, although FE-13™, Argon,and others exist as well. Both systemsoperate in a manner similar to Halonsystems in that the agent is stored infixed containers and is dischargedthrough fixed piping to discharge noz-zles. The properties of these alternativeagents, however, do not allow them tobe used as a direct substitute for exist-ing Halon installations. Therefore, pip-ing systems, agent storage containers,nozzle placement, and system hardwaremay need to be redesigned whenreplacing an existing Halon installation.

Unlike sprinkler systems, clean agentsystems are normally designed to dis-charge solely upon activation of theEWFD system within the protectedarea, permitting the suppression agentto be distributed early in the firegrowth period. This allows the agent tosuppress the fire before the heatrelease rate is low. In addition, thepresence of a VEWFD system coupledwith manual release stations allow facil-ity staff the opportunity to dischargethe agent even sooner in the firegrowth period. Of note is that if the fire

is of an electrical nature, as can be thecase with cable fires, the energy sourceshould be interrupted in order for thesuppression system to be effective.

One very unique feature aboutInergen™ systems is the ability foroccupants to remain in the protectedarea following system discharge.Inergen™ is actually a combination ofthree gases (nitrogen, argon, and car-bon dioxide) that basically inert theprotected volume. Depending upon thedesign concentration selected, thiswould typically reduce oxygen concen-trations within the protected area toapproximately 12 percent, which is suf-ficient to extinguish fires involving mostordinary combustibles. While this con-centration is also less than that requiredfor humans to survive, the presence ofadditional carbon dioxide stimulates thebody to breathe more deeply, increas-ing the absorption of oxygen by thebody. This physiological effect allowshumans to breathe normally, even in anoxygen-depressed atmosphere.

Another key difference between FM-200™ systems and Inergen™ systems isthe delivery method and agent dis-charge time. FM-200™ systems aremore like Halon systems in this regardand require the agent storage contain-ers to be located inside or within closeproximity to the protected area. This isdriven by the dynamics of the two-phase flow that occurs when the sys-tem discharges. Also, since FM-200™ isa halogenated system, it must be com-pletely discharged within 10 seconds.Inergen™, on the other hand, is storedand discharged as a gas. It operates at ahigher pressure (approximately 15 MPa[2,175 psi] vs. 2.5 MPa [360 psi]), whichallows the agent to be piped over asubstantial distance. Since it is an inert-ing system, the maximum dischargetime permitted by NFPA 2001,Standard on Clean Agent FireExtinguishing Systems, is 60 seconds.

There are many differences betweenthe two systems that will affect whetheror not they are appropriate for a givenfacility. Selection of equipment, overallperformance, agent storage methods,and cost are but a few examples. Bothsystems, however, are widely accepted.In addition to clean agents, water-misttechnology may be an acceptable alter-native for certain applications. Althoughwater mist systems have been shown to

be effective in localized extinguishmentof fires involving electronic equipment,their use as a total flooding system isstill relatively new and unproven.Water-mist systems are an emergingtechnology that may prove to be aneffective system for critical facilities inthe near future.

Clean agent suppression and similarfixed systems necessarily rely on theintegrity of the enclosure to help main-tain appropriate concentrations of theagent and effectively control or sup-press the fire. If the integrity of theenclosure is not maintained (for exam-ple, opened doors or holes in wallswhere cables have been run), then theagent would be less effective, or noteffective at all. This is one reason thatlocal authorities may not permit a direct“trade-off” between alternative suppres-sion systems and traditional sprinklersystems.

CONCLUSION

The fire protection methods that havebeen discussed by no means represent acomplete list of all the systems and strat-egies that are available to the fire protec-tion professional. The best overall fireprotection strategy for a specific facilityis very much a function of acceptablerisk levels, minimum code requirements,and interoperability of systems. Systemdesigns should be based upon a totalfire protection approach, not simplyindividual systems pieced together bydifferent trades. Only a total fire protec-tion concept approach will integratewith the facility’s goal of ensuring con-tinuous operation.

Ray Schmid is with Koffel Associates,Inc.

REFERENCES

1 Mulholland, G., “Smoke Production andProperties,” SFPE Handbook of FireProtection Engineering, Second Edition,National Fire Protection Association,Quincy, MA, 1995.

2 Babrauskas, V., “Burning Rates,” SFPEHandbook of Fire Protection Engineering,Second Edition, National Fire ProtectionAssociation, Quincy, MA, 1995.

For an online version of this article, goto www.sfpe.org.

Page 13: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 23

By John A. Alderman, P.E., CSP, and Marlon Harding, P.E.

ABSTRACT

Most companies have standards forfire protection on offshore installations.These standards typically are prescrip-tive in nature and require that fire pro-tection be installed, generally, withoutregard to the actual hazard. Fire protec-tion generally consists of passive fireprotection, water and/or foam systems,and detection systems.

The fire protection community isslowly beginning to use performance-based criteria in determining appropri-ate fire protection. Performance-basedcriteria that uses the latest modeling

programs to assess the hazards of firesand explosions on the offshore installa-tion are beginning to grow. Based onthe results of the risk calculations, thefire protection engineer can then deter-mine the appropriate fire protectionrequired for the hazard.

This paper addresses the use of per-formance-based criteria for fire protec-tion in the offshore industry.

INTRODUCTION – MORE IS NOTALWAYS BETTER

During the Piper Alpha incident in1989, 168 people lost their lives as aresult of an explosion and resulting fire.However, more fire protection mightnot have helped in that incident if it

was not adequately matched with theactual risks associatred with the facility.

The amount of fire protection neces-sary in an offshore environment hasalways been subject to debate. Under-protection can lead to potential loss oflife and property loss resulting inreduced production and possible envi-ronmental impact. Overprotection canlead to increased cost and maintenance.More importantly, protection based onstandards rather than the actual hazardcan lead to a false sense of security bymanagement that the facility is safe.

PRESCRIPTIVE FIRE PROTECTION

With prescriptively designed fireprotection, protection systems are

OFFSHORE INDUSTRY

Proactive vs. Prescriptive Fire Protection for the

Page 14: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

24 Fire Protection Engineering NUMBER 12

installed based on specific guidance orrequirements without much deviation.For example, one company providestwo 0.157 m3/s (2,500 gpm) fire pumpson all platforms, without considerationof the size of the platform or waterdemand. Prescriptive approaches tofire protection generally are a result ofregulation, insurance requirements,industry practice, and company proce-dures. Table 1 illustrates examples ofprescriptive approaches to fire protec-tion. Each of these tend to be basedon past incidents rather then trying tolook forward and determine whatcould happen.

APPROACH TO PERFORMANCE-BASED FIRE PROTECTION

Performance-based fire protection isdetermined by conducting some formof analysis or calculations to define thefire protection required to mitigate thehazards. The risk analysis process fre-quently used in the offshore industry isillustrated in Figure 1.

HAZARD ANALYSIS

The first step in any performance-based approach is to conduct a hazardanalysis. The hazard analysis tech-niques used to identify potential haz-ards in the process and facility areshown in Table 2. Typical offshorefacilities where hazard analyses are per-formed include platforms; Floating,Production, Storage, and Offloadingfacilities (FPSO); Floating, Storage,Offloading facilities (FSO); drilling(such as jack-up, semi, or ship); Sparsor Tension Leg Platforms (TLP); or anycombination of these.

The outcome of the hazard analysisis a list of potential fire hazards thatmay occur on the facility. A partial listcould include jet fire, pool fire, explo-sion, electrical fire, or Class A fire. Thelist would also include the correspond-ing location where each could occur.These hazards can then be turned intoscenarios for further analysis.

CONSEQUENCE ANALYSIS

Consequence analysis is the processto determine the impact of the scenar-ios. For example, one scenario could

be a seal failure that results in a vaporcloud forming with an explosion in theseparation area if ignited. In assessingthe consequences, two questions needto be answered:

• What is the range in size of theevents that can occur?

• What is the impact of the event? In assessing the impact, radiant heat,

overpressure, toxic effects on the tem-porary refuge, evacuation routes,escape equipment, and offshore equip-ment that could be involved in escala-tion are normally taken into account.

Toxic effects can include products ofcombustion from fires, such as smoke,carbon monoxide, and hydrogen sulfidecontained in the material.

In performing any consequenceassessment, analytical tools can bevery useful to determine the conse-quences of a scenario. In most cases,each scenario will have a variety ofconditions that need to be evaluatedin the consequence assessment. Theseinclude factors such as size of therelease, orientation of release, temper-ature and pressure of operation, andweather conditions (that will all vary).

During the consequence analysis, itis necessary to determine the neces-sary sophistication of the models thatwill be used. Programs range fromspreadsheets that use simple equationsto Computational Fluid Dynamics(CFD) modeling that can take a day

Table 1. Examples of Some Prescriptive Requirements

SOURCE REQUIREMENT

Regulation Mineral Management Service (MMS)US Coast GuardUK Health and Safety Executive (HSE)

Insurance Active fire protectionPassive fire protectionSafety systemsSpecific equipment requirements for compressors and heaters

Industry Practice American Petroleum Institute (API)• API 500 Electrical Classification• API 2030 Water Spray• API 2018 Fireproofing• API 2031 Gas Detection

International Maritime Organization – Safety of Life at Sea (SOLAS)Classification Societies

• American Bureau of Shipping (ABS)• Lloyd’s Register• Det Norske Veritas (DNV)

National Fire Protection Association (NFPA)• Fire Extinguishers• Carbon Dioxide Systems• Sprinkler Systems• Water Spray Systems• Fire Water Pumps

Company Requirements Standards or procedures for:• Equipment spacing• Electrical area classification• Water spray and sprinklers• Fireproofing• Safety shutdown systems• Isolation and blowdown• Relief and flare design• Pressurization systems• Drainage

Page 15: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

26 Fire Protection Engineering NUMBER 12

for one scenario evaluation. The modelsused depend on the level of the design,the time available to do the analysis,and the desired results. In the conceptu-al or feed-stage, simple models can beused, but as the design details increase,the complexity of the consequenceanalysis will also need to increase.

LIKELIHOOD DETERMINATION

If installed fire protection would bebased on only the consequence analy-sis, then the offshore industry would bevery well protected. In reality, the likeli-hood of the consequences must betaken into consideration. In determiningthe likelihood of the consequences, cer-tain key information is required, such asthe frequency of the initiating event, fre-quency of ignition, probability of escala-tion, likelihood that the weather will befavorable or not, etc. In any likelihooddetermination, there are generally alarge number of scenarios to be ana-lyzed.

Computer models are often used toperform the iterative process of evaluat-ing each variable of each scenario.

RISK

Risk is the product of consequenceand likelihood of each scenario. The riskfor each scenario can be combined byspecific areas or for the whole facility toobtain desired risk profiles. The risk iscalculated using event and fault trees thattake into account safety and protectionsystems.

The main problem associated with anyrisk assessment is the appropriateness ofthe data used in the calculations.Obviously, the use of generic industrydata may result in risk numbers that varywidely. It is best if company-specificdata can be used.

Identify hazards

Identify initiating events/scenarios

Determine frequencyof initiating events

Establish risk to personsfor each initiating event

Determine physicalconsequences of scenarios

Establish death andserious injury resulting

at time of initial incident

Assess escalationconsequences

Evaluate effectivenessof protective measures

Determine total andindividual risk

Assess risk and impairmentfrequencies against acceptance

criteria and performance standards

Is riskacceptable?

Determine potentialmodifications to

reduce risk

StopYes

No

Figure 1. Risk Analysis Process

Table 2. Hazard Analysis Methods

• Checklist

• Hazard identification

• What-if?

• Hazard and operability study

• Failure modes and effects analysis

Page 16: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 27

RISK TOLERANCE

After the risk is calculated, theresults must be compared to eithergovernmental or company criteriato determine if the risk is tolerable.If it is, then additional fire protec-tion is not required and the level offire protection used in the risk cal-culation is adequate.

If the level of risk does not meetthe risk criteria, then additionalprotection may be required. Theoptions for reducing the risk areselected and the analysis recalcu-lated to determine the impact onthe risk. In some cases, theoptions (for example, fireproofingon a quarters wall to reduceimpact of jet fire) provide signifi-cant risk reduction, whereas others(water spray of offshore vessels toprotect from jet fire) have very lit-tle impact on the risk.

One concept that has been usedextensively in the North Sea is “aslow as reasonably practical” (ALARP). Figure 2 shows theALARP concept. This concept suggests that at some point thecost to mitigate a hazard is so high that it is no longer practi-cal to implement the option.

OVERPRESSURE EXAMPLE

On facilities that produce hydrocarbons, there is alwaysthe potential for an explosion. The layout of a facility canexacerbate or reduce the impact of an explosion. Since thelayout of facilities may change several times during thedesign, the explosion analysis needs to be rerun each timethe design changes. The analysis will verify that an explo-sion can occur, calculate the resulting overpressure, anddetermine the design criteria for the blast wall. Using costlyand time-consuming CFD modeling is not practical in thissituation.

PC-CHAOS has been developed by AdvanticaTechnologies through extensive full-scale explosion testing.In order to use PC-CHAOS, a three-dimensional model of thefacility is required. This can easily be input from electronicfiles or by hand, if necessary. The program performs multi-ple runs of scenarios and can quickly be changed and rerunfor layout changes.

The results of the calculations can be used to determine ifa blast wall is required, the ideal location for the wall, andthe resulting design criteria.

John Alderman and Marlon Harding are with Risk,Reliability, and Safety Engineering.

For an online version of this article, go to www.sfpe.org.

Unacceptable region Risk cannot be justifiedexcept in extraordinarycircumstances

The ALARP or tolerabilityregion (risks undertakenonly if a benefit is desired)

Tolerable only if furtherrisk reduction isimpractical, or the cost is not proportionateto the benefit gained

Broadly acceptable region Negligible risk

Figure 2. ALARP concept

AQUA TANKFIRESimply The Best Quality By Design

• AC DC & Solar Battery115V 24 System230V Volt

• Tanks Can be Manifolded

• Lightweight

• Easy to Transport

• Dealer Inquiries Invited

Self-Contained Residential Fire Sprinkler SystemPackaged Pump & Aqua Tank System

• NFPA• 13D-13R System

• One-Man Installation

• Costs Less

• Easy to Install

• Less Freight to Pay

• New or RetrofitApplications

• Call for a FreeInformation Package

300GAL.

33" WIDE

1000GAL.

450GAL.

800GAL.

300GAL.

Tanks, From 30" Wide to go through DoorsWorld’s First Solar Electrical Powered Fire System

Simply the Best Quality by DesignAdvanced Fire Technology, Inc.

Phone: 800-536-7032 Fax: 314-351-3831

Page 17: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

30 Fire Protection Engineering NUMBER 12

By Jane I. Lataille, P.E.

When completed, the semi-conductor chips in comput-ers and other electronic

devices can withstand the nor-mal work environment. Butwhile these chips are beingmade, they are highly sensitiveto damage from even thesmallest particles. This is whythe semiconductor industrymust process chips in roomswith highly controlled environ-ments called cleanrooms.

Despite their high level ofcleanliness, cleanrooms stillhave fire protection concerns.The wet benches used to cleanchips between process steps arenot only made of plastic, theycontain heaters to heat baths ofcleaning solvents and wiring tocontrol automated processes.The concern is that a small wispof smoke from overheated plas-tic can ruin millions of dollarsof chips in process.

The semiconductor industryhas long been aware of the riskassociated with using plasticmaterials in cleanrooms. Oneway it has sought to manage this risk isby limiting the use of plastics inprocess equipment. However, some ofthe chemicals used in processing chipsare not compatible with any othermaterial. The semiconductor industry istherefore supporting the developmentof new plastics.

TEST BACKGROUND

Until recently, the only test for evalu-ating the combustibility of wet benchplastics was the Factory MutualResearch Corporation (FMRC) TestStandard 4910.1 This test measures firecharacteristics in a specially modifiedcalorimeter. Fire propagation, smokedamage, and corrosion damage indicesare then calculated from the measure-ments. It has been found that materialswith a fire propagation index of lessthan 6 in this test are not self-propagat-ing in the referred Parallel Panel test.

As the search for better plastics esca-

lated, numerous plastics manufacturersexpressed the need to test their ownproducts during development. Theywere not able to run the FM 4910 testbecause the nonstandard test apparatusis not readily available and the testresults were not reproducible. In addi-tion, the test is complex and expensive.At the semiconductor industry’s request,IRI and UL teamed up to explorewhether a simple, reproducible testusing standard apparatus could bedeveloped. The ASTM E1354 testmethod for the Cone Calorimeter wasselected for investigation.

With input from the semiconductorindustry, IRI and UL decided to basethe classification system resulting fromthe new test on fire propagation andsmoke damage properties.

The project was divided into six tasksas follows:

1. Characterize the Parallel Panel testignition source.

2. Conduct Parallel Panel tests onrepresentative plastics.

3. Conduct ASTM E1354 conecalorimeter tests on the same plas-tics.

4. Correlate Parallel Panel and conecalorimeter test results.

5. Develop a classification schemefor the combustibility of semicon-ductor plastics.

6. Develop a UL standard for classify-ing semiconductor plastics.

Tasks 1, 2, and 4 were necessary toconfirm that results from the new testwould correspond with results from thelarger-scale Parallel Panel test and todevelop a meaningful classificationscheme.

Samples of the following eight mate-rials were tested:

• Polypropylene• Fire-retardant polypropylene• Takiron PVC™• Corzan™

Page 18: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 31

• Kynar HFP™• Clear PVC• Polycarbonate• Halar 901™

This article summarizes the results ofthe six tasks in this project. For morecomplete information, see the ULreport.2

CHARACTERIZING THE PARALLELPANEL TEST IGNITION SOURCE

FM Test Standard 4910 describes theParallel Panel test apparatus, whichconsists of metal frames for holdingtwo 2-ft by 8-ft (0.6 m by 2.4 m) verti-cal panels 1 ft (0.3) apart. A 1-ft by 2-ft(0.3 m by 0.6 m) sand burner usingpropane fuel generates a 60 kW igni-tion source between the panels.

Both heat flux and smoke releasefrom the sand burner were measuredwith noncombustible panels in the testframe. Heat flux data were used toselect the radiant heat flux for exposingthe plastics samples in Task 2. Smokerelease data were used to compensatefor differences in smoke generationbetween the Parallel Panel and ASTM E1354 Cone Calorimeter tests.

CONDUCTING PARALLEL PANELTESTS ON THE SAMPLES

Three samples of each type of plasticwere tested. Each test measured oxy-

gen concentration, exhaust gas temper-ature, exhaust gas velocity, and smokeobscuration. Flame propagation wasnoted for each test.

The heat release rate was calculatedfrom these measurements by means ofthe oxygen consumption techniqueusing the following equation:

Where:= Heat release rate (kW)

kp = Constant (kJ × K/m3)= Volumetric flow rate (m3/s)

T = Exhaust gas temperature (K)x = Instantaneous mole fraction

of oxygen

The value of the constant kp includesfactors for the heat release per kg ofoxygen consumed, the ratio of the mol-ecular weight of oxygen to air, the den-sity of air at ambient temperature andpressure, the ambient temperature, andthe calibration constant for the parallelpanel apparatus.

The smoke release rate was calculat-ed from the following equation:

Where:= Smoke release rate (m2/s)= Volumetric flow rate (m3/s)

= Path length (m)I0 = Reference light beam signal

(V)I = Instantaneous light beam

signal (V)

The total smoke released was calcu-lated as a time integral using the trape-zoidal method. The specific extinctionarea is then the total smoke releaseddivided by the sample mass loss:

Where:= Specific extinction area

(m2/g)= Rate of smoke release (m2/s)= Sample mass loss (g)

Table 1 shows the average results ofthe Parallel Panel test for each type ofplastic.

CONDUCTING THE ASTM E1354CONE CALORIMETER TEST

This test measures oxygen concentra-tion, exhaust gas temperature, pressuredifference across an orifice plate, andsmoke obscuration. The heat releaserate was calculated from these mea-surements by means of the oxygenconsumption technique using the fol-lowing equation:

˙˙( . )( . . )

q kV x

T xp= × −−

0 20951 105 1 5

σ =∫ sdt

m

t f

0

TABLE 1Average Results of Parallel Panel Test

Flame Peak Heat Peak Smoke Total Sample SpecificPlastic Propagation Release Release Smoke Mass Ext. Area

[Ft (m)] Rate (kW) Rate (m2/s) (m2) Loss (g) (m2/g)

Polypropylene >8 (>2.4) * * * * *

Fire-retardant polypropylene >8 (>2.4) * * * * *

Takiron PVC™ 3.8 (1.2) 129 17.3 6107 7.88 0.759

Corzan™ 4.0 (1.2) 122 6.9 2719 8.03 0.339

Kynar HFP™ 6.5 (2.0) 219.3 18.3 4914 8.64 0.566

Clear PVC 4.2 (1.3) 192 26.1 11,420 8.94 1.275

Polycarbonate >8 (>2.4) * * * * *

Halar 901™ 3.0 (0.9) 105.3 13.1 5386 6.06 0.890

* Tests terminated when flame height exceeded 8 ft (2.4m).

˙˙

lnsV I

I= ×

l

0

V

q

sV

l

σ

sm∆

Page 19: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

32 Fire Protection Engineering NUMBER 12

Where:= Heat release rate (kW)

kc = Constant = Pressure difference across

orifice plate (N/m2)T = Exhaust gas temperature (K)x = Instantaneous mole fraction

of oxygen

The value of the constant kc includesfactors for the heat release per kg ofoxygen consumed, the ratio of the mol-

ecular weight of oxygen to air, and thecalibration constant for the ConeCalorimeter.

The total heat released was then cal-culated as a time integral using thetrapezoidal method. Smoke release rate,total smoke released, and specificextinction area were calculated thesame way as in the Parallel Panel test.

Table 2 shows the average results ofthe Cone Calorimeter test in the hori-zontal orientation for selected proper-ties of the tested samples. Several otherproperties were measured or calculatedin these tests. The tests were also donein the vertical orientation.

CORRELATING PARALLEL PANELAND CONE CALORIMETERRESULTS

The data collected in the ParallelPanel and Cone Calorimeter tests wereused to determine the ThermalResponse Parameter (TRP), FirePropagation Index (FPI), and SmokeDamage Index (SDI) as specified in theFM 4910 test. The TRP was determinedfrom the following equation:

TABLE 2Average Results of Cone Calorimeter Test in Horizontal Orientation

Peak Heat Total Heat Peak Smoke Total Sample SpecificPlastic Release Unit Area Release Smoke Mass Ext. Area

Rate (kW) (kW/m2) Rate (m2/s) (m2) Loss (g) (m2/g)

Polypropylene 6.0 179 0.11 29.3 52.4 0.557

Fire-retardant polypropylene 5.4 137 0.24 58.7 51.6 1.139

Takiron PVC™ 1.6 45 0.13 37.2 67.3 0.552

Corzan™ 0.7 24 0.06 6.9 72.5 0.094

Kynar HFP™ 1.4 70 0.13 46.6 111.4 0.418

Clear PVC 1.7 64 0.25 72.5 75.5 0.960

Polycarbonate 2.5 182 o.14 88.0 95.8 0.918

Halar 901™ 0.3 9 0.26 90.8 83.5 1.088

TABLE 3Comparison of Parallel Panel and Cone Calorimeter Tests

Peak FPI Peak FPI Peak SDI Peak SDIPlastic Parallel Panel Cone Calorimeter Parallel Panel Cone Calorimeter

(horizontal) (horizontal)

Polypropylene * 20.5 * 1.34

Fire-retardant * 21.5 * 2.88polypropylene

Takiron PVC™ 4.5 5.0 0.39 0.32

Corzan™ 1.0 0.5 0.04 0.01

Kynar HFP™ 8.5 8.0 0.57 0.39

Clear PVC 13.5 14.0 2.00 1.58

Polycarbonate * 9.0 * 0.97

Halar 901™ 2.0 1.0 0.22 0.13

* Tests terminated wen flame height exceeded 8ft (2.4m).

˙( . )

( . . )q k

P

T

x

xc= × × −−

∆ 0 20951 105 1 5

TRPm

= 4 1π

( )kJ m K kg( ) /×q

∆P

Page 20: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

TABLE 4UL 2360 Prescriptive Classification Scheme

Acceptance Criteria

Class 1: Class 2: Class 3:Non- Limited Slow

Test Property propagating Propagating Propagating

ASTM E1354 FPI 6 or less Parallel Panel Parallel Panelrequired required

SDI 0.4 or less 0.4 or less <1

Parallel Panel Propagation (ft) 4 or less 8 or less 8 or less at10 minutes

Pooling of melted No No Nomaterial

FALL 2001 Fire Protection Engineering 33

This equation is derived from equa-tions for flame height and flame propa-gation rate, and it is arranged to usemeasured and calculated data. (See theUL report for details.) The variable m isthe slope of the line of least square fitin a plot of 1/tig1/2 vs. radiant heat flux.The tests measured the time to ignition,tig. The radiant heat flux was calculatedfrom test measurements.

The FPI is then calculated as follows:

Where:k = Constant (1200 for Calorimeter

horizontal orientation, 1000for Parallel Panel)

= Peak heat release rate perunit sample area (kW/m2)

TRP = Thermal response parameter(kW × s1/2/m2)

Finally, the SDI is:

Where:= Specific extinction area

(m2/kg)

Table 3 compares the FPI and SDIcalculated from the Parallel Panel andCone Calorimeter tests.

The table shows that FPI and SDI val-ues obtained in the Cone Calorimetercorrespond well with those obtained inthe Parallel Panel test.

DEVELOPING A CLASSIFICATIONSCHEME

Using the results of the eight plasticstested, UL developed two classificationschemes, one a prescriptive schemeand one performance-based. The pre-scriptive classification scheme is shownin Table 4.

The performance-based classificationscheme lists the properties of the testedplastics for use in fire hazard assess-ments. The properties are determined ata radiant flux of 50 kW/m2.

Properties reported include the fol-lowing:Ignition Properties

• Critical Flux• Thermal Response Parameter

Combustibility Properties• Mass Loss• Heat Release• Smoke Release• Effective Heat of Combustion• Specific Extinction Area

DEVELOPING THE UL STANDARD

To conclude the project, UL devel-oped the standard UL 2360 – StandardTest Method for Determining theCombustibility Characteristics ofPlastics Used in Semiconductor ToolConstruction. The first edition wasissued on May 10, 2000. The standardwill be reissued with minor revisionssometime in 2001.

CONCLUSIONS

UL 2360 is a simple, reproducible testfor the combustion properties of plas-tics. It was developed from test infor-mation on plastics used in making wetbenches for semiconductor industrycleanrooms. However, this test can alsobe applied to plastics used in othersemiconductor tools, such as waferstockers and tools for metal vapordeposition.

Wet benches are just one source ofrisk in the semiconductor industry.See NFPA 3183 for information aboutother fire protection concerns of thisindustry.

Jane I. Lataille was formally withIndustrial Risk Insurers.

REFERENCES

1 FM 4910, FMRC Clean Room Materials –Flammability Test Protocol, Factory MutualResearch Corporation, Norwood, MA.

2 Report on the Research Investigation of the Combustibility of Plastics Used forSemiconductor Tools, UnderwritersLaboratories Inc., Northbrook, IL.

3 NFPA 318, Standard for the Protection ofCleanrooms, National Fire ProtectionAssociation, Quincy, MA, 1998.

For an online version of this article,go to www.sfpe.org.

FPI kQ

TRP= ′′( . ˙ ) /0 42 1 3

SDI FPI= σ8500

˙ ′′Q

σ

Page 21: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 35

By George Faller, C.Eng

1. IntroductionThe “Caixa Galicia” is a prominent Spanish bank

that has traditionally promoted the arts in Galicia, aregion in the northwest of Spain. In the mid-1990sthe “Fundación Caixa Galicia” expressed their inten-tion to build a new cultural centre to display itsimpressive collection of local artwork.

The building that was envisioned was one thatwould be accessible to the public at street level,one that would have a solid appearance but be fullof light, and the intention was that the buildingshould be a work of art in itself.

The site measures approximately 20 m across by30 m long, and the new building will be hemmedin between two existing six-story buildings. Thenew building has six-stories above ground level,with a roof height to match the eaves level of theadjacent buildings. The ground to fourth floor levelsconsisted of galleries and associated public areas,with the top two levels dedicated to administrativeuse. To meet the functional area requirements of thebrief on a restricted site, four basement levels wereintroduced to accommodate additional public galleryspace, an auditorium, and a services plant level.

An important feature of the design was an atriumthat bisects the building, allowing natural light topenetrate at all levels over the full height and depthof the building. The atrium forms a “canyon” abovethe public thoroughfare at street level, dividing thebuilding in two parts, and open circulation bridgeslink the accommodation on either side. Due to thespace restrictions of the site, the two escape stairsfrom the upper levels have been superimposed,one above the other in a “scissors” arrangement,and located to one side of the atrium. One of thetwo stairways in the “scissors” arrangement is a

FIRE SAFETY DESIGN OF THE

FUNDACIÓN CAIXA GALICIABUILDING IN SPAIN

Figure 1. Caixa Galicia atrium

Page 22: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

36 Fire Protection Engineering NUMBER 12

protected stair, the other an open stair.Escape from the galleries located onthe side remote from the protected stairis via open bridges through the atrium.

2. Fire Safety DesignThe fire safety design for the building

was based on the Spanish nationalcode NBE-CPI/96.1 This code had noprescriptive guidance that adequatelyaddressed the fire safety issues present-ed by this design. Key elements of thedesign were evacuation of the uppergallery areas via open bridges throughthe atrium, evacuation of a 300-seatauditorium in the third basement level,full-height glazing separating the gal-leries from the atrium, and appropriatefire-resistance requirements for thestructure and separating elements.

3. Smoke Control in the AtriumEvacuation of galleries on the first to

fourth floors of the building takes placevia two open bridges at each level,which link the two sides of the build-ing either side of the atrium. These gal-leries are served by two protected stair-ways, both of which are on the sameside of the atrium. There is thereforethe possibility that people would haveto travel over the open bridges andthrough the atrium “canyon” to the pro-tected stairs on the other side to maketheir escape.

The accumulation and control ofsmoke in the atrium was a fundamentalissue that had to be addressed toensure safe egress in the event of a fireat one of the lower levels. The firestrategy attempted first of all to min-imise the risk of smoke ingress into theatrium by a combination of:

i. fire-resistant separation of fire loadsfrom the atrium, and

ii. local smoke control with mechani-cal extraction.

Smoke from a fire in the accommo-dation on either side of the atrium wascontrolled at all levels by a combina-tion of these two methods. The atriumbase, however, could not be treated inthe same way. Although it is a “street”thoroughfare, which under normal cir-cumstances would be free of any fireload, there is always the possibility thatsomeone could, either inadvertently ormaliciously, introduce a fire load intothis space.

To investigate the effects of this sce-

nario, a 1.5MW design fire wasassumed, representing about 3m2 of atypical retail premises fuel load with amaximum heat release rate of 500kW/m2. It would be difficult to imaginea fire source of this magnitude beingleft unattended in the “street” area for abuilding such as this with a high levelof security and management, but anumber of design guides stipulate thisas a minimum design fire size.

Using automatically driven roofvents, a system of natural ventilationwas used to manage the smoke fromthis design fire. The natural ventilationsystem was designed so as to ensuretenable conditions within the atrium fora period of time well in excess of thecalculated egress time.

It was found to be impractical toremove smoke at such a rate that itwould not build down to any open

Figure 2. Smoke in atrium from a fire in the open “street”

Page 23: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

What is The Society of Fire Protection Engineers (SFPE)?SFPE, established in 1950, is a growing association of professionals involved in advancing the scienceand practice of fire protection engineering and fostering fire protection engineering education.

What are the benefits of SFPE membership?The Society will provide you with many new opportunities for professional advancement, education,and networking. The specific benefits members receive are:

Free access to SFPE’s periodicals This includes:▲ Fire Protection Engineering magazine. ▲ The peer-reviewed Journal ▲ SFPE Today - Our bimonthly Society newsletter. of Fire Protection Engineering.

Substantial discounts on continuing educationThis includes:▲ Technical symposia on current fire protection issues.▲ International conferences on state-of-the-art applications of fire protection engineering.▲ Short courses and seminars offering hands-on instruction.▲Discounts on fire-related publications.

Other benefits include:▲ Recognition of your professional qualifications. ▲ Contribute to the profession through technical task ▲ Opportunity to participate in the SFPE Annual Meeting. groups and committees.▲ Opportunity to network in local chapters. ▲ A periodic profile of the fire protection engineer, ▲ Low cost group life, health, and liability insurance. including salary information.

Advancing the Science and Practice of Fire Protection Engineering

I’m interested in learning more about joining SFPE. Please send me additional information.

Name Title

Company/Organization

Address City State/Province Zip/Postal Code

Country

Work Phone Fax E-mail

Fax to 301/718-2242 ▲ Visit the SFPE Web Site: www.sfpe.org

For more information, contact The Society of Fire Protection Engineers:7315 Wisconsin Avenue, Suite 1225 West ▲ Bethesda, MD 20814

Phone: 301/718-2910

Societyof Fire ProtectionEngineers

An Invitation to Join

Page 24: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 39

bridge level, so there was a possibilitythat some people escaping through theatrium would have to move throughthe smoke in order to reach the pro-tected stairways. An analysis was doneto determine the smoke temperatureand density of smoke particulates in theatrium, and from this the visibility, atany given time.

Due to the shallow depth of thegallery spaces on either side of the atri-um, escape distances to the protectedstairs are relatively short. Furthermore,clear views into the atrium from allaccommodation areas would enhanceearly awareness of any smoke accumu-lating in the atrium, and gallery man-agement procedures ensure that trainedstaff would be in attendance in thepublic areas at all times. Given thesecharacteristics, an upper limit to thetotal time for evacuation of the upper

levels into one of the two protectedstairs was calculated to be less than 3minutes.

With the limiting parameters for ten-ability assumed to be a smoke tempera-ture of 60°C and visibility of more than10m, it was found that acceptable con-ditions were maintained in the atriumfor more than twice the evacuationperiod.

Tenability of the open bridge links istherefore maintained during the escapeperiod, even for this unlikely scenario.The situation is shown schematically inFigure 2.

4. Fire Resistance RequirementsThe height of the top floor of this

building is in excess of 28 m, andtherefore the Spanish NBE-CPI/96 coderecommended a minimum 3-hour fire-resistance period for the structure. A

further implication of the strict codeinterpretation would have meant thatthe separating elements between thegallery accommodation and atriumwould require a fire rating of 90 min-utes. In order to allow natural lightingof the galleries at all levels, the archi-tect wanted full-height glazing alongthe gallery/atrium interface. Even itwere possible to achieve a 90-minutefire-resisting glazed partition to thegalleries with acceptable framingdetails, it would have been prohibi-tively expensive.

However, it was obvious from theoutset that the gallery fire loads weremuch lower than the fire load densityof 750 MJ/m2 typically associated withsuch a “public assembly” building.3

There were also large areas of potentialventilation from the gallery levels intothe atrium. Therefore, Article 14(a) ofNBE-CPI/96 was used, which states thatthe designer has the choice of eitheradopting the tabulated fire-resistancevalues or to determine the value byanalytical means using approved calcu-lation methods.

In the Caixa Galicia building, the fireloads in the galleries are less than nor-mal for public assembly buildings, andthe compartment sizes are far smallerthan the limiting dimensions assumedfor the tabulated fire-resistance values.Furthermore, it is unlikely that the fireloading for this building could changesignificantly without a major refurbish-ment. It was felt, therefore, that a per-formance-based approach should beused to derive a fire-resistance periodmore appropriate to this particularbuilding.

The method adopted for the calcula-tion of the fire-resistance period wasbased on the “equivalent time” calcula-tion approach given in the EurocodeENV 1991-2-2: 19962, which takes thefollowing form:

where te,d = equivalent time of fire expo-

sure (minutes)qf,d = design fire load density

(MJ/m2)kb = conversion factor for thermal

properties of enclosurewt = ventilation factorFigure 3. Section through the Caixa Galicia atrium bisecting the building

t q k we d f d b t, ,= × ×

Page 25: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

40 Fire Protection Engineering NUMBER 12

As a basis for the calculations, fireloads appropriate to the use of the dif-ferent areas were taken from statisticaldata based on a comprehensive surveycarried out on buildings throughoutEurope.3 The fire load densities usedfor the different areas in the CaixaGalicia building are given in Table 1,and it can be seen from this how thefire loads can differ substantially fromthe values assumed in the code tables.

The thermal inertia of the compart-ment is represented by the factor kb

and can be quite easily calculatedonce some basic fitting out details areknown, or alternatively a conservativedefault value can be used. The ventila-tion factor is calculated from a formulabased on the compartment geometry;height, floor area, and area of ventila-tion openings.

The “t-equivalent” value, therefore,can readily be calculated for eachcompartment using the relationshipgiven above.

But fire-resistance periods given innational regulations take into accountfactors other than fire load and ventila-tion – they make allowance for ease ofescape, access for firefighters, theprobability of a fully developed fireoccurring, and the consequence ofstructural failure. The “t-equivalent”value on its own, therefore, cannot beequated to a fire-resistance period.

The fire-resistance values were cal-culated by multiplying “t-equivalent”time period with factors for quantify-ing the risk of structural failure, assuggested in a UK ‘NationalApplication Document,4 intended tosupplement Eurocode guidance. Theapplication of these “gamma factors”((1) and (2)) relate the “t-equivalent”values to a fire-resistance period byassociating the risk of failure withheight of building above access level.The derivation of the “gamma factors”for the NAD4 approach was done insuch a way to ensure that the fireresistance values calculated using the“t-equivalent” approach are similar tothe prescriptive code values and, assuch, have no deeper scientific basis.The method, however, does give thedesigner some flexibility when the fireload or ventilation conditions are dif-

ferent to the typical values assumedfor the code tables. The results fromapplying the approach outlined aboveto a number of floors is given in theTable 2.

The method used for calculating thefire-resistance period described abovealso recognises the role of sprinklersin controlling the size of a compart-ment fire and makes allowance for thisby means of an additional reductionfactor. In the later development of thedesign, sprinkler protection was intro-duced into the Caixa Galicia buildingat all levels as a property protectionmeasure. This introduced an addition-al factor of safety into the design thatwas not used in the derivation of thefire-resistance values given above.

By using this “first principles”approach, it was determined that a 60-minute fire-resistance period was ade-quate for the structure of the CaixaGalicia building. The main benefit ofthis approach in this case was that itcould be used to justify a 60-minutecompartmentation between floors,

which allowed us to use a standard30-minute fire-resistant glazing to sepa-rate the galleries from the atrium.

George Faller is with Arup Fire.

REFERENCES

1. Norma Básica de la Edificación, NBE-CPI/96: “Condiciones de protección con-tra incendios en los edificios,” CSCAE,1996.

2. Eurocode 1: Basis of Design and Actionon Structures, Part 2.2 – Actions onStructures Exposed to Fire, DD ENV1991-2-2:1996.

3. CIB W14 Workshop. Design Guide –Structural Fire Safety, Fire Safety Journal,March 1986.

4. BSI, National Application Document,Eurocode 1: Basis of Design and Actionon Structures, Part 2.2 – Actions onStructures Exposed to Fire for use in theUK, DD ENV 1991-2-2:1996

For an online version of this article,go to www.sfpe.org.

TABLE 1Fire load densities

Occupancy Fire load densities

(MJ/m2)

Office 570

Assembly (entertainment) 750

Shops and Commercial 900

Gallery 250

TABLE 2Calculated fire resistance periods

Occupancy Height (m) t-equivalent Risk Risk Calculated(minutes) Factor(1) Factor(2) Resistance

(minutes)Consequence Probability

Ground Floor Gallery 0.00 33 0.8 0.8 21

Ground F1 Bookshop 0.00 65 0.8 0.8 21

Upper Levels Gallery 17.10 24 1.1 0.8 21

Upper Levels Offices 29.70 29 1.6 1.2 56

Page 26: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 43

By Morgan J. Hurley, P.E, andJames G. Bisker, P.E.

Ahigh-pressure, total-flooding, car-bon dioxide extinguishing systemdischarged without warning dur-

ing routine maintenance of electricalequipment resulting in one fatality andseveral serious injuries. At the time of theaccident, the newly installed systemreleasing panel was electronically dis-abled and considered out of service, yetit still actuated the system when workcrews disconnected primary power to thepanel.

Several lessons can be learned fromthis accident. In particular, designers mustexamine personnel safety in the contextof possible special extinguishing systemfailure modes as well as the protection ofthe facility from attack by fire.

BACKGROUND1

The Idaho National Engineering andTesting Environmental Laboratory (INEEL)is a government-owned, contractor-oper-ated facility located in a rural area ofsoutheastern Idaho. The site housesnuclear reactors that are used for testingand research, fuel storage buildings, andmiscellaneous support infrastructure.

Building 648 is a two-story building

that contains electrical support equipmentfor a reactor and support facilities withinthe Test Reactor Area. The building isprotected with a carbon dioxide (CO2)extinguishing system, with automaticsprinklers installed in an adjoining emer-gency generator room. The CO2 systemwas originally installed in 1971 and con-sists of a fire-detection system, 2,500 kg(5,500 lb) of carbon dioxide stored in 55high-pressure bottles, and discharge pip-ing and nozzles. The CO2 system wasdesigned to achieve a 50 percent concen-tration of carbon dioxide.

The fire-detection system consists of areleasing panel, initiating devices, notifica-tion devices, and actuation devices.Initiating devices include 14 heat detec-tors, two manual pull stations, two manu-al CO2 releasing stations, and a waterflowdetector for the building’s dry pipe sprin-kler system. Notification devices includebuilding evacuation signals. Actuatingdevices include releasing circuits for theCO2 supply and an interface circuit thatallowed the fire alarm to be monitored bythe facility’s fire alarm reporting system.

The CO2 system was designed to bereleased upon a signal either from thefire-detection system or manually via anemergency release lever located in theCO2 storage shed. The fire-detection sys-tem was programmed to discharge the

CO2 upon activation of a heat detector orupon activation of one of the two manualCO2 releasing stations. The emergencyrelease lever in the storage shed was notconnected to the fire-detection system.

When the CO2 system was dischargedvia the fire-detection system, dischargewas accomplished by operation of twoelectrically operated control heads. Thecontrol heads were connected directly totwo of the 45 kg (100 lb) CO2 storagebottles. Operation of a control headreleased the CO2 from the bottle to whichit was connected, which pressurized thedischarge manifold. Pressuriza- tion of thedischarge manifold opened pressure-operated valves on the other CO2 storagebottles. The CO2 would then flowthrough a piping network and dischargeout of nozzles located in Building 648.Manual operation of the emergencyrelease lever would directly release theCO2, independent of the fire-detectionsystem. See Figure 1 for a diagram of theCO2 system arrangement.

The CO2 and fire-detection systemsincorporated two distinct time delays.The control panel was programmed witha 30-second delay. Upon operation ofeither a heat detector or a manual CO2

releasing station, the control panel wasprogrammed to sound the evacuation sig-nals immediately and begin a 30-second

LESSONS

LEARNED

FROM A

Carbon Dioxide System Accident

Page 27: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

44 Fire Protection Engineering NUMBER 12

time delay sequence, after which the con-trol panel would actuate the control headsolenoids.

An additional 25-second mechanicaldelay device was installed between theCO2 storage bottles and the dischargepiping to the building. This mechanicaldelay would automatically retard the dis-charge of CO2 and only required thepressure of the CO2 for its operation.With the combination of the delay pro-grammed into the control panel and themechanical delay, the evacuation signalswould sound for a total of 55 secondsprior to the discharge of CO2 when dis-charge was initiated via the fire-detectionsystem.

THE ACCIDENT1

On the afternoon of July 29, 1998,workers were preparing to perform pre-ventive maintenance of the electricalswitchgear in Building 648. The CO2 sys-tem was electronically impaired by pro-gramming the control panel not to acti-vate the control heads located on the CO2

bottles. Later, the work crew began opening

circuit breakers in preparation for thepreventive maintenance work. Shortly

after the last breaker was opened, the CO2

system discharged, creating “near zero vis-ibility.”1 While the evacuation alarms mayhave briefly sounded for less than onesecond, they did not continuously soundin conjunction with the CO2 release.

After the CO2 discharge, the workersran towards the exits, which were visiblesince they were held open by cables run-ning into the building from portable gen-erators. Eight of the workers were able toexit on their own; however, five remainedinside of the building and were renderedunconscious by the CO2. Three were laterrescued by the workers who had earlierescaped, which left two people remainingin the building. One of the remainingworkers was later revived, and the otherperished.

INVESTIGATION BY DOE1

Following the accident, the Departmentof Energy conducted an investigation todetermine the causes of the accident. Theaccident investigation determined that alikely cause of the CO2 system release wasthe transmission of a spurious signal tothe control heads, which occurred whenthe power to the control panel was dis-connected.

During a recreation of the events lead-ing up the accident, the circumstancesthat would have led to another CO2 dis-charge were duplicated when primarypower was disconnected from the controlpanel. When primary power was discon-nected from the control panel, a spuriousand momentary signal developed in thereleasing panel’s energy power supplyand charging circuitry that migratedthrough the releasing circuits, causing thecontrol solenoids to actuate and the evac-uation signals to momentarily sound. Theevacuation alarms again failed to continu-ously sound for a duration that wouldadequately warn building occupants.

While the malfunction of the controlpanel was the direct cause of the acci-dent, several contributing factors wereidentified, which included:

• The releasing panel did not monitordischarge of the CO2 system.

• The alarm sounded for an insufficientperiod of time prior to the CO2 dis-charge.

• The CO2 system was not equippedwith a mechanism that would havesupported lockout.

• The releasing panel was not equippedwith a supervised circuit disconnectswitch and had to be electronicallydisabled for work in the facility.

HAZARDS OF CARBON DIOXIDEEXTINGUISHING SYSTEMS

Carbon dioxide is a commonly usedextinguishing agent. According to theEnvironmental Protection Agency, CO2 isused in approximately 20 percent (basedon cost) of all special hazard fire protec-tion systems.2 The U.S. EPA also estimatesthat special hazard applications comprise20 percent of all fire protection systemapplications;2 therefore, CO2 systems com-prise approximately 4 percent of all fireprotection systems.

Carbon dioxide possesses several prop-erties that make it a desirable fire protec-tion agent. These include:2

• CO2 is noncombustible and does notproduce other hazardous substanceswhen heated (as some “clean agents”do).

• CO2 is stored in a pressurized state,and the pressurization alone is suffi-cient to propel the gas into a protect-ed space.

• CO2 leaves no residue following adischarge.

• CO2 does not react with most othermaterials.

CO2 Storagebottles

CO2 Manifoldpiping

Dischargedelay

Electroniccontrol heads

Check valve

Safety outlet(Pressure relief valve)

To switchgearroom

2" (50 mm)Connecting pipeto bypass stop valve

Figure 1. CO2 system arrangement1

Page 28: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 45

• Since it is a gas, CO2 provides “three-dimensional” protection, i.e., it canprotect spatially complex hazardssuch as printing presses or marineengine rooms.

• CO2 does not conduct electricity.Although CO2 is a naturally occurring

substance, produced, for example, by firesand breathing, it also can pose dangers inhigh concentrations. At low concentrations(~4%), CO2 is relatively benign.2 However,at greater concentrations, CO2 is lethal tohumans, and at concentrations greaterthan 17 percent, death can occur in lessthan one minute. Other adverse physiolog-ical effects can occur from exposures lessthan 17 percent, including headache,shortness of breath, and unconsciousness.2

The minimum design concentration ofCO2 for fire suppression varies with dif-ferent fuels and ranges from 34 percentto 75 percent.3 Most CO2 system applica-tions are at the low end of this range.Nevertheless, at a minimum design con-centration of 34 percent, all total floodingsystems create an environment that islethal to humans.

Carbon dioxide for extinguishing sys-tems is typically stored in a liquid form.When CO2 is discharged, the endothermicexpansion absorbs heat, and the dischargedCO2 is cold enough to cause water vaporin the air to condense into small droplets,creating a fog. This fog can obscure visi-bility, making egress more difficult.

Walking speed through nonirritatingsmoke has been shown to decrease asthe smoke concentration increases.4 Sincethe effect of nonirritating smoke onhumans would primarily be a reductionin visibility, it can be concluded that thedecrease in visibility caused by the fogcreated by CO2 discharge would alsodecrease walking speed.

Taken together, these two effectsdemonstrate the importance of evacuatingall people from a protected space beforeCO2 discharge. Since for systems installedto extinguish surface fires, a CO2 designconcentration of at least 34 percent mustbe achieved within one minute,3 a lethalenvironment would be created veryquickly, and evacuation times wouldincrease due to the reduction in visibility.

ANALYSIS OF CONTRIBUTINGCAUSES IDENTIFIED BY DOE

LockoutThe U.S. Occupational Safety and

Health Administration (OSHA) lockout

provisions are contained in 29 CFR 1910,subpart J. The scope of 29 CFR 1910, sub-part J, (contained in 29 CFR1910.147(a)(1)(i)) states that the regulationapplies to “the servicing and maintenanceof machines and equipment in which theunexpected energization or startup of themachines or equipment, or release ofstored energy could cause injury toemployees.”5

“Lockout” is defined as “the placementof a lockout device on an energy isolatingdevice... ensuring that the energy isolatingdevice and the equipment being con-trolled cannot be operated until the lock-out device is removed.”5

NFPA 12 also contains lockout require-ments. The edition of NFPA 12 that was ineffect at the time of the accident stated:3

“To prevent accidental or deliberate dis-charge, a ‘lockout’ shall be provided whenpersons not familiar with the systems andtheir operation are present in a protectedspace.” The 2000 edition of NFPA 12added a definition of “lockout” that paral-lels the OSHA definition.

While the CO2 system was not fittedwith a lockout device that would havemet the definitions contained in theOSHA rules, INEEL operating procedurescalled for removal of the electronic con-trol heads “prior to maintenance thatcould cause a release of CO2.”1 However,at the time of the accident, the controlpanel was used to electronically disablethe system in lieu of removing the controlheads.

Engineering ControlsThe National Fire Alarm Code (NFPA

72) requires that6 “the operation of anautomatic fire suppression systeminstalled within the protected premisesshall cause an alarm signal...” TheNational Fire Alarm Code also states that“the operation of... fire extinguishing sys-tem(s) or suppression system(s) shall initi-ate an alarm signal by means appropriateto the system, such as agent flow or agentpressure, by alarm-initiating devicesinstalled in accordance with their individ-ual listings.”

The CO2 system at the INEEL wasinstalled to operate the evacuation alarmand began counting down the logic-basedpredischarge delay simultaneously.However, because the control panel didnot intentionally operate the control heads,the predischarge alarm did not sound.

Operation of the CO2 system wasaccomplished via two means – upon asignal from the control panel or by manu-

al release at the CO2 storage. This lattermeans of operation is referred to as“emergency manual operation” by NFPA12.3 Emergency manual operation isrequired by NFPA 12 for all CO2 systemvalves.3 Because the control panel did notmonitor release of the CO2 system, opera-tion of the emergency manual controlwould have caused the CO2 system todischarge, without sounding the evacua-tion alarms or notifying response organi-zations.

NFPA 72 also requires that fire alarmsystems used for releasing service be pro-vided with a supervised disconnectswitch to allow for system testing withoutactivating the fire suppression system.Had this switch been provided at thissite, it could have been used to physicallyseparate the releasing panel from the CO2

control heads as opposed to disabling thereleasing panel electronically. Since thisswitch is not required as a condition forlisting as a releasing panel, design engi-neers must specify and verify its installa-tion.

It is noteworthy that the appendix tothe 1996 edition of the National FireAlarm Code contained explanatory mater-ial (in A-5-7) that stated that evacuationalarms may be initiated by the fire-detec-tion system. While this explanatory mater-ial was deleted in the 1999 edition of theNational Fire Alarm Code, it is likely thatthere are many CO2 systems and otherreleasing panel controlled suppressionsystems in service that are configured in asimilar manner to the CO2 system atINEEL.

OTHER AVAILABLE TECHNOLOGY

With the exception of “small” systems,defined as those with less than 140 kg(300 lb) of CO2 storage, marine CO2 sys-tems in the U.S. incorporate a number offeatures that are not required in theirland-based counterparts. Neither automat-ic nor electronic operation is permitted inmarine CO2 systems; operation must beby mechanical means. In marine CO2 sys-tems, electronic pre-discharge delays arenot permitted. Predischarge delays mustalso be mechanical and depend only onthe pressure of the CO2 to operate. Also,the audible evacuation alarms used inmarine CO2 systems must be powered bythe CO2 pressure; electronic alarms arenot permitted.7

The types of components that are usedin marine CO2 systems are readily avail-

Page 29: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 46

able and provide an increase in safetyover the types of components typicallyused in shore-based systems. Althoughthey are not required in shore-based sys-tems, their use should be consideredsince they would eliminate many readilyforeseeable failure modes that could leadto injury or death of people in a spaceprotected by CO2.

IMPLICATIONS FOR CO2 DESIGN

The CO2 system in Building 648 nearlycomplied with NFPA 12 and NFPA 72.However, the accident could have beenprevented by one of two methods: (1)the provision of pressure switches in theCO2 discharge piping upstream of themechanical delay that would cause theevacuation alarm to sound, or (2) the

provision of mechanically operated CO2-powered sirens in the CO2 discharge pip-ing that operated by the pressure of theCO2 alone.

While the provision of a lockoutdevice or a circuit disconnect switchpotentially could have prevented theaccident, given that it would require adeliberate action to operate, and that theprocedures that were in place to removethe control heads were not followed, it issuggested that a lockout device alone isnot sufficient. Addition- ally, a failuremode would still be present where thesystem can discharge without the evacua-tion alarms sounding.

This accident demonstrates that codecompliance alone may not be sufficientfor CO2 system design, and engineersshould consider possible failure modesand effects during the design of CO2 sys-tems. Similarly, existing CO2 systemsshould be evaluated to determinewhether they provide adequate workersafety.

Morgan Hurley is with the Society ofFire Protection Engineers. James Bisker iswith the U.S. Department of Energy.

REFERENCES

1 “Type A Accident Investigation Board Reportof the July 28, 1998, Fatality and MultipleInjuries Resulting from Release of CarbonDioxide at Building 648, Test Reactor AreaIdaho National Engineering andEnvironmental Laboratory,” U.S. Departmentof Energy, EH2PUB/09-98/01A1, September,1998.

2 “Carbon Dioxide as a Fire Suppressant:Examining the Risks.” EPA430-R-00-002, U.S.Environmental Protection Agency, 2000.

3 NFPA 12, Standard on Carbon DioxideExtinguishing Systems, National FireProtection Association, Quincy, MA: 1993.

4 Jin, T., “Studies on Human Behavior andTenability in Fire Smoke,” Fire SafetyScience: Proceedings of the FifthInternational Symposium, Dr. Y. Hasemi,Ed., Society of Fire Protection Engineers,Bethesda, MD: 1997.

5 29 CFR 1910.147

6 NFPA 72, National Fire Alarm Code,National Fire Protection Association, Quincy,MA: 1999.

7 46 CFR 76.15.

For an online version of this article, goto www.sfpe.org.

Page 30: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

48 Fire Protection Engineering NUMBER 12

Established in 1939, Schirmer Engineering was the first independent fireprotection engineering firm to assist insurance companies in analyzing

and minimizing risk to life and property. Since then, Schirmer Engineeringhas been a leader in the evolution of the industry, innovating for tomor-row with science and technology, using insight from tradition and experi-ences of our past. Today, Schirmer Engineering, with a staff of 195employees, is synonymous with providing high-quality engineering andtechnical services to both national and international clients.

Career growth opportunities are currently available for entry-level andsenior-level fire protection engineers, design professionals, and codeconsultants. Opportunities available in the Boston, Chicago, Charlotte,Dallas, Las Vegas, Los Angeles, Miami, Phoenix, San Diego, SanFrancisco, and Washington, DC, areas. Our firm offers a competitivesalary and benefits package, including 401(k). EOE

Send résumé to:

G. JohnsonSchirmer Engineering Corporation707 Lake Cook RoadDeerfield, IL 60015-4997

Fax: 847.272.2365E-mail: [email protected]

SCHIRMER ENGINEERING

Fire protection engineering is a growing profession with many challenging career opportunities. Contact the Society of Fire Protection Engineers at www.sfpe.org or the organizations below for more information.

Arup Fire

The F.P. ConnectionAn electronic, full-service fire protection resource Web site.

The F.P. Connection offers posting of employment opportunitiesand résumés of fire protection professionals. If, as a fire protec-

tion service provider or equipment manufacturer, your Web site is dif-ficult to locate using search engines and keywords, let us post yourbanner and provide a direct link for use by our visitors who mayrequire your services.

Please visit us at www.fpconnect.com or call 724.746.8855.

For posting information, e-mail [email protected].

Fax: 724.746.8856

As part of our global expansion, Arup Fire is seeking to add staff through-out the USA, including Massachusetts, New York, California, Washington,

DC, and Florida. Successful candidates will play a very active role in develop-ing Arup Fire in the USA and will work closely with many of the world’s lead-ing architects and building owners, developing innovative, performance-baseddesign solutions for a wide range of building, industrial, and transport pro-jects.

This is an excellent opportunity for qualified fire engineers to develop theircareer within one of the leading international fire engineering consultancies.

Candidates should possess a degree in Fire Protection Engineering or relat-ed discipline, approximately five years of experience in code consulting anddeveloping performance-based strategies in a project design environment, anda PE in Fire Protection Engineering or related field. Risk management, quanti-tative risk assessment, industrial fire engineering, and computer modelingskills will be highly regarded.

Successful candidates will have excellent report writing and interpersonalskills. They should also have an ability to present clear and technically soundfire engineering strategies and have good negotiating skills in order to gainapproval of clients, architects, and authorities for fire engineering designs.

Arup Fire offers competitive salaries and an excellent benefits package.

Please submit your résumé andsalary history to:

Arup Fire901 Market StreetSuite 260San Francisco, CA 94103

http://www.arup.com/

Attention: Jim QuiterTelephone: 415.957.9445Fax: 415.957.9096E-mail: [email protected]

ClassifiedsCareers/

DecoShield Systems, Inc.The Creative Cover Support System

COVERS IT ALL!!!

FOR MORE INFORMATION

&SAMPLES

CALL800-873-0894

OR VISIT OUR WEBSITE @

www.decoshield.com

CONCEALING sprinkler piping, plumbing,electrical wire and cabling has never beeneasier, thanks to the UL-ListedDecoShield® Cover and Support Systems.

We have been serving the fire sprinklerretrofit industry for more than tenyears. Our system is clean, attractive,paintable, durable, tamper-resistant,economical and quick to install. That is why we are the #1 choice for retrofitting all types of buildings.

Page 31: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

50 Fire Protection Engineering NUMBER 12

ResourcesResourcesRefresh your

FIREPROTECTION

ENGINEERINGBOOKSHELFwith these

new titles

available from

SFPEPrinciples of Fire Behavior, JamesG. Quintiere, Ph.D.Provides a comprehensive treat-ment of fire behavior for the fire-fighter, building official, investi-gator, and product manufacturer.Includes chapters on heat trans-fer, ignition, flame spread, fireplumes, heat flux, and other top-ics. Special sections on combus-tion products and compartmentfires help readers understandprinciples of fire behavior inspecific situations.

Price: SFPE Member, $49.00;

Nonmember, $60.95

From Model Codes to the IBC: A Transitional GuideThe IBC is different from every other building codeyou’ve ever referenced. This user-friendly, 900-pageguide includes side-by-side comparisons of IBCrequirements to the National Building Code (NBC),Standard Building Code® (SBC), the Uniform BuildingCode©®(UBC), and NFPA 101® Life Safety Codes.The guide also features a quick-find index linkingmodel code items to their IBC equivalents. Publishedin 2001 by Protection Knowledge Concepts.

Price: SFPE Member, $91.99;

Nonmember, $114.95

The Technical Basis for Performance-Based Fire Regulations: A discussion of capabilities, needs, and benefits of fire safety engineeringIn January of 2001, the United Engineering Foundation, in cooperation with SFPE,sponsored a conference in San Diego, CA, designed to further develop the SFPEResearch Agenda for Fire Protection Engineering. Three keynote lectures and fourfocused sessions on current knowledge and research needs resulted in the nextstep in better defining the technical basis for performance-based fire regulations.The conference report, prepared by Geoff Cox, conference organizer, can befound at SFPE’s Web site at www.sfpe.org.

Structural Design for Fire SafetyAndrew H. Buchanan, University of Canterbury, New Zealand

Presenting a comprehensive overview of the fire resistanceof building structures, this text provides expert guidance on:• Interpreting code requirements for fire safety• Understanding the concepts of fire severity and fire resis-

tance• Estimating time-temperature curves for fully developed

compartment fires• Understanding the behavior of structural elements and

buildings exposed to fires• Designing steel, concrete, and timber structures to resist fire exposure• Assessing the fire performance of existing structures

Published in 2001 by John Wiley and Sons.

Price: SFPE Member, $118.00; Nonmember, $150.00

To Order Any Publication, Call SFPE at 301.718.2910

Page 32: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

FALL 2001 Fire Protection Engineering 51

Design of Special Hazard and FireAlarm Systems, Robert M. Gagnon,P.E. This is the most current guideto the design of state-of-the-art spe-cial hazard and fire protection sys-tems. It provides information on asimplified methodology for calculat-ing special hazard systems, withillustrations and solved examples ineach of 15 chapters. Includes achapter on ethical practice.Published in 2001 by DelmarPublishers.

Price: SFPE Member, $62.50;

Nonmember, $77.95

Design of Water-Based Fire ProtectionSystems, Robert M. Gagnon, P.E.This abundantly illustrated, step-by-step approach is a vital reference forevery designer of fire protection sys-tems. Hydraulic calculations of themost commonly encountered water-based fire protection systems areaddressed in detail, and a disk isincluded. The latest technology inwater-spray and water-mist systemsis presented. Published in 2001 byDelmar Publishers.

Price: SFPE Member, $62.50;

Nonmember, $77.95

Clean Agent Systems Video SeriesThis special hazards fire protection program compares the clean agent systems onthe market today. The 4-tape program covers agent types and typical applications toNFPA and EPA regulations and maintenance guidelines. Published in 2001 byProtection Knowledge Concepts.

Price: SFPE Member, $249.00; Nonmember, $299.00

Page 33: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

November 10-14, 2001

NFPA Fall MeetingDallas, TXInfo: www.nfpa.org

December 2-5, 2001

Symposium on Thermal Measurements:The Foundation of Fire Standards

Dallas, TXInfo: www.astm.org

December 3-6, 2001

5th Asia-Oceania Symposium on Fire Science &Technology

Newcastle, AustraliaInfo: http://www.eng.newcastle.edu.au/cg/AOSF-

ST5/welcome.html

January 10-11, 2002

The 3rd International Forum on Safety Engineeringand Science (IFSES III) Virtual Symposium

Info: http://hugoniot.chem.t.u-tokyo.ac.jp/IFSES3-1.htm

March 18-19, 2002

Structures in FireUniversity of Canterbury, Christchurch, New ZealandInfo: http://www.civil.canterbury.ac.nz

March 20-22, 2002

4th International Conference on Performance- Based Codes and Fire Safety Design Methods

Melbourne, AustraliaInfo: www.sfpe.org

May 19-23, 2002

NFPA World Fire Safety Congress and ExpositionMinneapolis, MNInfo: www.nfpa.org

52 Fire Protection Engineering NUMBER 12

ResourcesU P C O M I N G E V E N T S

UPCOMING EVENTS

Page 34: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

54 Fire Protection Engineering NUMBER 12

• Advanced Fire Technology, Inc. ..................Page 27• Ansul, Inc. .....................................................Page 21• Bilco ...............................................................Page 52• Blazemaster Fire Sprinkler Systems .............Page 49• Central Sprinkler..............................................Page 2• Commercial Products Group..........................Page 8• DecoShield Systems, Inc...............................Page 48• Draka USA.....................................................Page 25• Edwards Manufacturing ................................Page 46• Edwards Systems Technology.................Page 28-29• Fike Protection Systems..................................Page 6• Fire Control Instruments, Inc. ......................Page 47• Gem Sprinkler Company.........................Back Cover• Great Lakes Chemicals..................................Page 53

• Koffel Associates ...........................................Page 17• Master Control Systems.................................Page 22• NOTIFIER Fire Systems..Page 18 & Inside Back Cover• Potter Electric Signal Company....................Page 55• Protection Knowledge Concepts..................Page 41• Protectowire ..................................................Page 51• The RJA Group ............................Inside Front Cover• Reliable Automatic Sprinkler ........................Page 42• Siemens............................................................Page 5• SimplexGrinnell .......................................Page 12-13• TVA Fire & Life Safety, Inc. ..........................Page 37• Viking Corporation .......................................Page 14• Wheelock, Inc. ..............................................Page 34

Index ofAdvertisers

CORPORATE 100

Solution to last issue’s brainteaser

Starting with the letter A, number the letters in thealphabet in order from 1 to 26. Using these values forthe letters, find a word with the product of 69,888.

Answer: “Pump”

The difference between any two numbers in the set {2,3, 4} is equal to their greatest common factor. The same istrue of any two numbers in the sets {6, 8, 9, 12} and {8, 9,10, 12}. Find a set of five numbers for which this is true.

Thanks to Jane Lataille, P.E., for providing this issue’sbrainteaser.

B R A I N T E A S E R The SFPE Corporate 100 Program wasfounded in 1976 to strengthen the relation-ship between industry and the fire protec-tion engineering community. Membershipin the program recognizes those who sup-port the objectives of SFPE and have agenuine concern for the safety of life andproperty from fire.

BENEFACTORSRolf Jensen & Associates, Inc.

PATRONSCode Consultants, Inc.Edwards Systems TechnologyHughes Associates, Inc.The Reliable Automatic Sprinkler Company

MEMBERSArup FireAutomatic Fire Alarm AssociationBFPE InternationalFactory Mutual Research CorporationFike CorporationGage-Babcock & Associates, Inc.Grinnell Fire Protection SystemsHarrington Group, Inc.HSB Professional Loss ControlHubbell Industrial ControlsJoslyn Clark Controls, Inc.James W. Nolan Company (Emeritus)Industrial Risk InsurersKoffel Associates, Inc.Marsh Risk ConsultingNational Electrical Manufacturers AssociationNational Fire Protection AssociationNational Fire Sprinkler AssociationNuclear Energy InstituteThe Protectowire Co., Inc.Reliable Fire Equipment CompanyRisk Technologies, LLCSchirmer Engineering CorporationSiemens Cerberus DivisionSimplexGrinnellUnderwriters Laboratories, Inc.Wheelock, Inc.W.R. Grace Company

SMALL BUSINESS MEMBERSBourgeois & Associates, Inc.Demers Associates, Inc.Fire Consulting Associates, Inc.MountainStar EnterprisesPerformance Technology Consulting Ltd. Poole Fire Protection Engineering, Inc.S.S. Dannaway & Associates, Inc.The Code Consortium, Inc.

Page 35: Mission - c.ymcdn.com in flammable liquid and ... Test data accumulated in our laboratory, as ... The inference to be drawn

56 Fire Protection Engineering NUMBER 12

Performance-based fire protectiondesign requires doing manythings differently than prescrip-

tive-based design. Some of these dif-ferences are obvious; for example, theuse of more sophisticated models andcalculation methods and the amount ofengineering analysis required. However,there is a key difference that is notalways immediately obvious: the met-ric that an engineer uses to gage theacceptability of their own design.

In the case of prescriptive-baseddesign, determination of what generaldesign attributes constitute an accept-able level of safety is one step removedfrom the engineer. These determina-tions were made during the develop-ment of codes and standards and bygovernmental jurisdictions as theydecided which codes and standards toadopt and what amendments, if any, toimpose.

With such designs, a designer neededonly to ensure that all attributes of adesign fit within the minimums andmaximums specified by a code or stan-dard to determine if their design wasacceptable. Similarly, from an enforce-ment standpoint, verification that adesign complied with the code or stan-dard was relatively straightforward.The engineer’s final metric of successwas frequently whether or not enforce-ment officials accepted the design.

However, with performance-baseddesign, the attributes of an acceptabledesign are not as clearly stated. Whilean “acceptable level of safety” is stillestablished through the developmentand adoption of codes and standards,the responsibility falls to engineers tospecify a design that is acceptably safe– and to demonstrate why the design issufficiently safe. Before asking enforce-ment officials to evaluate a design, theengineer should first prove to himselfor herself that the design is safe. In fact,the engineer should be their own worst

critic, as the engineer who prepared adesign has the best perspective ofwhether or not the design provides anacceptable level of safety.

Models, correlations, and other engi-neering methods are often used duringthe development of a performance-based design to establish the designattributes and to verify that the designmeets the performance criteria in thedesign fire scenarios selected. Each ofthese models and correlations will havea limited range of applicability.Similarly, input data that is used in themodels or correlations will only have alimited range of applicability. Also allmodels, correlations, and data havesome uncertainty associated with them.

There is nobody better suited to con-sider the applicability of the models,correlations, and data that is used in thedevelopment and evaluation of perfor-mance-based designs than the engineerthat performed the analysis. Whileenforcement officials or third-party engi-neers may review a performance-baseddesign, they will not consider the devel-opment of the design and the selectionof methods and data as closely as theengineer who developed the designdid.

No engineer would intentionally pre-pare a design that he or she knew didnot provide an acceptable level of safe-ty. In most cases when designing toprescriptive codes and standards, codecompliance is sufficient to provide anacceptable level of safety. However,when using models or other engineer-ing methods to develop or evaluate aperformance-based design, an engineermust first evaluate the applicability, lim-itations, and inherent assumptions ofthe methods and data used to fullyunderstand the level of safety provided.

A Paradigm Change

Morgan J. Hurley, P.E.Technical DirectorSociety of Fire Protection Engineers

from the technical director