based design analyses often involve a comparison of the contributions of
a variety of fire protection systems to achieve specified design
objectives. Fire protection systems can be divided into two categories:
passive and active systems. Passive systems include fire-rated barriers
and protection of openings in fire-rated barriers, while active systems
include systems such as fire detectors and sprinklers.
analysis of the contribution of a particular fire protection system to
the achievement of specified objectives should include an assessment of
the effectiveness and reliability of the proposed fire protection
systems. For the purposes of this article, "reliability” is defined as
the probability that a product or system will operate under designated
operating conditions for a designated period of time or number of
cycles. "Effectiveness” refers to the ability of a system to achieve
While reliability data for sprinkler systems has been included in NFPA analyses,1
few compilations of reliability data are available in the open
literature for other types of fire protection systems. This article
provides an overview of the results of recent research that has been
conducted to assemble such data from a variety of available sources.
data may be derived from fire incident statistics considering the
entire fire protection system to be a single entity. Alternatively, the
reliability of a system may be determined from an engineering analysis
based on failure and repair rates of the components of the system,
accounting for any redundancy in system components. For this article,
the reliability data comes from considering the system to be one entity.
Figure 1: Effectiveness of Sprinklers in U.S. Fire Incidents1
|Occupancy||Fire death rate* without auto extinguishing system||Fire death rate* with wet pipe sprinkler||Percent reduction|
|All public assembly||0.4||0.0||100%|
Table 1: Fire Death Rates With and Without Sprinklers1
|Reason for failure||All||wet pipe||Dry pipe|
|System shut off||65%||61%||74%|
|Manual intervention defeated system||16%||19%||8%|
Lack of maintenance
|System component damaged||7%||6%||10%|
|Inappropriate system for type of fire||5%||6%||3%|
|Total fires per year||738||564||130|
Table 2: Reasons for Sprinkler Ineffectiveness1
of information include expert opinion, NFIRS data, surveys by insurance
companies, surveys by researchers, and in-depth fire incident analyses
by selected sectors of the industry. This review will begin with a
review of the effectiveness and reliability for two types of active
systems – sprinklers and fire alarms, then proceed to passive systems.
The effectiveness of sprinkler systems in U.S. fire incidents is summarized in Figure 1.1 As indicated in the figure, when sprinklers operate in fire incidents, only one sprinkler operates in almost 70% of all fires and that one sprinkler is effective in 98% of the incidents. An interesting trend indicated in the figure is that the effectiveness of sprinklers declines with an increasing number of operating sprinklers.
The significant reduction in fire death rates in a wide variety of occupancies with sprinklers is presented in Table 1.1
The fire death rate in sprinklered warehouses is likely an anomaly due
to the small number of fire deaths that occur in warehouses and should
not be used as justification to eliminate sprinklers from warehouses.
The reasons for sprinklers to be ineffective are indicated in Table 2.1 As indicated in the table, the dominant cause for ineffectiveness is the system being turned off.
Figure 2: Fires "Too Small” for Smoke Detector and Sprinkler Operation3
Fire Alarm Systems
Analyzing data from fire incidents that occurred from 2003-2006 in U.S. homes, Hall2 determined fatal casualty rates in homes without operating smoke detectors to be 1.10 deaths per 100 reported fires compared to 0.52 deaths per 100 fire incidents with operating smoke detectors, a decrease of more than 50%. In a study by Milke, et al.,3 the effectiveness of sprinklers and smoke detectors was assessed via an analysis of approximately 200,000 U.S. fire incidents that occurred from 2003 to 2007 in residential, commercial residential, and health-care facilities. The casualty rate (including both fatal and non-fatal casualties) was substantially less in residences with operating sprinklers (2.06 casualties per 100 fire incidents) than operating smoke detectors (3.17 casualties per 100 fire incidents).
indication of the relative sensitivity of smoke detectors and
sprinklers is provided in Figure 2. This indication is gleaned from the
fires that were judged by the individual completing the fire incident
form to be too small for the operation of sprinklers and smoke
detectors. As indicated in the figure, with fewer fires judged to be
"too small” for smoke detector operation than sprinklers, smoke
detectors are characteristically more sensitive to early fire conditions
In a second study by Milke, et al.,4
the response of individuals to the operation of smoke detectors in U.S.
fire incidents in commercial occupancies is indicated in Figure 3. This
study included 30,900 fire incidents that occurred from 2003-2010. As
indicated in Figure 3, occupants responded in only 36% of the fire
incidents where an audible alarm was produced as a result of an
operating smoke detector. In the previous study by Milke et al.3,
involving residential occupancies, using the same categories as
reflected in Figure 3, 83% of the occupants in residences responded to
an audible alarm produced by an operating smoke alarm.
Figure 3: Response of Occupants to Audible
Fire Alarms in Commercial Occupancies4
a survey of experts in the 1980s as part of the Warrington Study in
Australia, the reliability of fire-resistant-rated construction in
commercial occupancies was estimated to be 70% (as compared to
reliabilities of 95% for sprinklers and 75% for smoke detectors).5 Rosenbaum
conducted an analysis of data from U.S. fire incidents that occurred
from 1989-1994 in commercial occupancies to compare the extent of
thermal fire damage where different fire protection systems were
installed. In his study, the extent of thermal damage was divided into
three categories: fires that were limited to the room of fire origin,
the floor of origin or multiple floors (i.e., the structure). The
results of Rosenbaum’s analysis are included in Table 3.
|Fire-resistant-rated construction (FRRC)||77%||4%||19%|
Fire death rate: civilian deaths/1000 fires
Table 3: Extent of Damage in U.S. Fire Incidents, 1989-19945
indicated in Table 3, fires are limited to the room or floor of origin
in 63% of all fire incidents where no protection is provided. In
comparison, fires in buildings of fire-resistant-rated construction are
limited to the room or floor of origin in 81% of all fire incidents, a
difference of 18% over the baseline case consisting of buildings without
any protection systems. In contrast, the proportion of fires limited to
the room or floor of origin is 89% when only a detection system is
installed and 92% when only a sprinkler system is installed.
Photographs depicting the desired performance of a fire door are presented in Figure 4. These photographs were taken during the FEMA World Trade Center analysis during a tour of WTC 5. The office space adjacent to the door had a well-developed fire that consumed virtually all of the combustibles. As illustrated in the photographs, the inside of the stairwell is almost absent any effects of the severe fire that occurred on the other side of the door.
Reliability data for fire doors is available from four sources. FM Global examined 1,600 installations of fire doors.6
Several types of fire doors were included in the survey, including
rolling steel, horizontal sliding on inclined tracks, counterweight
closures or spring closures, vertical sliding and swinging doors.
average, 82% operated properly, with rolling steel doors having the
lowest reliability of 80% and vertical sliding doors having the greatest
(93%). CIGNA Property and Casualty Loss Control staff evaluated the
in-place performance of 805 fire doors. "41.1% of all doors had some
type of physical or mechanical problem that would prevent them from
operating properly during a fire event”.7 This results in a modest reliability of 58.9%.
Figure 4: Fire Door Performance in WTC 5 during 9-11 (on left looking
inside stairwell, on right viewed from inside stairwell)
Dusing, Buchanan and Elms8
surveyed 91,909 installed fire doors in several occupancies. Of those
doors, 12,349 (13.4%) were propped open. The problem was at its worse in
institutional occupancies (39% of the doors were propped open) and
least in assembly occupancies (only 5% of the doors were propped open).
Scarff reported that fire doors were blocked open or had kick-down stops
in 50 of the 275 hotels surveyed (18%) from June 1992 to June 1993.9
The performance of a fire damper is depicted in Figure 5 in WTC 5. This damper is near the location of the stairway door included in Figure 4. In this photograph, the damper has closed with the steel HVAC duct having collapsed away from the wall.
Figure 5: Fire Damper Performance in WTC 5
Through Penetration Seals
In a survey of large loss fires in the early 1970s, it was found that in 38% of all large loss fires (based on direct property damage of at least $250K), horizontal fire spread was attributed to non-fire stopped areas including floors and concealed spaces above and below floors and ceilings.10 This was prior to the development of ASTM E81411 (the first edition was in 1981) and may have been part of the motivation for development of the standard test method for through penetration firestop systems. Spruce12 estimated that in buildings that were at least five years old, 95% of the openings in fire-rated barriers were inadequately protected (i.e., a reliability of only 5%).
quality of any data highly depends on the input inserted onto
inspection or incident forms. Further, while data from the U.S. National
Fire Incident Reporting System is publically available, those from
other sources is rarely reported openly. The availability of limited
data or lack of quality data constrains the ability of engineers to make
informed decisions on design options. This constraint can be withdrawn,
but requires a concerted effort starting with those entering the
information from inspections and fire incidents.
James A. Milke is with the University of Maryland.
- Hall, J., "U.S. Experience With Sprinklers,” National Fire Protection Association, Quincy MA, 2011.
- Hall, J. "US Experience with Sprinklers and Other Automatic Fire Extinguishing Equipment,” National Fire Protection Association, Quincy MA, 2009.
- Milke, J., Campanella, A., Childers, C., and Wright, B, "Performance of Smoke Detectors and Sprinklers in Residential and Health-Care Occupancies,” Department of Fire Protection Engineering, University of Maryland, College Park, MD, 2010.
- Milke, J., Hanson, R., and Mizrack, D. "Comparison of the Life Safety Performance of Smoke Detectors and Sprinklers in Commercial, Industrial and Education Institution Housing,” Department of Fire Protection Engineering, University of Maryland, College Park, MD, 2013.
- Rosenbaum, E., "Evaluation of Fire Safety Design Alternatives,” MS Thesis, Department of Fire Protection Engineering, University of Maryland, College Park, MD, 1996.
- "Can You Count on Your Fire Doors?” Record, Factory Mutual Research Corporation, July/August, 1991: pp. 12-15.
- "Fire Doors, Friend or Foe?” CIGNA Property and Casualty Companies, CP-9N70, 1970.
- Dusing, J., Buchanan, A. and Elms, D. ”Fire Spread Analysis of Multi-compartment Buildings,” Research Report 79112, Department of Civil Engineering, University of Canterbury, New Zealand, 1979.
- Scarff, S. "Field Performance of Fire Protection Systems,” Balanced Design Concepts Workshop, Richard W. Bukowski, ed., NISTIR 5264, National Institute of Standards and Technology, Gaithersburg, MD, 1993.
- Fire Protection Handbook, National Fire Protection Association, Quincy, MA, 1974.
- ASTM E814, Standard Test Method for Fire Tests of Penetration Firestop Systems, ASTM International, West Conshohocken, PA, 2013.
- Spruce, C. "In Praise of Passive Fire Protection,” Fire Prevention, London Fire Protection Association, 32, 1994, p. 275.