Available online at www.sciencedirect.

com
Proceedings
ScienceDirect of the
Combustion
Institute
Proceedings of the Combustion Institute 35 (2015) 2723–2729
www.elsevier.com/locate/proci

Characterization of interactions between hot air

plumes and water sprays for sprinkler protection
Xiangyang Zhou ⇑
FM Global, 1151 Boston-Providence Turnpike, Norwood, MA 02062, USA

Available online 22 October 2014

Abstract

One important aspect of the complex sprinkler protection process is the interaction between the water
spray and the fire plume. In order to provide suitable data for the development and validation of a LES-
based fire protection models, such as FireFOAM, a series of small-scale experiments were conducted to
examine the interaction of hot air plumes and water sprays through combined gas–liquid velocity and
droplet size measurements. Laser-based particle image velocimetry (PIV) was used to acquire the spa-
tially-resolved velocity data; and a shadow imaging system (SIS) was used to measure the water droplet
size and volume flux. Hot air plumes with three convective heat release rates (1.6, 2.1 and 2.6 kW) were
selected to interact with a water spray at a discharge rate of 0.084 Lpm. The velocity field of the hot air
plume and ceiling flow with/without water spray, the droplet size and volume flux of water spray with/
without hot air plume were measured. The interaction between the hot air plume and water spray was char-
acterized by the location of the interaction boundary with the momentum ratio of the hot air to that of the
spray. The results showed that that the momentum ratio and the evaporation effect due to hot air on the
water droplets played a significant role to change the interaction structure and the ceiling layer pattern.
Ó 2014 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Fire plume; Sprinkler spray; Interaction; Flow velocity; PIV

1. Introduction evaporation) and heat transfer (due to droplet

Water sprays have been one of the most reli-
able and effective methods to control or suppress
a fire. One important aspect of the complex fire
protection process is the interaction which occurs
between the droplet spray and the turbulent buoy-
ant gas flow induced by a fire. When a downward
water spray is injected above a fire source,
the droplet undergoes momentum transfer (due
to drag force), mass reduction (due to water
⇑ Fax: +1 (781) 255 4024.
E-mail address: xiangyang.zhou@fmglobal.com
heat-up and evaporation). The trajectories of fall-
ing droplets are modified by the upward fire
plume so that only a part of the spray can pene-
trate through the plume to the top of the burning
fuel, and a part of the spray is deflected back
toward the ceiling, finally falling to the floor well
away from the fire source. In addition to the mod-
ification of the upward fire plume, there is also an
impact on ceiling-jet velocities and temperatures
induced by the droplet spray. The outcome of this
interaction determines how the droplet size and
momentum of the spray, the momenta of the
flames and hot plume, and the evaporation of
spray in the flames influence the cooling of the fire

http://dx.doi.org/10.1016/j.proci.2014.05.078
1540-7489/Ó 2014 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
2724 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
plume, and the degree of water penetration
through the fire plume to the burning fuel
surface.
The fire research community has long been
interested in the experimental measurement and
numerical modeling of the interaction between a
fire plume and a water spray from a sprinkler.
The actual delivered density (ADD) test appara-
tus was developed at FM Global to characterize
and evaluate the capability of sprinkler sprays to
penetrate the fire plume of a rack storage com-
modity arrangement by collecting the water in
pans placed beneath a fire [1]. When the ADD is
greater than the RDD (required delivered
density), fire suppression is expected to occur.
The ADD apparatus has provided much useful
data; however, little experimental data are avail-
able to detail how the characteristics of a fire
plume and a droplet spray change during the
interaction. Most of the knowledge of this interac-
tion was gained through computer modeling [2–6].
A new computational code, FireFOAM, based on
the OpenFOAM framework has been developed
at FM Global to further the numerical modeling
of water-based fire protection including turbulent
reactive flow [7], solid fuel pyrolysis [8], soot
radiation [9], water transport [10] and sprinkler
sprays [11].
This work was purposed to extend existing
studies of the interaction between the hot air
plumes and water sprays through measurements
of gas–liquid two-phase velocity fields and droplet
size variation in a series of small-scale experi-
ments. A diagnostic technique of laser-based par-
ticle image velocimetry (PIV) was used for
acquisition of spatially-resolved velocity data in
the gas–liquid interaction field. This measurement
was motivated by the need to develop and validate
numerical simulation tools for such flows. New
numerical techniques, such as Large Eddy Simula-
tion (LES), have combined spatial and temporal
resolution that cannot be fully validated by tradi-
tional point-measurement techniques, in which
data are not spatially correlated. The variation
of water droplet size and volume flux in the inter-
action field was measured by a shadow imaging
system (SIS). The goals of this work were to
examine the characteristics of turbulent hot air
plume interacting with the water sprays through
combined gas–liquid velocity and droplet size
measurements, and to use the data for the devel-
opment and validation of LES-based fire protec-
tion model. The hot air plume experiments were
selected because the generation of turbulence
due to buoyancy and water spray can be studied
but the complexities of combustion chemistry
are removed. The region of interest in this study
was the velocity field of hot air plume and ceiling
flow with/without spray, and droplet size and
volume flux of water spray with/without hot air
plume.
2. Experimental setup
A small-scale plume-spray test facility was set
in a laboratory with dimensions of 8  9 m and
4 m in height. Figure 1 shows the schematic of
the experimental setup, which mainly comprises
an upward air flow, a downward water spray
and a horizontal ceiling. The four measurement
windows (or field of view) illustrated in the figure
were used for the PIV measurements. The ADD
test apparatus [1,4] was referenced to design the
current small-scale test facility.
The hot air source was designed to be a circular
nozzle with 72 mm in diameter. The origin of a
two-dimensional (2D) cylindrical coordinate sys-
tem (r–z) was located at the center of the top of
the nozzle, where r is the radius from the centerline
and z is the height above the nozzle. Hot air was
forced into the nozzle through a Leister electric
air heater and a tube. The tube was filled with
metal screens to make the exit velocity profile uni-
form. Several fire sizes were simulated by changing
the convective heat release rates (HRRs) of the hot
air. The convective HRR was calculated from the
air flow rate and the temperature rise above ambi-
ent. The vertical exit velocity was up to 5.0 m/s,
and the exit air temperature was up to 205 °C.
The hot air temperatures were measured using
seven type K thermocouples (28 Gage).
A Delavan CT-1.5–30 B full cone nozzle was
selected to discharge the water spray. This spray
nozzle is rated to flow 0.08 Lpm of water at
690 kPa with a 30° initial spray angle. Figure 1
shows that the spray nozzle was installed at
z = 560 mm above the hot air nozzle and 30 mm
below the center of an aluminum ceiling plate.
The ceiling plate size was 1.22  1.22 m with
3 mm thickness. Water flow to the nozzle was
accomplished by pressurizing a water storage tank
with dry grade nitrogen. The nitrogen supply pres-
sure was regulated with a two-stage regulator and
the tank pressure was monitored with a pressure
Fig. 1. Schematics of experimental setup: upward air
flow, downward water spray, horizontal ceiling and four
measurement windows (A, B, C and D) used for PIV
measurements.
 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
gauge. The water tank had a 151 L capacity and
was connected at its bottom by a high-pressure
flexible tube to the spray nozzle. The water flow
rate was monitored by a flow meter.
In PIV measurements, flow velocity is deter-
mined from measurement of displacement of
imaged droplets in a thin light sheet and the time
for displacement. In this work, the water droplets
in the spray were directly used for imaging and
velocity calculation. The hot air and the surround-
ing air were seeded with oil droplets (soybean oil,
0.25-1 microns droplet size), which were produced
from a LaVision Aerosol generator. The droplets
were illuminated by two light sheets expanded
from a dual-cavity-pulsed Nd:YAG laser with an
optimized time difference. The maximum energy
in each pulse was 120 mJ and the wavelength was
532 nm. A 14-bit dual-frame CCD camera (LaVi-
sion Imager ProX4M) with a 2048  2048 pixel
resolution was used to image droplets. The light
sheet was aligned vertically to the z-axis at the
spray center. The camera was aligned at 90° to
the light sheet and focused on the sheet/droplets.
The SIS was used to measure water spray
droplet size, number density, and velocity [12]. It
is based on the shadowgraph technique with
high-resolution imaging and pulsed backlight illu-
mination. After calibrating the detection system,
the current camera’s field of view was
4.41  4.41 mm. The number of pixels in the cam-
era’s field of view was 2048  2048. The minimum
droplet size that could be detected was 0.012 mm.
LaVision’s DaVis 7.2 software was used to pro-
cess the PIV and SIS data.
To provide a plume-dominated flow and a
spray-dominated flow, three convective HRRs
(1.6, 2.1 and 2.6 kW) were generated by supplying
three air flow rates and maintaining a constant
exit air temperature. The water spray nozzle was
operated the pressure of 750 kPa and the mea-
sured water flow rate was 0.084 Lpm. Four mea-
surement area windows were selected for PIV
measurements and each window had the same size
of 257  257 mm. As shown in Fig. 1, two
windows A (z = 0–257 mm) and B (z = 253–
510 mm) were placed in the spray centerline, and
two windows C (r = 56-313 mm) and D
(r = 308–565 mm) were placed at the right side
of the spray nozzle and under the ceiling. A low
measuring frequency of 0.9 double exposures per
second was selected so that more images could
be stored directly in the computer hard disk.
3. Results and discussion
3.1. Initial velocity and temperature of hot air
plume without spray
Figure 2 shows measurements of the radial
variation of the average vertical velocity
2725
component (V) and three Reynolds stresses (uv,
uu and vv) 30 mm above the hot air nozzle for a
low convective HRR of 1.6 kW. The data were
averaged from a vector set with 200 pairs of PIV
images or vector fields. Figure 2 shows that a near
top-hat velocity profile existed at the nozzle exit
and the maximum velocity was Vm = 3.3 m/s.
The turbulent Reynolds stresses were low in the
centerline and then increased to peak values at
the nozzle edges (r = ±36 mm). The other two
hot air plumes had the same velocity profiles but
a higher value of Vm = 4.2 m/s for 2.1 kW and
Vm = 5.1 m/s for 2.6 kW.
Figure 3 shows horizontal profiles of the air
temperatures 30 mm above the hot air nozzle mea-
sured for three hot air plumes. The data were
averaged over 3 min with a 1 Hz measuring fre-
quency and the standard deviation was about
0.5 °C at the nozzle center. The peak temperature
was Tm = 205 °C and the room temperature was
T0 = 21 °C. The peak value of Tm was controlled
to be constant for the three different convective
HRRs.
3.2. Initial droplet size and velocity of water spray
without hot plume
Figure 4 shows a scatter plot of the droplet ver-
tical velocity versus its size for a large number
(>4000) of droplets detected in the spray center
and at two elevations: z = 530 mm (30 mm below
the spray nozzle) and z = 260 mm. In the near
field (z = 530 mm) of the spray nozzle, Fig. 4
shows a general trend with the velocity magnitude
increasing with droplet diameter from 1 m/s to
27 m/s. In the far field (z = 260 mm) below the
nozzle, however, the small droplets (0.012 mm)
are in the same velocity range from 1.5 m/s to
7 m/s as the larger droplets. This result means that
most water droplets in the far field are moving
Fig. 2. Horizontal profiles of averaged vertical velocity
(V) and three Reynolds stresses measured at 30 mm
above the hot air nozzle with convective HRR of
1.6 kW.
2726 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
Fig. 3. Horizontal profiles of averaged air temperatures
measured at 30 mm above the hot air nozzle for three
convective HRRs.
Fig. 4. Scatter plot of vertical droplet velocity versus
diameter in the near field (z = 530 mm) and the far field
(z = 260 mm) measured in the spray center.
following the air flow induced by spray entrain-
ment. Because current PIV cannot measure the
two-phase velocity field simultaneously in the
spray region, the air velocity in the far field was
then approximated by imaging the water droplets.
Figure 5 shows the horizontal profiles of the
volume-median diameter (dv50), averaged vertical
velocity, and vertical volume flux of water drop-
lets in the near field measured by the SIS. The vol-
ume-median diameter (dv50) denotes that 50% of
the cumulated water volume is represented by
droplets having a diameter smaller than dv50.
The volume flux (Lpm/m2) was calculated from
all detected droplets with droplet diameter, num-
ber density and velocity. Because droplet velocity
is related to droplet size, the average droplet
velocity was calculated by dividing the volume
flux by the spray volume density. Figure 5 shows
that the initial spray in the near field was very
dense, so that the spray center had a peak down-
ward volume flux of 170 Lpm/m2, a maximum
Fig. 5. Horizontal profiles of volume-median diameter
(dv50), average vertical velocity and vertical volume flux
of water droplets in the near field (z = 530 mm).
Fig. 6. Horizontal profiles of volume-median diameter
(dv50), average vertical velocity and volume flux of water
droplets measured by SIS, and a PIV velocity profile at
z = 260 mm.
downward velocity of 18 m/s and large size
droplets, dv50 = 0.056 mm. By integrating the vol-
ume flux and area, the total flow rate was calcu-
lated to be 0.075 Lpm, which is 9% lower than
the measured value of 0.084 Lpm. This verifica-
tion confirms that the SIS measurements were
reliable.
The initial spray pattern changed with distance
from the nozzle. In the far field, Fig. 6 shows that
the radial distribution area was larger, the down-
ward droplet velocities and volume fluxes were
lower, and larger droplets moved toward the outer
spray edge. As a comparison, the velocity profile
obtained from the PIV is also presented in
Fig. 6. Because the spray was not so dense and
droplet velocity was less size related, the figure
shows that the two results obtained from PIV
and SIS are in good agreement in the far field.
3.3. Interaction between hot air plume and water
spray
After characterizing the initial conditions of
hot air plume and water spray, measurements
 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
were conducted to investigate the interaction
between the plume and the spray. For three con-
vective HRRs and one spray flow rate, Fig. 7
shows the response of the velocity structure of
the hot plume to the spray. The vectors were aver-
aged from a set with 500 pairs of PIV images. The
measurements in the two windows A and B were
merged into one. The current PIV could distin-
guish two droplets (small oil droplets and larger
water droplets) by light intensity in the region
where there were no (or very little) water droplets;
however, this discrimination was not possible for
a dense spray. For similar densities of small and
large droplets, the PIV measurements would bias
toward the large droplets, which means that the
velocity measured was close to the velocity of
the larger droplets. Figure 7 shows that a stagna-
tion point was formed in the interaction boundary
where the upward air flow met the downward flow
of air entrained with the spray, and the interaction
boundary moved up with a higher HRR. At the
low HRR of 1.6 kW, the hot air column was
almost suppressed and the interaction boundary
was pushed closer to the base of the plume. As
the HRR increased to 2.1 kW, the upward flow
pushed the interaction boundary to a higher posi-
tion (z = 320 mm) and formed two large vortexes
Fig. 7. Averaged vector fields measured by PIV for the interaction between the hot air plume with three convective
HRRs and the water spray with one discharge rate.
2727
around the spray. At the higher HRR of 2.6 kW,
the interaction boundary was pushed closer to the
spray nozzle and the spray entrainment effect was
reduced. The position of the interaction boundary
was determined as the location characterized by a
zero vertical velocity at the centerline. Although
the average velocity was zero, the interaction
region itself was very turbulent. The peak value
of the turbulent Reynolds stresses (vv) at the inter-
action boundary (z = 320 mm) was found to be
20 times larger than that at the free hot air plume
(2.1 kW).
The other important effect of the hot plume is
that water droplet sizes and the volume flux varied
with the plume-spray interaction. Figure 8 shows
a comparison of the horizontal profiles of vol-
ume-median diameter and downward volume flux
obtained from the spray with/without hot air
plume (2.1 kW) at z = 320 mm. The figure shows
that the droplet size in the spray center increased
with the hot air plume, indicating that some small
droplets disappeared due to water evaporation.
Small droplets might also be pushed out from
the centerline so that the droplets became smaller
toward the outer spray edge. The SIS measure-
ments showed that the number of droplets and
the downward velocity were reduced in the
2728 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
Fig. 8. Horizontal profiles of volume-median diameter
(dv50) and vertical volume flux of water droplets com-
pared at z = 320 mm for spray with/without hot air
plume (2.1 kW).
interaction region. Therefore, Fig. 8 shows that
the volume flux reduced significantly with the
hot air plume, indicating that the degree of water
penetration through the hot plume to the hot
source was reduced.
The interaction between a sprinkler and fire
has often been discussed in terms of the competi-
tion between the momentum of the downward
spray and the momentum of the upward fire
plume [2,4]. In this work, the momentum (Ma)
of the hot air plume without spray was calculated
directly from the velocity and temperature mea-
surements near the nozzle exit. The initial momen-
tum (Mw = 0.043 N) of the water spray was
approximated from the nozzle discharge rate
and the injection velocity. Figure 9 shows the
position of the interaction boundary versus the
momentum ratio (Ma/Mw) for the three hot air
plumes. The figure shows that the location of
the interaction boundary increased from
z = 60 mm to z = 440 mm with the momentum
ratio going from 0.6 to 1.5. Because a ceiling
was installed, the height of the interaction bound-
ary with the 2.6 kW plume might have been
reduced due to the effect of the ceiling. To under-
stand the effect of hot air temperature on the
interaction, a cold air flow with Ma/Mw = 1.2 at
room temperature was also measured and the
interaction boundary location is presented in
Fig. 9. The figure shows that the cold air flow
had a lower interaction height of z = 180 mm,
although its momentum was stronger than that
of the hot air plume with 2.1 kW. The meaning
of this observation is that the interaction structure
was changed not only by the momentum ratio but
also by the evaporation effect due to hot air on the
water droplets.
3.4. Ceiling air flow
The ceiling flow is important to sprinkler acti-
vation. Figure 10 shows the averaged vector fields
Fig. 9. Position of the interaction boundary versus the
momentum ratio measured for three hot air plumes and
one cold air flow.
and horizontal velocity (U) contours of the ceiling
layer at the interaction between the hot air plume
(2.1 kW) and the water spray. The vectors were
averaged from a set with 200 pairs of PIV images.
The measurements in the two windows C and D
were merged into one. The spray nozzle was
located beyond the left bound of the figure at
r = 0 mm. It is clear that a horizontal ceiling layer
formed under the ceiling and moved out from
the centerline. The left side of Fig. 10 shows a
circular zone with negative horizontal velocities,
which means that air was entrained into the spray.
Figure 10 also shows a peak value of U = 0.43 m/s
in the ceiling layer appeared at r = 310 mm. For
the two other hot air plumes interacting with the
same spray (shown in the supplemental material),
the maximum U appeared at different locations:
r = 210 mm for 2.6 kW and r = 550 mm for
1.6 kW. This result indicates that a relative
stronger spray will push the ceiling layer toward
the outer edge.
Figure 11 shows a comparison of the vertical
profiles of averaged velocity (U) and turbulent
Reynolds stresses (uu) in the ceiling layer obtained
from the hot air plume (2.1 kW) with/without
water spray at r = 310 mm. The figure shows that
the peak value of U was reduced by 50%, and that
of uu was reduced by 72% under the spray
Fig. 10. Averaged vector fields and horizontal velocity
(U) contours of the ceiling layer for the interaction
between the hot air plume (2.1 kW) and the water spray.
 X. Zhou / Proceedings of the Combustion Institute 35 (2015) 2723–2729
Fig. 11. Vertical profiles of averaged horizontal velocity
(U) and Reynolds stress (uu) in the ceiling layer
compared at r = 310 mm for hot air plume (2.1 kW)
with/without spray.
interaction. In the lower region below the ceiling,
however, the values of U and uu were larger with
the interaction than those without the interaction.
If the ceiling layer edge is defined as the location
with U = 0.01Umax, the ceiling layer thickness
was 4 times increased from 60 mm for free hot
air (2.1 kW) to 250 mm for the hot air with spray,
meaning that the plume-spray interaction can
make the ceiling layer more diffuse and thicker.
4. Summary and conclusions
The characteristics of the interaction between
the hot air plume and the water spray have been
measured by using particle image velocimetry
(PIV) and a shadow imaging system (SIS) with
the objective to generate a data set for validation
of computational fire models. Hot air plumes with
three convective heat release rates (1.6, 2.1 and
2.6 kW) were selected to interact with a water
spray at a discharge rate of 0.084 Lpm. First the
initial conditions of velocity and temperature of
the hot air without spray, and droplet size and
volume flux of water spray without hot air were
measured. The interaction between the hot air
plume and water spray was then characterized
by the location of the interaction region. The
velocity vector fields showed that the height of
the interaction boundary above the base of the
plume increased with the heat release rates.
Finally the velocity fields of the ceiling layer
2729
with/without the plume-spray interaction were
measured. The results of the plume (2.1 kW)
showed that the peak velocity in the ceiling layer
was reduced by 50%, the layer thickness was
4 times increased and the location of the maxi-
mum velocity moved toward the outer edge due
to the interaction. It was shown that both, the
momentum ratio of the hot air to the spray and
the variation of droplet size and volume flux due
to water evaporation, played a significant role in
the plume–spray interaction.
Acknowledgements
This study was funded by FM Global in its
sprinkler technology research program. The
author is grateful to Hong-Zeng Yu and Stephen
D’Aniello for their assistance and discussions.
Appendix A. Supplementary material
Supplementary data associated with this article
can be found, in the online version, at http://
dx.doi.org/10.1016/j.proci.2014.05.078.
References
[1] H.Z. Yu, T.S. Chan, H.C. Kung, W.R. Brown, P.
Stavrianidis, Sprinkler performance evaluation sys-
tem, US Patent number 6085585, July 11, 2000.
[2] R.L. Alpert, Fire Safety J. 9 (1985) 157–163.
[3] R.G. Bill Jr., Fire Safety J. 20 (1993) 227–240.
[4] S. Nam, Fire Safety J. 32 (1999) 307–329.
[5] Y.F. Li, W.K. Chow, Fire Technol. 44 (2008) 351–
381.
[6] N. O’Grady, V. Novozhilov, Combust. Sci. Technol.
181 (2009) 984–1006.
[7] Y. Wang, P. Chatterjee, J.L. de Ris, Proc. Combust.
Inst. 33 (2011) 2473–2480.
[8] M. Chaos, M.M. Khan, S.B. Dorofeev, Proc.
Combust. Inst. 34 (2013) 2583–2590.
[9] P. Chatterjee, J.L. de Ris, Y. Wang, S.B. Dorofeev,
Proc. Combust. Inst. 33 (2011) 2665–2671.
[10] K.V. Meredith, J. de Vries, Y. Wang, Y. Xin, Proc.
Combust. Inst. 34 (2013) 2719–2726.
[11] Y. Wang, K. Meredith, X. Zhou, P. Chatterjee, Y.
Xin, M. Chaos, N. Ren, S. Dorofeev, Fire Safety
Science 11 (2014), (in press).
[12] X. Zhou, H.Z. Yu, Fire Safety J. 46 (2011) 140–150.