Escolar Documentos
Profissional Documentos
Cultura Documentos
buildings
Lougheed, G.D.
NRCC-40106
www.nrc.ca/irc/ircpubs
The following paper was published in ASHRAE Transactions, Vol.
103, Pt. 1. 1997 American Society of Heating, Refrigerating
and Air-Conditioning Engineers, Inc.
ABSTRACT and only two operating sprinklers, the heat release rate, radiant
flux, room temperatures, and buoyancy pressures were substan-
As part of a joint research project between ASHRAE and
tially reduced in a compartment. It was also noted that, although
the National Research Council of Canada (ASHRAE Research
total smoke production was reduced by the sprinklers, there were
project RP-838), the probability of occurrence and expected
still significant quantities of “cold” smoke produced and that this
size of shielded fires in sprinklered high-rise office buildings
smoke contained potentially harmful concentrations of carbon
were investigated. Phase 1 of the project included surveys of
monoxide.
typical office buildings to identify shielded fire situations and
The previous investigation (Mawhinney and Tamura 1994)
the fuel loads in the shielded areas. Based on these surveys,
identified a number of areas in which additional research was
medium- and full-scale fire tests were conducted. The results
needed. In 1994, a second joint research project was initiated
of these investigations are summarized in this paper. In partic-
between ASHRAE and NRCC (ASHRAE research project RP-
ular, the effect of a sprinkler system on heat release rate, com-
838). This current study was established to address one of the
partment temperatures, carbon monoxide (CO) production,
fundamental concerns about the previous study raised by the fire
and smoke generation and movement for fires in an open-plan
protection engineering community—namely, what are the prob-
office scenario are discussed. An initial analysis of the impli-
ability of occurrence and probable size of shielded, sprinklered
cations of a sprinkler system on a zoned smoke control system
fires in high-rise office buildings? The objective of the investi-
is also provided.
gations conducted in this second project was to provide infor-
mation that would answer those questions and thus improve
INTRODUCTION
confidence in the design of smoke management systems in
The ability of sprinkler systems to control and extinguish buildings protected by sprinklers. In particular, in this paper, the
unshielded fires is well established. However, in recent years, effect of a sprinkler system on heat release rate, compartment
concerns have been raised regarding shielded fires that are not temperatures, CO production, and smoke generation and move-
rapidly extinguished by sprinklers (Klote 1990). It was ment for fires in an open-plan office scenario are discussed. An
suggested that such fires could produce significant quantities of initial analysis of the implications of the sprinkler system on a
smoke and toxic fire gases; therefore, smoke control systems, in zoned smoke control system is also provided.
addition to the automatic sprinkler systems, may be useful for
life safety of building occupants. SHIELDED FIRE SCENARIOS
In 1991, a joint research project was initiated between the In phase 1 of the project, three tasks were completed:
American Society of Heating, Refrigerating and Air-Condition- • Surveys were conducted to quantify typical shielded fire
ing Engineers, Inc. (ASHRAE) and the National Research scenarios in high-rise office buildings.
Council of Canada (NRCC) to investigate the effects of sprin- • A review of North American fire loss databases was
klers on the performance of a zoned smoke control system undertaken to identify shielded fire incidents in sprin-
(ASHRAE research project RP-677). The results are docu- klered buildings.
mented in the final report for the project (Mawhinney et al. 1992) • Recent literature on fire statistics and full-scale tests with
and are summarized in a paper by Mawhinney and Tamura sprinklers and smoke management systems was
(1994). These investigations indicated that with shielded fires reviewed.
Gary D. Lougheed is a senior research officer at the National Fire Laboratory, Institute for Research in Construction, National Research Coun-
cil Canada, Ottawa, Ont.
Instrumentation for Open-Plan Office Fire Tests 2. At two minutes, the propane burner was ignited. The burner
design and heat release rate (approximately 25 kW) were
Instrumentation for the full-scale open-plan fire tests based on the ignition source used for standard tests for
included the following: furniture (ASTM 1994).
• Seven thermocouple trees were located in the test facility 3. At five minutes, the propane burner was shut off.
as shown in Figure 2. The thermocouples on each tree
4. The sprinklers were allowed to activate normally and the
were located 0.15, 0.61, 1.22, and 1.83 m (0.5, 2, 4, and
water flow rate was adjusted manually to provide the
6 ft) below the ceiling. Metal shields (approximately 400
required flow rate.
mm by 400 mm [1.3 ft by 1.3 ft]) were placed above each
thermocouple to shield it from direct water spray from 5. All tests were continued until the temperatures measured in
the sprinkler. the shielded areas indicated minimal fire remained in these
• Thermocouples were located adjacent to the fusible ele- locations.
ment of each of the four sprinklers (Figure 1).
• Pressure taps to measure the differential pressure MEDIUM-SCALE FIRE TEST RESULTS
between the test facility and the burn hall were located at A series of 23 medium-scale fire tests was conducted with
three heights on the west wall of the test facility (Figure a horizontal shielded area representative of a typical table or desk
2). The pressure taps were located 0.15, 1.22, and 2.29 m (0.9 m2 [9.6 ft2]). The medium-scale tests were conducted with
(0.5, 4, and 7.5 ft) below the ceiling. selected portions of the fuel packages used for the full-scale fire
tion of less than one minute. At this temperature, the pressure door to the stairshaft was opened, the stairshaft was rapidly
at or shortly after initial sprinkler activation would exceed the contaminated with smoke, even at the low buoyancy pressures
12.5 Pa minimum pressure difference recommended in NFPA produced by the sprinklered fires.
92A (NFPA 1993). However, the time during which the pres- In addition to the temperatures measured in the test facility,
sure exceeded 12.5 Pa (temperature increase > 200°C the temperatures were also measured in the duct in tests with and
[392°F]) was only approximately 60 seconds. After this initial without the exhaust operating. Peak temperatures were up to
high-temperature spike, the temperature increase at the ceiling 125°C (257°F) at, or shortly after, sprinkler activation. Full-scale
was less than 200°C (392°F) and, in many cases, less than tests with a propane burner system indicated that heat release
100°C (212°F) even in the immediate fire area. These results rates of more than 1,000 kW would be required to sustain
indicate that the 12.5 Pa minimum design pressure suggested temperatures in excess of 100°C (212°F) in the duct with the
by NFPA 92A (NFPA 1993) would be adequate for zone sprinkler system operating (Lougheed and Carpenter 1996).
smoke control applications in sprinklered office buildings. In
fact, in most cases, a lower pressure (approximately 7 Pa) CO Concentrations
would still exceed the pressure produced by the fire gases. Incapacitation by CO depends upon the dose accumulated
The temperature/pressure results measured in the full- over a period until carboxyhemoglobin concentrations exceed
scale tests are consistent with those obtained by Mawhinney incapacitation levels (Purser 1995). The intake of CO is a func-
and Tamura (1994) in the tests conducted in a single compart- tion of a number of factors, including the level of activity of the
ment under steady-state conditions. They measured pressure subject. Estimates for tenability for incapacitation by CO
differences of 4 to 6 Pa in the tests in a compartment with the provided by Purser (1995) are: 6,000 to 8,000 ppm (0.6% to
sprinkler system providing the design density for a light hazard 0.8%) for 5-minute exposures and 1,400 to 1,700 ppm (0.14% to
occupancy. 0.17%) for 30-minute exposures.
It is accepted practice to only pressurize adjacent floor The CO concentrations measured at the ceiling at the north-
zones and to rely on mechanical exhaust from the fire zone and west corner and southeast/center of the room are shown in
leakage from the pressurized floors into a stairshaft to develop a Figures 5 and 6, respectively. The CO levels measured in these
pressure differential between the stairshaft and the fire floor. tests were in the range that could produce incapacitation with
Tests conducted in a 10-story tower facility (Mawhinney and extended exposure. They were, however, by a factor of two to
Tamura 1994) indicated that a zoned smoke control system with three lower than the peak levels measured by Mawhinney and
a pressure difference of 12.5 Pa was able to prevent smoke Tamura (1994). There are a number of factors that can affect the
spread into a stairshaft or between floors from a shielded fire as CO levels, including the heat release rate from the fire, the type
long as the door to the fire floor was closed. However, when the of fuel, the volume of the test space, and the ventilation rate.
Further analysis is required to fully assess the differences production as the primary fuel load changed from the cardboard
between the two results. boxes to the paper stock contents. As with the full-scale tests, the
In the current tests, the CO levels remained relatively rate of CO production did not decrease even though there was a
constant over an extended period (20 to 30 minutes) even though decrease in the heat release rate with the sprinklers activated
the heat release rate was decreasing. The CO2 levels, measured (Lougheed and Carpenter 1996).
at the same locations in the test facility as the CO measurements, The CO results obtained from the full- and medium-scale
indicated that there was a higher concentration of the fire gases tests do not indicate an increase in CO production in an office fire
in the upper corners of the test facility. However, the general with sprinklers. In fact, taking into consideration recent studies
trend showed a decreasing concentration consistent with the that indicate that, for furniture (Sundström 1995), the rate of CO
decrease in heat release rate (Lougheed and Carpenter 1996). As production is related to the heat release rate, a considerably
such, the steady-state CO results indicate that, although there higher rate of CO production would be expected for an uncon-
was a reduction in the heat release rate with decreasing fire load trolled office fire without sprinklers.
in the shielded area and from the effects of the sprinklers, the Further analysis of the full-scale test results is required
smoldering fire in the paper stock in the shielded areas, which before a CO production rate based on the suggested design fire
was the primary source of CO, was not affected by the sprinklers. can be determined. However, initial estimates can be determined
These observations are consistent with the results obtained using the CO concentrations measured in the exhaust duct
in the medium-scale tests with paper stock fuel loads. In the (Figure 7). For the tests with the fan operating, CO concentra-
medium-scale tests, it was noted that there was an increase in CO tions were 0.05% to 0.15% in the duct. Assuming an air density
of 1 kg/m3 (a temperature of approximately 350 K) and a flow 0.2 OD/m, respectively, for subjects not familiar and
rate of 3 m3/s in the exhaust duct, the CO measurements indicate familiar with an escape route.
approximately 0.1 to 0.5 g/s of CO were produced in the full- • Rasbash (1975) suggested a 10-m (33-ft) visibility limit,
scale tests. The initial CO production rate is consistent with the which is equivalent to 0.08 OD/m.
results of furniture calorimeter tests on office chairs conducted • Babrauskas (1979) suggested a tenability limit of 0.5
by Sundström (1995). Office chairs with peak heat release rates OD/m.
of 486 kW and 829 kW produced peak CO rates of 0.33 and 0.46 • Wakamatsu (1968) suggested a tenability limit of 0.043
g/s, respectively. OD/m, which was assumed to be 1:100 of the optical den-
sity in the fire region.
Smoke Production and Movement
The smoke obscuration measured in the full-scale open-plan
The smoke obscuration for an open-plan office test office tests indicated the smoke levels in the test facility were
measured in the northwest and southeast corners of the test facil- in the range of 1 to 10 OD/m. These levels were higher than
ity is shown in Figures 8 and 9, respectively. The smoke obscu- the generally accepted tenability limits (0.045 to 0.5 OD/m)
ration measured in the exhaust duct is shown in Figure 10. suggested above.
Various tenability limits for smoke obscuration have been There is limited information available regarding the maxi-
proposed as follows: mum level of smoke likely to be found in fires. Tamura (1994)
• Jin (1981) suggested tenability limits of 0.06 OD/m and suggests, based on a review of existing data, that 4 to 10 OD/m
could be considered as the limiting level for adverse fire condi- concentration measured in the duct were in the range of 1 to 2
tions. As such, the smoke levels measured in the test facility are OD/m and 0.2% to 0.25%, respectively (Figures 7 and 10).
comparable to those noted in the literature for tests without sprin-
These results are consistent with those obtained by
klers. Taking into consideration the lack of mobility of cold
Mawhinney and Tamura (1994). They noted that, in the tests
smoke, this result is not unexpected and, as with the CO gener-
conducted in the tower facility, the use of a zoned smoke control
ation, does not indicate that smoke production is comparable to
system confined the smoke to the fire floor when the door to the
that of the unsprinklered fires.
stairshaft was closed. For these tests, a mechanical system was
In test 31, there was smoke flow through the exhaust duct
used to exhaust the fire floor at a rate of 4.8 to 5.1 air changes per
even though the fan was not operating. The volumetric flow rate
hour (ACH). However, for tests without smoke control, smoke
is shown in Figure 11. The duct was connected to the outside of
did spread to floors above the fire floor through leakage open-
the main test building, with the initial airflow produced by the
ings. Also, the stairshaft became smoke-logged within minutes
temperature difference between the test facility and outside
after opening the door, with complete loss of visibility and
ambient conditions. The volumetric flow measurements indicate
temperatures of more than 65°C (149°F).
a peak flow rate of up 0.25 m3/s (8.82 ft3/s) through the duct.
Since the duct had a cross-sectional area of 0.25 m2 (2.7 ft2), For the tests with the smoke control system, the mechanical
these results indicate a maximum smoke flow rate of approxi- exhaust of the fire floor was not sufficient to prevent the spread
mately 1 m3/s (35.3 ft3/s) per square meter (square foot) of leak- of smoke into the stairshaft when the door on the fire floor was
age area across a smoke barrier. For approximately 10 minutes opened. The flow of smoke into the stairshaft could be stopped
after first sprinkler activation, the smoke obscuration and CO by providing a minimal flow of air into the stairshaft by either
opening one of the doors on a pressurized floor or by opening the furniture arrangement. The results of these tests, including the
door to the building exterior at the bottom of the shaft. impact on zoned smoke control systems, are summarized as
The potential impact of the smoke flow through leakage follows.
areas will depend on smoke movement and dilution in the build-
1. Fire Spread. The sprinklers were able to control the fire
ing outside the fire zone. Without computer modeling, it is diffi-
within the limits of the 3 m by 3 m (10 ft by 10 ft) office area.
cult to assess the extent of the smoke hazard in a multistory office
However, the fire was able to spread to secondary shielded areas
building. Computer studies similar to those conducted by Waka-
in the same office while the sprinklers were operating. After the
matsu (1968) and McGuire et al. (1970) for nonsprinklered
first sprinkler was activated, the fire was localized in the areas
buildings are required. The full-scale open-plan office tests can
shielded by the desks and tables.
be used to determine the conditions (heat release rate, tempera-
ture, pressure, CO production rate, and smoke generation) for a 2. Heat Release Rate. Peak heat release rates of up to 800
design fire for use with a multizone model, such as CONTAM to 900 kW were measured for tests that initially involved two
(Walton 1993), to investigate the extent of the hazard in a sprin- shielded areas (desk/table). The heat release subsequently
klered office building. decayed exponentially, with heat release rates of 100 to 200 kW
measured 25 to 30 minutes after ignition. The heat release rates
SUMMARY measured in the open-plan office tests were consistent with those
Zoned smoke control systems are used to limit smoke flow measured in a series of medium-scale tests and with the office
through cracks in floors and partitions and through nonpressur- tests conducted by Madrzykowski and Vettori (1992a, 1992b).
ized shafts that threatens life and damages property at locations Based on the heat release rates measured in the open-plan office
remote from the fire (NFPA 1993; Klote and Milke 1992). In fire tests, a heat release rate for a design fire for use in sprinklered
addition, measures to limit fire size (mass burning rate) can be office buildings was developed. This design fire is consistent
used to increase the reliability and viability of such systems with the suppression algorithm developed by Madrzykowski
(NFPA 1993). Methods to limit fire size include fuel control, and Vettori (1992a, 1992b) and the steady-state heat release rate
compartmentation, and automatic sprinklers. In this paper, the used in the U.K. (Hansell and Morgan 1994) (see Figure 12). The
effects of a sprinkler system on heat release rate, compartment proposed design fire, in principle, assumes an initial fire involv-
temperatures and pressures, CO production, and smoke genera- ing two shielded areas subsequently spreading to two secondary
tion and movement were summarized based on the results of a shielded areas and thus could be considered a conservative chal-
series of full-scale open-plan office fire tests. lenge for a light hazard sprinkler system in an office building.
Seven full-scale open-plan office tests were conducted The proposed heat release rate curve, determined on this basis,
using a “typical” 3 m by 3 m (10 ft by 10 ft) open-plan office is less conservative than the U.K. steady-state design fire but
more conservative than the suppression algorithm developed by Higher smoke levels were measured in the duct in test 31 when
Madrzykowski and Vettori. the fan was not operated. Also in this test, there was smoke flow
3. Temperature and Pressure. The highest temperatures through the duct during the test. These results indicate that the
in the test compartment were measured at or shortly after the operation of a sprinkler system is not sufficient to completely
activation of the first sprinkler. The highest temperatures stop smoke flow through leakage areas in a smoke barrier.
measured at the ceiling in the immediate fire zone were up to However, the potential impact of such smoke flow will depend
325°C (617°F). Such temperatures could result in pressures on the smoke movement and dilution in the building outside of
slightly higher than the 12.5 Pa generally required by standards the fire zone. Computer studies similar to those conducted by
for smoke control systems in office buildings. However, the Wakamatsu (1968) and McGuire et al. (1970) would be required
duration for temperatures in excess of 200°C (392°F) was to determine the extent of the hazard of the smoke in a multistory
limited to less than 60 seconds. As such, the 12.5 Pa minimum office building.
design pressure suggested by NFPA 92A (NFPA 1993) should 6. Opposed Airflow. The airflow through the ventilation
be adequate for zoned smoke control system design. In most openings produced by the exhaust fan was sufficient to keep the
cases, a minimum design pressure of 7 Pa would be sufficient. smoke produced by the sprinklered fires from leaking through
4. CO Concentrations. The CO levels (0.1% to 0.5%) the openings. However, this may be due, in part, to the effect of
measured in the test facility were generally higher than the long- the exhaust system on temperatures at lower levels (Lougheed
term tenability limits. The CO concentrations remained rela- and Carpenter 1996). Based on temperatures measured in test 31
tively constant throughout the test, with the CO produced in the and the equation for opposed airflow in NFPA 92B (NFPA
latter stages of the test attributed to fires in the paper stock in the 1995), airflow velocities of 0.5 to 1.0 m/s (1.64 to 3.28 ft/s) may
shielded areas. The initial CO production rate was consistent be required to keep the smoke from entering adjacent areas. The
with peak rates measured in furniture calorimeter tests for office results are consistent with the recommendation by Mawhinney
chairs (Sundström 1995). and Tamura (1994) that small airflows between the area to be
5. Smoke Levels and Smoke Movement. Smoke levels in protected and the fire zone were helpful in preventing the spread
excess of the normally accepted tenability limits were measured of smoke.
throughout the test facility shortly after the first sprinkler acti-
ACKNOWLEDGMENTS
vated. With the exhaust fan operating, lower smoke obscuration
was measured in the exhaust duct than in the test facility. It was This research was carried out as a joint research project
suggested that the smoke entering the duct was diluted by fresh supported by the American Society of Heating, Refrigerating
air entering the test room through the ventilation openings. and Air-Conditioning Engineers, Inc., and the National