Invited Lecture

Overview of Sprinkler Technology Research

CHENG YAO
Factory Mutual Research Corporation
Norwood, Massachusetts 02062 USA

ABSTRACT

Sprinkler fire protection has existed for over 100 years. Until recently, however,
understanding how a sprinkler protects against fire had not been very important, because
sprinklers were so greatly overdesigned for fires in industrial buildings of the past. Since
World War 11, the traditional sprinkler has been severely challenged to its limit of effectiveness
by ever-changing manufacturing and storage practices. In response to this challenge, FMRC
initiated a sprinkler technology research program in 1970, to study sprinkler performance
principles. This ongoing research identifies and quantifies the c o n t r o h g variables, such as
RTI (Response Time Index), ADD (Actual Delivered Density) and RDD (Required Delivered
Density), and develops engineering tools and deterministic computer models to predict
sprinkler fire protection performance. Emerging scientific understanding has stimulated new
sprlnkler developments: the Large-Drop Sprinkler (1971-1980), the Residential Sprinkler
(1976-1979), and the ESFR Sprinkler (1984.1986). It is the intention of this paper to present
an overview of continuing sprinkler technology research activities, including some practical
applications.

KEYWORDS: Automatic Sprinkler Systems, Fire Control, Fire Suppression, Sprinkler

Performance Characteristics, Sprinkler Technology, ESFR Sprinkler, Large Drop Sprinkler

INTRODUCTION

The automatic sprlnkler was designed originally in the late 1880's to protect against fires in
combustible ceiling structures of New England textile mills. The old style sprinkler directed
about half of the water upward to protect the ceiling structures; the remainder of the water
and that which bounced off the ceding was distributed toward the fire on the floor. The 13
mm orifice diameter sprinkler was so overdesigned and the loss experience was generally so
favorable, it was essentially unchanged until 1950 when FMRC developed the '-Spray"
sprinkler in recognition of the important role of water distribution in the protection of storage

FIRE SAFETY SCIENCE-PROCEEDINGS OF THE FIFTH INTERNATIONAL SYMPOSIUM, pp 93-1 10 93

Copyright © International Association for Fire Safety Science

fires. The spray sprinkler directed all the water downward toward the fire on the floor"'. The
spray sprinkler was later accepted by the NFPA as the "Standard" sprinkler in 1953, replacing
the old style sprinkler.

Since World War 11, there have been continuous changes in manufacturing and storage
practices, especially the introduction of lift trucks to stock more hazardous synthetic materials
in higher buildings, dramatically increasing the challenge to sprinkler protection systems.
Under these circumstances, a relatively large number of sprinklers are supplied with sufficient
water to control and limit the fire spread within a particular area by keeping the surrounding
combustibles wet enough so that they do not ignite. This defensive approach was in practice
for many years before it was formally defined as the "Fire Control Concept" in 1984(". Under
the control concept, the protection requirement for each storage condition has to be
individually determined. FMRC built the world's largest sprinkler fire test facility in 1967 to
seek immediate solutions to storage problems through large-scale rack storage and plastic
storage fire test programs, which were expensive. In order to provide the data needed with a
reasonable number of fire tests, a concept of "Parallelism" was adopted by a committee
governing the "Rack Storage Program". This involved the establishment of a base density /
area of demand curve for a standard test commoditv and a set of test conditions using a
specific sprinkler brand. Additional curves for other commodities and variables, such as aisle
width, sprinkler temperature rating, and type of rack, were then constructed through a single
test point drawn parallel to the base curve. All of the tests were conducted with the ignition
source centered below four sprinklers. These guidelines are complex; NFPA 231C, Standard
for Rack Storage of ~aterials(~', has 14 figures and 7 tables. By d e f ~ t i o n the
, density/area
rule assumes that for a given water density, the performance of all listed sprinklers in a given
category is the same, regardless of their make, orifice size, spacing, or pressure.
Unfortunately, over the years, test results have shown that different sprlnkler models and
ignition locations can cause a significant difference in area demand^'^'. Furthermore, fire tests
in the Plastic Storage Program revealed that rack storage of a plastic commodity over 4.6 in
high cannot be protected without supplementing ceiling sprinklers with in-rack sprinklers. In-
rack sprinkler systems, however, are expensive, and a warehouse incorporating them is
difficult to operate.

In summary, the traditional sprlnkler system has been challenged to its limit of effectiveness,
and the areatdensity rule which has been used as the basis for sprinkler system design is no
longer appropriate. In spite of hundreds, if not thousands, of large-scale fire tests under
various conditions, very little was known and very few adequate conclusions could be drawn.
At the same time, the urgent need for improved fire protection in residential occupancies had
raised doubt as to the suitability and effectiveness of an industrial sprinkler in residential and
office buildings.

FMRC initiated a long-range sprinkler technology research program in 1970 to study the
principles of sprinkler fire protection in an effort to provide a sound technical basis for solving
the above mentioned problems. The knowledge which came out of this study has been used
to formulate guidelines for developing new and effective fire control and fire suppression
sprinkler systems: the Large-Drop sprinkler system (1971-1980) to control high-challenge
storage fires, the Residential sprinkler system (1976-1979) to maintain a survivable
environment in residential areas, and the Early Suppression Fast-Response (ESFR) sprinkler
system (1984.1986) to suppress rather than control high challenge fires.
SPRINKLER TECHNOLOGY RESEARCH

In order to establish a scientific and quantitative basis for the understanding of sprinkler fire
protection principles, detailed knowledge of the following conditions is needed: 1) fire
growth inside the storage array, 2) buoyant f r e plume above the storage array, 3) transient
ceiling flows, 4) sprinkler response, 5) sprinkler spray and fire plume interaction, and 6) fire
suppression criteria.

Fire Challenge at the Time of Sprinkler Operation

Most of the heat detector response prediction models(5,6,7) determine the detector response
from cumulative convective heat transfer to the heat sensing element from the ceiling flow.
Ceiling flow data have been measured for small-scale steady fires or T-square fire, i.e., fires in
which heat release rate increases with time to the second power. All the response prediction
models assumed that the heat sensing element is located at an elevation where the ceiling flow
temperature and velocity are maximum.

In large-scale rack storage fire tests, the first few sprinklers usually operate while the f ~ ise
still confmed to the ignition array, represented by an array two-pallet loads wide and two-
pallet loads deep (2x2 array). A series of free-bum fire tests were conducted with two
different classes of commodities: Standard Class lI1and Standard plasticZ test commodities,
arranged in 2x2 array at three different heights: two, three and four tiers at 2, 4 and 6 m high,
respectively. These fire tests have been conducted under the FMRC Fire Products Collector
(FPC) - a large calorimeter which collects all fire products and determines heat release rates@).
Thermocouples and hi-directional probes were used to measure the fire plume temperature
and velocity pro~es'9'. Some fires were repeated under a large, smooth ceiling to measure
the transient ceiling gas temperatures and vel~cities"~'.Empirical correlation of instantaneous
gas temperature (T,) and velocity (U) of the ceiling flow at the sprinkler location and the
radiative heat gain from the fire plume (Q,) for both transient and quasi-steady state fire
growth characteristics, are used as input data in Equation 1 to predict the response of
sprinkler (T,) at different elevations below the ceiling, for storage fires in large warehouses'").

-d x- - U"'(T-X) + c(T,-T-)
QL -
dt RTI mc RTI

U"' m c
RTI = -
h A

where m is mass of the element, c is specific heat of the element, h is the convective heat
transfer coefficient, A is surface area of the element, T, is gas temperature, and T- is ambient

' Standard Class I1 Commodity is an FMRC standardized test commodity made of double triwall corrugated
cartons with metal liners.
Standard Plastic Commodity is an FMRC standardized test commodity made of polystyrene cups packaged
in compartmented cxdboxd cartons.
temperature. RTI (Response Time Index) represents the thermal responsiveness of the heat
sensing element and C (Conduction Parameter) characterizes the conductive heat loss from
the element to the sprinkler support. The values of RTI"" and c(13) are intrinsic properties of
a sprinkler design and can be determined experimentally in a test tunnel. Testing of numerous
sprinklers shows RTI values ranging from 124 to near 440 m" s"; for fast response
(Residential and ESFR) sprinklers the RTI is 28 m" s". The C value, measured for 11
sprinkler models, ranges from 0.5 to 1.56 mln Is ".

Figure 1 shows the measured and predicted thermal responses of sprinklers at 3 m x 3 m

square spacing and centered directly above a 6 m high rack storage of standard plastic
commodity fire in a 9.1 m high building. The predicted fast response sprinkler link
temperatures at locations both 0.203 and 0.330 m below the ceding agree very well with the
measured values. Arrows labeled as ESFR are connected to flustrate that the ESFR sprinkler
"
with fast response (RTI = 24.3 m s "), low temperature rating (79 OC) link, located at 0.33
m below the ceding and 2.1 m from the fire axis, will respond at 53 s from ignition with a
convective heat release rate of 2.6 MW; the Large Drop sprinkler with conventional response
(RTI = 173), high temperature rating (142 OC) at 0.203 m below the ceiling responded at 78 s
with a convective heat release rate of 4.9 MW. Not shown here is that if an ESFR sprinkler is
centered directly above the same fire, the sprinkler will respond at 50 s and a convective heat
release rate of about 1.5 to 1.8 MW.

The principal elements of challenge that a storage fire presents to a sprinkler system is the
upward momentum of the fire plume. Figure 2 shows the fire plume temperature and velocity
profiles measured at 2.8 m above a 6 m high rack storage of standard plastic commodity fire
at the convective heat release rates of 1.76 MW and 2.64 MW. Here the center-core
momentum flux (Mf") values of 21 ~ l m 'and 24 ~ / were m ~calculated from the local fire
plume density (pf ) and velocity (vf ) data integrated over a 0.093 m2 circular area around the
center-line of the fire plumes.

Sprinkler Performance Characteristics

At the present time, there are a range of sprinkler sizes available:

Sprinkler Type Nominal Orifice % of 112 in. K Factor

Diameter Sprinkler Flow
I / m i d (bar)'" nvm/ (psi) '"
Spray Sprinkler
Small Orifice 10 318 50
Standard Orifice 13 112 100
Large Orifice 14 17/32 140
Extra Large Orifice 16 0.64 200
Large-drop Sprinkler 16 0.64 200
ESFR Sprinkler 18 0.7 250
 Agure 1. Responses of ESFR Sprlulder (fast
response,low temperahlre Ilnk, located 0.33 m
below the cedmg) and Large-Drop SpnnWer
(conventional response, hlgh temperature Ilnk,
located 0.203 m below the callng) for a 6 m
high rack-storage of FMRC Plastic Commodity
fire m a 9 m bmldlng, with ign~tloncentered
bdow four sprinklers.

0 10 20 30 40 50 60 70 80 90
TIME FROM IGNITION ( SEC )

Drop Diameter (mm)

F ~ g m e3 Drop SIC? h t n b u a o n mearmed at 2 bar

$ 5.331 -I
&charge pressure for spray and large-drop sprinklers

1.33
Figure 2. Gaussian plot of fire plume temperahxe and . ,
velocity data measured at 2.8 m above a 6
storage of plastic fire at convect~veheat release rates of - 1.5 - 1.0 -0.5 0.0 0.5 1.0 1.5
1.76 kW and 2.64 kW. Radius ( m )
Volumetric sprinkler discharge rate is usually expressed as: Q, = K P'", where K is flow
constant commonly referred to as K factor, and P is the water gage pressure. The sprinkler
provides a well-defined umbrella-shaped discharge pattern. Water discharge from four
sprinklers, when they are spaced 3 m apart, overlaps at about 1 m below the deflectors. By
definition, design density, DD, is determined from: DD=Q, / A, =Q, / s2where A, is
designed coverage area per sprinkler and S is the spacing in square arrangement.

Fire Plume Penetration - ADD: One of the most important measures of sprinkler performance
is the ability to deliver water drops through the fire plume to reach the burning surface of the
storage array. Under a no-fire condition, water density actually applied (distributed) per unit
area within the designed coverage area is the Local Applied Density (LAD); LAD averaged
within the designed coverage area is the Local Density (LD). The LD at different elevations
below the deflector may be different from the DD. In the presence of a fire, the water density
actually penetrating the fire plume and delivered onto the top cross sectional area of the
burning array is the Actual Delivered Density (ADD). Therefore, the fire plume penetration
performance of a sprinkler can be expressed by a Penetration Ratio (Pe = ADDILAD).

Water drops can penetrate the buoyant fire plume in two modes: the gravity mode and the
momentum mode. In the gravity mode, the plume flow is not affected by the downward
momentum of the water spray: only large drops in the hot plume having a terminal velocity, u,,
greater than the fire plume velocity will penetrate the fire plume to reach the seat of the fire(14).
As shown in Figure 3, the volume median diameter (dm) measured at 2 bar pressure for the
Large Drop sprinkler is about 1.5 mm, versus 0.9 mm for large orifice Spray sprinkler(15).
Figure 4 shows the measured terminal velocity('6) as a function of drop diameter in air at
various temperatures. For fires in the range of 4.9 MW with about 10 m!s plume velocity,
drops smaller than 1 mm will have a great amount of difficulty in penetrating the fire plume by
gravity alone. Drops larger than 2 mm are needed in this situation. However, there is a limit
as to how large the drops can be, because drops larger than 3 to 4 mm in diameter become
unstable and break up easily. In the gravity mode, the penetration ratio (Pe) is proportional to
drop size (dm). Based on dimensional analysis of pertinent variables, the penetration ratio for
sprinklers of similar geometry can be expressed"7' in terms of drop size factor:

where D is the orifice diameter of the sprinkler and Q, is the sprinkler flow rate. In the
gravity mode, an increase in orlfice diameter will increase the drop size and an increase in
pressure will decrease the drop size slightly.

In the momentum mode, the penetration ratio is a function of the Momentum Parameter (M)
which is the ratio of the downward momentum of the water spray and the upward momentum
of the fire plume. The momentum parameter is important only when it exceeds a critical
value, beyond which the sprinkler spray creates a significant effect on the plume, and drop size
is no longer important.
The penetration performance of geometrically similar prototype sprinklers of different
scales'"' was measured at different pressures with four sprinklers spaced 3 m apart and
centered at 3.5 m above a 2.28 MW simulated f i e plume generated from the FMRC ADD
Apparatus No. I ~ .Lumping all the values characterizing the simulated fire plumes and water
properties into a constant, the momentum parameter for this series of experiments was
expressed as:

Figure 5 shows the measured penetration ratio as a function of the drop size factor as defined
in Eq. 4. Penetration curves for both scale 1 and scale 1.5 sprinklers turn sharply upward at a
point representing the critical condition (drop size factor) for transition from the gravity
regime to the momentum regime. For the scale 1 sprinkler, the averaged threshold value on
the abscissa is about 0.007 s2" and occurs at a water pressure of 3.1 bar. From Eq. 5, the
associated value which can be considered as a critical momentum parameter (M,) for
transition from the gravity to the momentum regime, is determined as: M, = 0.204. This
value is found to be the same for all geometrically similar sprinklers.

In order to operate well in the momentum regime, the momentum parameter (M) has to be at
least 1.5 times as high as the critical value (M, ). From Eq. 5, the critical pressure for the
initiation of momentum regime, is influenced strongly by the K factor, sprinkler spacing, and
convective heat release rate. Methods to achieve reduction in the critical pressure to enter
more easily into the momentum regime include: 1) use of fast response sprinklers to reduce
the convective heat release rate at sprinkler operation, 2) increase in the orifice size, 3)
reduction in the spacing, and 4) reduced clearance. Furthermore, the critical value M, can
vary by a factor of 4, depending on the sprinkler design and location of the sprinkler in
relation to the ignition location.

Large-scale fire test and ADD test results('*' show that ADD is very sensitive to ignition
location and ceiling clearance. High clearance is most critical when ignition is located directly
under a sprinkler. Low clearance becomes critical when ignition is centered below two or
four sprinklers. Typical ADD performance of three different types of sprinklers, operating at
3.4 bar, with one sprinkler located 3 m dxectly above the FMRC ADD Apparatus No. I I ~are ,
compared in Figure 6. The upright 16 mm Large Drop sprinkler is operated in the gravity
regime: the higher the heat release rate and the lower the ADD value. The pendent 18 mm
ESFR-A sprinkler is clearly operated in the momentum regime, the ADD value does not
change significantly with f i e size, approaching the unity penetration ratio. For the pendent 16
mm prototype sprinkler, the ADD value decreases sharply beyond a critical heat release rate.

FRlRC ADD Apparatus No. I consists of a 1.3 m diameter water collection pan located below a non-reactive
fire plume generated from six commercial premixed gas burners.

FMRC ADD Apparatus NO. I1 consists of four sunulated storage commodities. In each of the simulated
commodities, there are four water collection systems. The fire plume simulator is a heptane spray fire placed
below the bottom of the simulated commodities. Four additional water collection pans are located at the floor
level to collect water falling through the FMRC flue spaces between the simulated commodities.
 x
-60

4
- 0 15--50
0

-- T=ZIT

d,EquiValent Dmp Diameter, mm

Figure 5. Penehatxon Ratio for FMRC Geame~cdlySilar Spnnlden,
Figvn 4. Terminai Velociw of water dmps in air at variou temp-. Scale 1 (D = 13 mml Scale 1 10 (D= 16 mm)

ESFR
.-
C ,
l a m

Pendent 16- Figure 6. ADD performance of 16 mm

Upright (LDS), 16 mm p e n d e ~ and
~ t 18 mm
ESFR Sprinklers. operated at 3.4 bar, one
Upright Large Drop 16- sprinkler located 3 m directly above the fire.
0 1000 2000 3000
Convective Heat Release Rate (W

Figure 7 . Air flow in the interaction between

water spray and the fire: the entrained flow
i%dominated neither by the plume nor the
spray.
This is a strong demonstration of the effect of critical condition for transition from momentum
regime on the left to gravity regime on the right.

The table below shows the sprmkler discharge characteristics of two different models of ESFR
sprinklers, ESFR-A and ESFR-B, together with a Large Drop sprinkler and a 13 mm standard
orifice size Spray sprinkler.

Water Pressure 3.4 bar Water Pressure 5.1 bar

Sprinkler dm MI(, LAD dm M: LAD
Tv~e Imm) ( ~ l m)' (mmlmin) (mm) ( ~ l m ' ) (mmlmin)
ESFR-A 0.72 48.6 25.8 0.63 61.1 36*
ESFR-B 0.67 97.1 33.3 0.59 145.7 42"
LDS 1.43 16
Standard 0.73 < 4.3 5

The center-core momentum flux of the spray, M i , was determined experimentally from the
total thrust force impinging on a 0.093 m2 circular load cell disk which was centered directly
under a sprinkler, at 2.13 m below the ceiling. The LAD is water density averaged over an
area of 5.23 m2, simulating the top cross-sectional area of a 2 x 2 storage array at 3 m below
the ceiling; those marked with * were measured at 4.42 m below.

The ESFR-A sprinkler, measured with higher ADD value , near 100% penetration, in Figure
6, has a smaller drop size, dm,than the Large-Drop sprinkler. Comparison of the center-core
momentum flux of sprinkler spray with that for fire plume (Figure 2), at the time of ESFR
sprinkler operation: M,/M, = 48.6 1 21 = 2.3, shows that the center-core momentum flux of
the sprinkler has to be at least twice as large as the f i e plume in order to approach 100%
penetration. Although both ESFR-A and ESFR-B were approved for protection of 9.1 m
high warehouses, ESFR-B, with smaller drop size and higher center-core momentum flux than
ESFR-A, was proven to be more effective in suppressing plastic fires in a 12.2 m high
building"9).

Computer Modelm of Sprinkler Spray and Fire Plume Interaction: Numerical simulation of
the interaction between a sprinkler spray and a f i e plume was first attempted by ~ l ~ e r t ( ~ O
A) .
SPRAY Code was developed to simulate two-dimensional Cartesian or Axisyrnmetric flows
induced by sprinkler spray and fire. In comparing the predicted ADD value using the SPRAY
Code with the measured ADD value, ill'^') found that while the model successfully captures
the trend of the experimental results, it has a few drawbacks and becomes unstable in many
instances.

In a recent attempt with a significantly modified PHOENICS code, am(^^' simulated the
interaction by combining two separate numerical models: a sprinkler spray without fire and a
fire plume without spray. Using the measured drop size, density distribution and thrust force
distribution as input data, and an accurately modeled free bum fire plume to predict ADD
values of ESFR-A and ESFR-B sprinklers, the numerically predicted ADD values are in good
agreement with the measurements at low water flow rate (low pressure). The predicted
values become much lower than the measurements at high water pressure when the sprinkler
momentum has significant impact on the f i e momentum. The calculated ADD values from
low to high sprinkler discharge rates had to be adjusted to match the measurements by gradual
reduction of the heat release rate from 100% to 40% of the free-bum-fire sizes. The
experiments were performed with the FMRC ADD Apparatus No. III', in which the heptane
spray is completely exposed to sprinkler water spray. It is possible that the actual sizes of the
spray fire, at high water flow rates, were reduced, due to the water drops washing away the
liquid heptane spray droplets before they were vaporized. The ADD value, as defined, should
not include the effects of suppression or extinction of a heptane spray fire plume, because the
pyrolysis and most of the combustion process were completed within the storage fuel array.

Figure 7 shows the numerically simulated air flow movement in the transition from the gravity
mode to the momentum mode of penetration, where the entrained flow is dominated neither
by the fire plume nor the spray.

Fire Suppression Criteria

Protection requirements are usually stated in the standards and codes as densitylarea of
demand or total water volume needed over a time period. Actually only a few sprinklers
discharge water toward the burning area, and a fraction of these can actually deliver water to
the burning surfaces for fire suppression. In a typical three-dimensional rack storage fire, at
the time of the first few sprinkler responses, the fire is still spreading vertically from the
burning commodities stored in the bottom two or three tiers. Flames appearing over the
storage array result from the burning of the pyrolyzed volatiles. The customary view is that
cooling of the flame near the ceiling by the spray droplets will have an insignificant effect on
the fire spread in the portion of the storage array which is shielded from the spray discharge.
Under these circumstances, the primary mode of sprinkler fire suppression is the c o o h g of
the burning and pyrolyzing surfaces by the flowing and trickling water delivered onto the top
surfaces of the stored array. In fact, there is a limit as to how far the trickling water can reach
the burning products if they are deep-seated inside the cardboard cartons. In the clean up
stage of a large-scale fire test, even hose streams are not expected to extinguish the residue
fires without the help of physical removal of burning commodities from the storage racks.
This is the reason that we emphasize fire suppression instead of fire extinguishment or
extinction in sprinkler technology research.

Fire suppression tests were conducted under the FMRC Fire Products Collector using various
standardized test commodities in a realistic rack storage arrangement to gain insight on fire
suppression. There are also two categories of standardized tests to class@ or evaluate the
fire hazards of real world storage commodities: 1) commodity evaluation for fire control and
2) Required Delivered Density (RDD) tests for fire suppression. In these tests, water is
applied uniformly on top of the storage array when the fire grows to a pre-selected size.

Critical Water Application Densitv for Fire Suppression: The status of fire extinction
~ ~ ) as bash'^^'. Most of
research with water up to 1986 has been reviewed by ~ e s k e s t a d ' and
these fire extinction data were converted to a critical water application density, m,", grams of
water per second per unit of entire surface area (gls m2), below which the fuel will be
completely consumed; above the critical density, the time to extinguishment decreased rapidly.

ADD Apparatus No. 111 consists of twenty water collection pans which are assembled to simulate the top
surface of a 2 x2 rack storage array, including the flue spaces between the simulated commodities. The fue
plume simulator is a heptane spray fue placed on top of the water collection system.
For wood slabs, cribs and pallet fires with 5-60% preburn, m,/ values vary from 1.5 to 3 gls
m2, and those for wood cribs and pallet fires are higher than wood slabs. For a fire initiated at
one end of a wood crib, water application density necessary to l h t lateral fire spread to the
entire crib (for fire control) were considerably smaller (0.7-1.1 gls m2 ). m," values for
plastic slabs range from 1.2 g/s m2 for horizontal PMMA to 4.4 gls m2 for Polyethylene. m,/
values were considerably higher with external radiation.

About one hundred forty fire suppression tests on FMRC standardized Class I1 and Plastic test
commodities were conducted using 2x2 rack storage arrays ranging from two to five tiers
h~gh'~*'.A fire suppression was developed to analyze the test results. This model
employs a global energy balance between the heat flux received by the fuel surfaces from the
flame and the heat absorbed by water due to evaporation to determine the total rate of
pyrolysis of the commodities stored in the test rack, during a time period after water
application. Based on this model, a fire suppression parameter, k, was identified to correlate
the fire suppression results, expressed as a ratio of AQ, / AQ,, AQ, is the cumulative total
heat release in a time period starting from water application to. AQ,, is a reference cumulative
heat release in a time period assuming the fire stopped growing at the start of the water
application.

where Q, is the heat of evaporation of water. T, is average temperature of the layer of file1
undergoing pyrolysis and T_ is the ambient temperature. The fire suppression parameter k,
which increases linearly with water application density (m,"), is a function of fuel
flammability, expressed in parameters such as density (pf ), specific heat (c), heat of
combustion (A&), burning rate per unit burning surface area (m, "), heat of pyrolysis (Q,),
fraction of total heat release rate (flame heat flux) transferred to the fuel surface (0) and fuel
geometric factor (a).

Figure 8 represents the ratio, AQ, 1 AQ,,, versus k for both the class I1 and the plastic
commodities stored in racks of different heights. The ratio, AQ, / AQ,,, measured during a
period after the water application of t - to = 240 s, correlated well with the fire suppression
parameter, k, as defied in Eq. 7. By setting AQ, / AQ,, = 1 in Eq. 6, the critical water
density per unit ex~osedsurface area for water a ~ ~ l i c a t i o(the
n entire surface area minus the
bottom area of the commodity) for the Standard Class I1 commodity was determined as 6 gls
m2 ; and that for plastic commodity was found to be in the range of 17 g/s m2 to 20 g/s m2.
The critical water density for Standard Class I1 commodity can be converted into m,," = 5 gls
per unit of entire surface area, in comparison to 1.5 to 3 gls mZ for wood slabs, cribs, and
pallets. It appears that combustible type, geometry, packaging and storage arrangement have
significant impact on the magnitude of the critical water density. These factors, especially
 + - theory
2.5 - O 2-tier.CLASS (I
0 3-tier.CLASS II
A 4-tier.CLASS II
2 - 0 2-tier.PIASTIC
? * 3-tier,PLASTIC
a + 4-tier,PLASTIC

2
4
1-

.5 -

Clearance to Sprinkler (m)

50 ; 4 3 2 1 0
I

....................................... ..........
......
........ ADD

.......

ADD Wre m e &la4.4 rprtnklen)

...............ADD W n m r r e &la2, I .prlnkkml

Storage Height (m)

Figure LO. A nons plot of me& ADD and RDD valuw as a function of slarage
height, clearance in a 9 m buildlng for diffetent ipinon Locations
with their influence on the flame heat flux to the exposed fuel surfaces or the J3 value inside
the storage array, have to be clarified and quantified before the fire suppression results of
small-scale tests can be used to assess the fire suppression results of the corresponding large-
scale tests.

There have been numerous studies of flame heat fluxes on the single and parallel burning
walls. In the last two years, ~ n g a s o n ( ~ has
~ ) provided measurements of the heat flux
distribution at the surface of an idealized fire spread in the 2x2 storage array, i.e., four equally
separated rectangular towers, exposed to flames from a circular gas burner at the floor. It
was found that the flame height data and excess gas temperature data within the storage array,
appear to follow axisymmetric fire plume laws. Comparison of the flame height data with
that measured from large scale rack storage of Standard Class I1 commodity tests shows good
agreement. Formulas for the flame heat flux distribution in terms of the total fire heat release
rate, fuel sootiness and separation distance are provided. It is expected these data can be used
to predict fire growth in storage geometries and employed in the fire suppression model as
defined in Eqs. 6 and 7 above.

Comrnoditv Evaluation for Fire Control: Sprmkler protection requirements of storage

occupancies for fire control systems were developed from large-scale fire tests using different
standardized test commodities. A less costly intermediate-scale fire test protocol was
developed to classlfy fire protection (control) challenges of different storage c~mmodities'~~).
The test apparatus consists of a special water applicator that is capable of spreading a fairly
uniform distribution of water directly over the top surface of the burning fuel array, thus
simulating water that has penetrated the fire plume (ADD). This apparatus is positioned
below the FMRC Fire Products Collector. The water applicator is actuated at a specific heat
release rate (as measured by the FPC), whlch corresponds to that of a sprinkler actuating in a
building with 3 m clearance.

This test procedure is to classlfy the hazard of a commodity stored on a 2 x 2 ~ 2tiers high rack
segment based on the delivered density that merely controls the fire. With this method, the
potential fire protection challenge of an unknown storage or test commodity can be measured
in terms of heat release rates (total, convective and radiative) under different water application
rates. Heat release rate curves, measured for an unknown storage commodity, are then
compared with those measured for the standard test commodities, especially under the water
application densities where the fires are not suppressed (Figure 9). This quantitative test
method is being used to classlfy an unknown storage commodity into proper fire protection
category.

Reauired Delivered Density (RDD) for Fire Suppression: Different from the commodity
evaluation test protocol for fire control, the RDD test protocol was developed to determine
the water density which must be delivered to the top of a burning storage array in order to
achieve a pre-determined degree of f r e suppression. The test procedure is similar to the
commodity evaluation tests except that RDD values are determined for different storage
heights in a 9.1 m high building. accepting only the results obtained from a successfully
suppressed test based on a criterion of no more than 10% fuel consumption of the test array
during the entire fire suppression period.
For the ESFR sprinkler located 0.33 m below a 9.1 m high cediug, the fire sizes at sprinkler
actuation for fires of different storage heights are all within the range of 1.5 and 2.64 MW
convective heat release rates. The measured RDD values for a given storage height do not
change significantly within this range of heat release rates. Consequently, a single RDD value
measured at 2.64 MW was selected for each of the three-, four- and five-tier high rack-
storage of plastic commodity in a 9.1 m high building as 16.3, 20.4 and 26.5 rnm/min"8'. It
was found recently'27' that the above RDD values corresponded to a AQ,/ AQ,, ratio of 0.67
in Figure 8. Based on this ratio, the fire suppression parameter, k, is determined to be
0.00357. Using the same ratio, the RDD values for Standard Class I1 commodity were about
6.9, 10.5 and 13.4 mmlmin for 2-, 3- and 4- tier fuel arrays.

APPLICATION OF SPRINKLER TECHNOLOGY

Knowledge gained from sprinkler technology research has been used to develop the following
three new performance oriented sprinklers: 1) Large Drop sprinkler - designed to provide
better fire control by producing a higher proportion of large drops to penetrate the fire plume,
2) Residential sprinkler - customized to maintain a survivable environment in residential areas,
and 3) ESFR sprmkler - optimized to overpower the fire plume and deliver enough water at
the early stage of development to suppress the fire before it becomes a severe challenge.

Development of Large Drop Sprinkler (LDS) System

The Large-Drop sprinkler was developed during the period from 1970 to 1980. It was
designed to provide: 1) higher water discharge using a maximum orifice size of 16 mm
(0.64 in), the highest water discharge possible in a retrofit to a 112 in pipe thread; and 2)
significantly large drops in comparison to large orifice Spray sprinkler (Figure 3). Against a
45 literlmin heptane spray fire plume, four Large Drop sprmklers centered above the fire,
deliver three times as much water to the seat of the fire as four large orifice Spray sprinklers
under the same discharge pressure'29).

Results are tabulated below from a series of four large-scale fire tests, using Large Drop
sprinklers installed at different sets of spacing and operating pressure combinations to provide
the same designed density (DD) of 24.4 mmlmin, to protect a 6 m high palletized storage of
plastic commodity in a 9.1 m high building which could not be protected with Spray
sprinklers, at the ceiling only, under any operating conditions :

Sprinkler Type LDS LDS LDS LDS

Orifice Size, in 0.64 0.64 0.64 0.64
mm 16 16 16 16
Density, mmlmin 24.4 24.4 24.4 24.4
Spacing, m2 12 9.3 7.4 3.1
Pressure, bar 3.3 1.9 1.2 0.5
No. Open 10 18 33 152
Total Flow, literlmin 2530 3710 5150 13600
Area of Demand, m2 120 167 24.4 465
This comparison shows that at the same density, the area of water demand increased from 120
to 465 m2 as the spacing and pressure were decreased from 12 m2 at 3.3 bar to 3.1m2 at 0.5
bar. It is clear that the area-density rule is not valid. Therefore, the Large-Drop sprinkler
standards are presented in terms of number of sprmklers at the minimum specified pressure
(no less than 1.7 bar) for the occupancy. The spacing of Large Drop sprinklers is limited to
a minimum of 7.4 m2 and a maximum of 9.3 m2. Against the 6 m high palletized storage of
plastics, Large Drop sprinkler operation at 1.9 bar resulted in a 50% reduction in water
demand in comparison with the standard sprinklers as shown in the last column on the right.
Furthermore, the Large Drop sprinkler, operating at 3.3 bar and 9.3 m2 spacing, has extended
the upper boundary of sprmkler effectiveness to 6 m high rack storage of plastic commodity in
a 9.1 m building. Currently, there are two FMRC Approved large drop sprinklers on the
market.

Development of the Residential Sprinkler System

From 1976 to 1982, research results demonstrated the need for fast-response (low RTI)
sprinklers with different discharge patterns for providing adequate life-safety protection in
residential fires. Based on the Information gained (30,31),fast-response (low RTI) Residential
sprinklers have been developed by the U. S. SpnnMer Industry. Since 1983, work in this area
has been expanded to include: 1) fast-response sidewall sprinklers for residential and hotel
applications, 2) effects of cathedral and beamed ceiling construction on residential sprinkler
performance, and 3) application of fast-response sprinklers in office occupancies.

Residential sprmklers were developed before it was fully established that ADD and RDD are
key factors in determining early fire suppression performance. Suppression capabilities of
Residential sprinklers and fast (quick)-response Spray sprinklers under a residential fire
scenario have been investigated using the procedures outhed for the ESFR development
program in a series of seven full-scale fire tests(32333).Suppression or no-suppression
predictions based on ADDJRDD relationships were verified in five tests. Suppression
occurred in two tests even with ADD<RDD. The ADDRDD approach provided a
conservative means for predicting fire suppression in this test program.

Development of ESFR Sprinkler System

ESFR stands for Early Suppression Fast Response. ES was placed intentionally before FR to
emphasize the end rather than the means. The ESFR sprinkler system was developed
originally to provide early suppression performance on high challenge storage fires
represented by up to 7.6 m high rack storage of standard plastic commodity in a 9.1 m high
building. The basic ESFR concept is that if a required quantity of water can be delivered to a
fire in the early stages of its development, suppression will be quickly achieved; thus, fewer
sprinklers will operate, and less fire and water damage will result(".

In principle, early suppression performance is determined by three fact~rs(~~~*),which

can be
independently measured: 1) Response Time Index (RTI), 2) Actual Delivered Density
(ADD), and 3) Required Delivered Density (RDD). The RTI and the time-dependent nature
of ADD and RDD determine the early suppression performance of the ESFR sprinkler system.
The premise of the ESFR system is to ensure ADD in excess of RDD, regardless of ignition
location.

As shown in Figure 5 , the most effective way of delivering a large quantity of water to the
base of the fire plume, is to ensure that the sprmklers are always operating in the far left of the
momentum regime. Based on Eq. 5, two methods can be used to achieve this at a fast
response heat sensing element, and 2) increasing the K factor of the s p r d e r . In addition,
the critical value for transition, M,, can also be reduced significantly through an innovative
sprinkler design to provide a sprinkler spray with strong momentum to overpower the upward
momentum of the fire plume. This is the guiding principle behind the ESFR Sprinkler
development.

Figure 10 is a cross-plot of measured ADD and RDD values as a function of storage height
and clearance in a 9.1 m high building for different ignition locations. It shows that for only
one sprinkler centered over the fire source, high clearance (low storage height) is critical; with
two and four sprinklers, low clearance (high storage) is critical. There were three years of
intensive ADD testing to develop the first ESFR prototype and a series of large-scale double-
and multiple-row rack storage fire tests, during the period of 1985 - 1986. Results of these
fire tests 'I8' not only confirm that the prototype ESFR sprinkler would perform well in the
critical conditions, they also demonstrate that even under the worst failure mode condition
tested, ignition between two sprinklers with one of them plugged, the fire was still controlled
with 11 sprinklers.

ESFR sprinklers have been shown to be effective for most common materials up to and
including unexpanded plastics, in cardboard cartons or exposed, in single-row, double-row
and multiple-row rack storage or in solid-piled, palletized or open-frame storage up to 10.6 m
high in a 12.2 m high warehouse. The list of commodities protectable by ESFR sprinklers has
been extended to include expanded plastic, nonwoven and aerosol products. Also, lesser
hazard commodities, lower storage heights, or ceiling heights lower than 9.1 and 12.2 m are
protected with the same water demand: 12 sprinklers at a minimum water pressure of 3.4 bar
and 5.1 bar for 9.1 to 12.2 m high warehouses, respectively.

Currently there are four FMRC Approved ESFR sprinklers on the market. They are of
pendent design with an orifice diameter of 18 mm (0.7 in.). The sprinkler link used is rated
74" C with a RTI of 22-36 mln sl".

FUTURE TECHNOLOGICAL TRENDS

The successful applications of sprinkler technology in residential and hlgh challenge

warehouse (up to 12.2 m high) fire protection represent only two segments of an entire
spectrum of fire challenge. There is a need to expand the application: to develop ESFR
sprinklers of smaller orifice sizes for medium challenge occupancies, such as manufacturing
and assembling plants, public and institutional buildings as well as ESFR sprinklers of larger
than 18 mm orifice size. Undoubtedly, a more tolerable design method than the initial ESFR
program with its "one size fits all" approach is needed. Design methodology of the next
generation of ESFR sprinklers will require technically sound predictive models. These models
will be used to predict whether a sprlnkler spray can deliver the RDD for the suppression of a
particular fire situation. The future performance-oriented sprlnkler standard will be backed up
with a computer simulation design guide. One ultimate goal is to develop a
phenomenological sprinkler performance model intended to simulate a high-challenge
sprinklered warehouse fire. The model will be expanded to provide a systematic method for
the mitigation of the consequences such as fire damage and nonthermal damage. as well as
potential damage from water run-offs.

REFERENCES

Thompson, N. J., "Fire Behavior and Sprinklers," National Fire Protection Association,
Quincy, MA, 1964.
Yao, C., Marsh, W. S., "Early Suppression - Fast Response: A Revolution in Sprinkler
Technology," Fire Journal, National Fire Protection Association, Quincy, MA, Jan.
1984, pp. 44-46.
NFPA 231 C: Standard for Rack Storage of Materials," National Fire Protection
Association, Quincy, MA USA, 1991 Edition.
Vincent, G. B. and Kung, H-C., "Comparison of European Conventional and U.S. Spray
Sprinklers," J. of Fire Protection Engineering, 5 (1) 1993, pp. 17-28.
Alpert, R.L., "Calculation of Response Time of Ceiling Mounted Fire Detectors,'' Fire
Technology, Vol. 8, No.3, August 1972.
Evans, D.D., and Stroup, D.W., "Methods to Calculate the Response of Heat and
Smoke Detectors Installed Below Large Unobstructed Ceilings," Fire Technology, Vol.
22, No. 1, 1985, pp. 54-65.
Heskestad, G., and Delichatsios, M.A., "The Initial Convective Flow in Fire," The
Seventeenth Symposium ( International ) on Combustion, 1978, pp. 1113-1123.
Yu, H-Z, 'Transient Plume Influence in Measurement of Convective Heat Release Rates
of Fast-Growing Fires Using a Large-Scale Fire Products Collector", Journal of Heat
Transfer, Vol. 112, 1990, pp. 186-191.
You, H. Z. and Kung, H. C., "Strong Buoyant Plumes of Growing Rack Storage Fires,"
Twentieth Symposium on Combustion, The Combustion Institute, 1984.
Yu, H-Z., and Stavrianidis, P., "Transient Ceiling Flows of Growing Rack Storage
Fires," Proceeding of the Third International Symposium on Fire Safety Science,
International Association for Fire Safety Science, pp. 281-290, 1991.
Yu, H-Z., "Sprinkler Response Prediction for Warehouse Rack Storage Fires,"
Publication in preparation, 1996.
Heskestad, G. and Smith, H. F., "Investigation of a New Sprinkler Sensitivity Approval
Test: The Plunge Test," FMRC 22485, Factory Mutual Research Corporation,
Norwood, Massachusetts, December 1976.
Heskestad, G. and Bill, R. G., "Quantification of Thermal Responsiveness of Automatic
Sprinklers - Including ConductionEffects", Fire Safety J., 14, 1988, pp. 113-125.
Yao, C. and Kalelkar, A. S., "Effect of Drop Size on Sprinkler Performance," Fire
Technology, Vol. 6, No. 4, November 1970, pp. 254-268.
Yu, H-Z., "Investigation of Spray Pattern of Selected Sprinkler with the FMRC Drop
Size Measuring System," Proceeding of the First International Symposium on Fire
Safety Science, pp. 1165-1176, Hemisphere, 1986.
Yao, C., 'Development of Large Drop Sprinklers" FMRC Technical Report No. 22476,
Factory Mutual Research Corporation, Norwood Mass., June 1976.
Heskestad, G., "Sprinkler Performance as Related to Size and Design, Vol. 1-
Laboratory Investigation,' FMRC Technical Report No. 22437, 1979.
Yao, C., 'The Development of the ESFR Sprinkler System", Fire Safety Journal, 14
(1988) pp. 65-73.
Kung, H-C., Vincent, B., Chan, T-K., Yu, H-Z., and Stavrianidis, S., 'Zarly
Suppression Fast Response (ESFR) Sprinkler Protection for 12 m High Warehouses" to
be published in the Proceeding of the Fifth International; Symposium on Fire Safety
Science, 1996.
Alpert, R. L., "Numerical Modeling of the Interaction Between Automatic Sprmkler
Sprays and Fire Plumes," Fire Safety J., 9, 1985, pp. 157-163.
Bfl, R. G., Jr. "Numerical Simulation of Actual Delivered Density Measurements," fire
Safety J., 20, 1993, pp. 227-240.
Nam, S., "Development of a Computational Model Simulating the Interaction Between a
Fire Plume and a Sprinkler Spray," Fire Safety Journal, 26, 1996.
Heskestad, G., "The Role of Water in Suppression of Fire: A Review", Journal of Fire &
Flammability, Vol. 11 (October 1980) p.254.
Rasbash, D. J., "The Extinction of Fire with Plain Water," The Proceedings of the First
International Symposium of Fire Safety Science, Hemisphere Publishing Corporation,
pp. 1145-1163, 1986.
Lee, J. L., "Extinguishment of Rack Storage Fires of Corrugated Cartons Using Water,"
The Proceedings of the First International Symposium of Fire Safety Science,
Hemisphere Publishing Corporation, 1986, pp. 1177.
Yu, H-Z., Lee, J. L., and Kung, H-K., "Suppression of Rack Storage Fires by Water,"
The Proceedings of the fourth International Symposium of Fire Safety Science,
International Association for Fire Safety Science, pp. 901-912, 1994.
Ingason, H., and De Ris, J., "Flame Heat Transfer in Storage Geometries," Publication
in Preparation, 1996.
FMRC Update "Advances in Commodity Classification", A progress report from
Factory Mutual Research Corporation, Vol. 4, No. 1, 1990.
Chicarello, P. J., "Large Drop Sprinkler," Technical Report, FMRC J. I OGOE7. RR,
Factory Mutual Research Corporation, Nonvood, MA, February 1982.
Kung, H. C., Spaulding, R. D. and Hill, E. E., "Sprinkler Performance in Residential Fire
Tests," Technical Report FMRC Serial No. 22574, Factory Mutual Research
Corporation, Nonvood, MA July 1980.
Kung, H. C., Spaulding, R. D., Hill, E. E., and Symonds, A. P., "Field Evaluation of
Residential Prototype Sprmklers, Los Angeles Fire Test Program," Technical Report,
FMRC J.I. OEORS.RA, February, 1982.
Bill, R. G., et al, "Predicting the Suppression Capability of Quick Response Sprinklers in
a Light Hazard Scenario - Part I: Fire Growth and Required Delivered Density (RDD)
Measurements," Journal of Fire Protection Engineering, Vol. 3, No. 3, September 1991.
Bill, R. G., et al, "Predicting the Suppression Capabihty of Quick Response Sprinklers in
a Light Hazard Scenario - Part 11: Actual Delivered Density (ADD) Measurements and
Full-Scale Fire Tests," Journal of Fire Protection Engineering, Vol. 3, No. 3, September
1991.
Yao, C., "Early Suppression Fast Response Sprinkler Systems" Chemical Engineering
Progress, September 1988, pp.38-43.