SPRINKLERS: INTERPRETING THE STATISTICS

BY RICHARD R. LICHT

Automatic sprinkler systems are highly regarded as an effective fire protection tool that can reduce loss of life and property damage in a fire. Based on reliability figures and statistics provided by the National Fire Protection Association (NFPA), the National Fire Sprinkler Association, and other groups interested in the subject of sprinkler performance, architects and building owners are relying more heavily on automatic sprinkler systems.

This trend is driven by the level of confidence designers and building owners place on sprinkler reliability. However, before you can evaluate statistics purporting to show sprinkler reliability, you must understand how sprinklers are intended to operate and what can be expected from their operation.

Automatic fire sprinklers are individually heat-activated and tied into a network of piping with water under pressure. When the heat of a fire raises the sprinkler temperature to its operating point (usually 165°F), a solder link will melt or a liquid-filled glass bulb will shatter to open a single sprinkler, releasing water directly over the source of the heat.

Sprinklers operate automatically in the area of fire origin, preventing a fire from growing undetected to a dangerous size while simultaneously sounding an alarm. Automatic fire sprinklers keep fires small until additional fire suppression can be brought to bear (usually firefighters). Although small fires may be extinguished by automatic fire sprinklers, it is important to keep in mind that automatic fire sprinkler systems are designed to control but not extinguish fires.

TWO MEASURES OF SPRINKLER RELIABILITY

Statistics purporting to demonstrate the reliability of sprinklers often are cited publically, but these statistics usually are not broken down in terms of the two factors that govern the overall reliability of automatic sprinkler systems-operational reliability and performance reliability.

  • Operational reliability is a measure of assurance that a system or component will operate as intended when needed.
  • Performance reliability is a measure of the adequacy of the system once it has operated. In other words, did the system perform in accordance with its design and purpose?

Total reliability may be expressed as follows: operational reliability multiplied by performance reliability.

Most of what we know about national fire trends and patterns comes from the National Fire Incident Reporting System (NFIRS). NFIRS is the world’s largest collection of information on incidents to which fire departments respond. The NFIRS database comprises roughly one-half of all reported fires that occur annually.

Nationally, more than 18,000 (out of an estimated 31,000) fire departments report in the NFIRS each year. Participating departments report an average of 11 million incidents, including 800,000 fires, each year. State participation in NFIRS is voluntary, but nearly all states participate. However, there is no system for collecting information on sprinkler or building performance during fire incidents responded to by the thousands of fire departments that do not participate in the NFIRS.

Another source of fire data is the NFPA survey, which is based on a random sample of roughly 3,000 U.S. fire departments. It includes the following information:

  • Total number of fire incidents, civilian deaths, and civilian injuries and total estimated property damage (in dollars) for each of the major property use classes defined by NFPA 901, Standard Classifications for Incident Reporting and Fire Protection Data.
  • The number of on-duty firefighter injuries, by type of duty and nature of injury.
  • Information on the type of community protected and the size of the population protected, which is used in the statistical formula for projecting national totals from sample results.

CALCULATING NATIONAL STANDARDS USING NFIRS AND NFPA DATA

Many people erroneously believe that the data reported for sprinklers and other fire protection systems are actual hard data. However, the available data and the resulting statistics are not actual numbers but projected figures that may include discrepancies.

National estimates are calculated using the NFIRS data reports supplemented by the annual stratified random-sample survey of fire experience conducted by the NFPA, which is used for calibration. Most national estimates analyzed for this article allocate unknowns reported in their system-i.e., reports in which actual data are not given but are labeled “unknown.” This allocation of unknowns is accomplished by using scaling ratios defined by NFPA survey “estimates of totals” divided by only those NFIRS fires for which the dimension in question was known and reported.

For example, the scaling ratio for 2002 civilian deaths in residential structures is equal to the total number of 2002 civilian deaths in residential structure fires reported to fire departments, according to the NFPA survey (2,695), divided by the total number of 2002 civilian deaths in residential structure fires reported to NFIRS (1,029). Therefore, the scaling ratio is 2,695/1,029 = 2.62, which represents a substantial discrepancy.

Another source of possible error has to do with the practice of rounding numbers. The possibility of rounding errors exists in all calculations. The impact of rounding is greatest when the stated number is small relative to the degree of rounding.

For example, if a table presented five-year averages (all rounded to the nearest one) of estimated deaths by the item first ignited, and there were a total of 30 deaths in the five-year period covered by the table, the total average would be 30/5 = 6. If 10 of the possible items first ignited each accounted for three deaths in five years, there would be 10 entries of 3/5 = 0.6, rounded to one, and the sum would be 10 instead of the actual total of six.

NFIRS REPORTING FORM REVISIONS

The forms for gathering NFIRS data are constantly being revised and updated to improve on the information available. Unfortunately, it takes time to get all involved to switch to the new forms and to get the data to the proper end source.

NFIRS Version 4.1 had one data element for sprinkler performance, with six values (Table 1). NFIRS Version 4.1 did not separate types of suppression systems, so one could not separate automatic fire sprinklers from other types of suppression systems. NFIRS Version 5.0 (Table 2) has multiple data elements, including two that divide the information on presence and operation of sprinklers that are contained in only one data element in Version 4.1.


Table 1. NFIRS Version 4.1 Reporting Form
# Sprinkler Performance
1 Equipment operated
2 Equipment should have operated but did not
3 Equipment present but fire too small to require operation
8 No equipment present in room or space of fire origin
9 Performance of automatic sprinklers not classified above
0 Performance of automatic sprinklers undetermined or not reported

Click here to enlarge image

Sprinkler reliability based on NFIRS 4.1 is based solely on sprinkler operational reliability (assurance of operation) and does not include data on performance reliability (effectiveness in operation). Thus to calculate the percentage of operational reliability, you need to divide # 1 by the sum of # 1 and # 2 times 100.

Sprinkler reliability based on data from NFIRS Version 5.0 can now include both operational reliability (M3 – 1, 2, and 4) and performance reliability (M3 – 1, 2). Using the NFIRS version 5.0, you can tell what type of automatic extinguishing system was present as well as its operational reliability and its performance reliability.

NFIRS changed from Version 4.1 to Version 5.0 in 1999, but not every reporting entity made the switch to the new reporting form, which means that there are limited data from NFIRS Version 5.0.

The NFPA report (“U.S. Experience with Sprinklers and other Fire Extinguishing Equipment,” August 2005, by Kimberly D. Rohr and John R. Hall, Jr.1) was the first NFPA report with data from NFIRS version 5.0. The report covers two periods of data, 1989-1998 based on NFIRS Version 4.1 and 1999-2002 based on data from NFIRS Versions 5.0 and 4.1, converted data. In this report, sprinkler reliability was 84 percent for the data from 1989-1998 and 82 percent (un-recoded2) for the data from 1999-2002.

NFPA AUGUST 2005 SPRINKLER REPORT

One must read very carefully to get a clear picture of overall reliability of sprinklers from the August 2005 NFPA report. The report focused on data3 stating a sprinkler reliability of 93 percent, which is accurate if you consider only the recoded data and limit this consideration to operational reliability. To provide a more complete picture of reliability, NFPA could have pointed out that with NFIRS Version 5.0 we can now obtain total sprinkler reliability, which includes both operational and performance reliability, instead of just highlighting the 93 percent operational reliability. The first report based on limited data from version 5.0 of NFIRS should have noted that overall (recoded) reliability for automatic fire sprinklers was in the range of 89 percent when both operational and performance reliability were considered.

Coding of sprinkler data may or may not be misleading, but the August 2005 NFPA report recoded sprinkler data without providing the actual numbers changed. The report indicates that, in all cases, sprinkler systems are not designated with a “failure to operate” or a “failure to perform” if the sprinklers were not present in the area of the fire, in accordance with NFPA 13R, Standard for the Installation of Sprinkler Systems in Residential Occupancies up to and Including Four Stories in Height. This raises significant questions: Is this a system where the sprinklers are only designed to be installed in limited areas, or was the design inadequate when it failed to operate? The NFIRS would not provide enough data to make this determination, so eliminating all the data from the statistics labeled as “not in the area of the fire” may not be correct.

The August 2005 NFPA report4 does not provide the same level of data as was provided in previous reports; it lists percentages, not actual numbers. Therefore, it is not possible for a person reading the report to see what is accomplished by recoding the data or how significant the data are in actual numbers vs. the older NFPA reports on sprinkler reliability.

SIGNIFICANCE OF USING PROPER SPRINKLER DATA

The significance of accurate sprinkler reliability data is clear when one sees that the voting members of the International Code Council (ICC) often rely on these data as one means to justify code changes. A number of sources are using the single failure criteria of “operational reliability” (93%) when they speak or write about sprinkler reliability, simply because of the way the NFPA reported it.

According to the National Fire Sprinkler Association’s Guide to Fire Sprinklers, there are more than 450 trade-offs in the 2003 International Building Code, many based on a misinformed understanding that sprinkler system failure is very low. To reduce the cost of construction, the codes have removed or eliminated many passive protection features when sprinklers are installed. Does using the data alone justify all the trade-offs when it is known that sprinklers can fail and leave building occupants vulnerable in a fire?

Promoting sprinkler performance and effectiveness has been a key in the successful marketing of sprinklers as a lifesaving tool. However, there is an ongoing debate among professionals in the fire protection industry as to what constitutes a “sprinkler failure” or “satisfactory sprinkler performance,” since both terms are ambiguous and subjective.

Until 1970, the NFPA measured sprinkler effectiveness by looking at the percentage of fires in sprinklered properties showing satisfactory sprinkler performance. This statistic was then discontinued by NPFA because of the possibility of a growing bias in the database from 1925 through 1969. The NFPA claims that cases from that period produced larger fires that were more likely to require fire department intervention or insurance company attention and therefore were more likely to be reported to the NFPA. It is theorized that small- and medium-sized fires would be handled so successfully by sprinklers that those fires would not be reported. The NFPA’s data-collection procedures could therefore be biased toward cases of poor sprinkler performance.

Based on that assertion, the NFPA decided to discontinue tracking sprinkler failures because of its belief that this bias toward poor sprinkler performance was becoming worse each year and was causing some people to believe that sprinkler effectiveness was declining. To adjust the public’s perception, the NPFA began using the data to report sprinkler effectiveness in terms of lives or property saved, rather than analyzing successes and failures of both active and passive systems.

Whereas many believe that this means of promoting sprinklers has justified the “end”-a national campaign to put the power of sprinklers to work-others believe that NFPA’s decision to report sprinkler performance based on lives or property saved instead of on actual system performance has created a reverse bias which has led people to believe that sprinklers are a panacea with a record of near perfection to the detriment of other essential elements of fire protection.

There also is a conspicuous 35-year gap of vital and useful information about active and passive fire protection features. In reality, the data indicate that all elements of fire protection tend to show more problems with human error and inferior design, installation, or maintenance than with intrinsic mechanical or electrical reliability. However, the NFPA’s reporting method has caused some to overstate the effectiveness of a sprinkler system by focusing too narrowly on the reliability of the components of the system or by giving sprinklers credit for life savings that are actually produced by the whole integrated system of balanced fire protection in which sprinklers are an essential part but not the only part.

SYSTEM DESIGN SHOULD NOT RELY SOLELY ON STATISTICS

Collecting the available data, the use of scaling ratios, rounding up or down, projecting based on estimates rather than actual data, and other methods of distilling the available information have had a significant effect on how sprinkler performance data are interpreted by code enforcement, fire services, and design professionals.

Collecting the available data, the use of scaling ratios, rounding up or down, projecting based on estimates rather than actual data, and other methods of distilling the available information have had a significant effect on how sprinkler performance data are interpreted by code enforcement, fire services, and design professionals.

As indicated above, the new version of NFIRS 5.0 provides more fire data on sprinklers and other extinguishing equipment than the older version, NFIRS 4.1. The new version of NFIRS as used in the NFPA reported time-frame has limited data (not all jurisdictions are using the new NFIRS Version 5.0) compared with the older version of NFIRS. Unfortunately, the August 2005 NFPA report focuses just on operational reliability and not the total reliability now available with the data from NFIRS Version 5.0. For a more complete picture of sprinkler reliability, one needs to consider operational as well as performance reliability.

If it is clarified that the actual sprinkler performance data collection system is limited and that the resulting statistics are based on the projections from estimates, the total recoded reliability of sprinklers (based on performance and operational reliability) is most likely to be approximately 89 percent. In other words, the sprinklers fail once in every nine fires.

Whether one is convinced that the reliability of automatic sprinkler systems is 89 percent or some other value, the NFPA data seem to indicate that the commonly stated reliability of automatic sprinkler systems in the range of 96 percent (fails once in every 25 fires) overstates the reliability of sprinkler systems.

A properly installed and maintained sprinkler system is an effective fire protection tool that should be designed as part of a system. This system should not solely be based on statistics that can be manipulated but on a proven, balanced design in which the system has the support and reinforcement of other integrated fire and smoke protection features, based on the needs of the owner, the occupants, and the community.

Endnotes

1. Rohr, Kimberly and John R. Hall, Jr., “U.S. Experience with Sprinklers and Other Fire Extinguishing Equipment,” National Fire Protection Association, August 2005.
2. Recoding – refers to changing the data based on other information.
3. USFA/National Data Center, “Fire in the United States 1992 – 2001,” Thirteen Edition, Oct. 2004.
4. U.S. Fire Administration, “2005 NFIRS Data by Incident Version,” www.usfa.fema.gov/nfirs/status/data_wers.shtm.

RICHARD R. LICHT is the technical director of the Alliance for Fire and Smoke Containment and Control (AFSCC) in Tarrytown, New York. He has more than 30 years of experience with fire protection methods, materials and testing, and fire code standards. The AFSCC was established in 1999 by building enforcement, construction, design, and manufacturing professionals in response to the need for a well-coordinated educational effort to promote the value of balanced fire protection design in the built environment. Its members consist of companies, organizations, and individuals in the construction industry with an interest in fire safety.

No posts to display