.03 Spk Experimental And Numerical Study Of Interactions Between Spk And Natural Smoke Vents [623034]

RESEARCH ARTICLE
Experimental and numerical study of interactions between
sprinklers and natural smoke vents
Nicolas Trévisan1,2,3|Pascal Boulet1,2
|Zoubir Acem1,2|Alexandre Jenft3|
Grégoire Pianet3|Arnaud Breton3|Armelle Muller3
1Université de Lorraine, LEMTA, UMR 7563,
Vandœuvre ‐lès‐Nancy F ‐54500, France
2CNRS, LEMTA, UMR 7563, Vand œuvre ‐lès‐
Nancy F ‐54500, France
3CNPP, Route de la Chapelle Réanville, BP
2265 F ‐27950 Saint Marcel, France
Correspondence
Pascal Boulet, Université de Lorraine, LEMTA,UMR 7563, Vand œuvre ‐lès‐Nancy F ‐54500,
France.
Email: pascal.boulet@univ ‐lorraine.frAbstract
Real scale experimentations have been conducted in order to investigate interactions that may
occur in a compartment where sprinklers and Smoke and Heat Exhaust Vent Systems coexist.
Fuel oil spray was used as fire source with steady heat release rate of 400 and 800 kW. Bothsprinkler system and natural Smoke and Heat Exhaust Vent Systems were designed according
to the French standards. Effect of vent on sprinkler activation time was studied. An analysis of
the temperature field inside the compartment was conducted while vent was opened prior orafter sprinkler activation. Simulations of the experiments were also carried out with FireDynamics Simulator v.6 for validation purpose and also to provide supplementary data regarding
soot flow rate and energy extracted at the vent.
KEYWORDS
FDS, real scale experiment, smoke vent, sprinkler
1|INTRODUCTION
Smoke venting goal is to exhaust heat and smoke from a building in the
event of a fire emergency. Its main purpose is to ensure safe egress ofthe occupants by slowing the loss of visibility and keeping clear ofsmoke a path of escape close to the floor. Other safety objectives
aim at delaying flash ‐over occurrence and assisting firefighters'
operations. Automatic fire sprinkler purpose is to detect and controlor extinguish fire in its early stage. While both of these systems are
well known, their use in conjunction is subject to controversy due to
a lack of scientific data.
The main problem is that smoke vents and sprinkler systems may
have opposite side effects. When sprinkler activation occurs in a fire
event, a possible smoke destratification is expected, resulting in a lossof visibility close to the floor and a mixing of smoke and droplets with
possible toxic and thermal consequences on people. This goes against
smoke and heat venting purpose which is to extract smoke andenhance visibility, allowing safe egress. On the other hand, byremoving heat from the facility before sprinkler activation, smoke
and heat vents could delay or prevent this activation.
Several studies have been conducted since 1950s in order to
determine whether or not smoke venting should be used with a
sprinkler protection system. In 2001, Beyler and Cooper
1reviewed13 experimental studies and 34 position papers on the subject and
evaluated the negative and positive claims of combined use of both
protection systems. These studies ranged from 1954, in response to
a particularly dramatic fire that devastated Livonia automobile factory,to the last large fire tests conducted in 1998. This review showed no
evidence that venting has a detrimental effect on sprinkler perfor-
mance. However, it brought out the lack of repeatability due to thenature of the tests and the cost of large ‐scale fire tests.
A previous paper from Persson and Ingason
2also pointed out the
lack of data, responsible for the lack of consensus on the subject. Theypresented needs for future works in order to finally draw conclusions
and bet on the evolution of CFD.
Since 2001, few works have been conducted on the subject. Li
et al
3,4highlighted experimentally and analytically a relationship
between sprinkler operation pressure and exhaust flow rate from a
horizontal vent. Zhang et al5observed situations without
destratification after sprinkler activation. These observations were alsoreported more recently by Morlon et al,
6at least apart from the area
directly impacted by the sprinkler.
The present study was carried out as a collaborative work
between CNPP and LEMTA laboratory, aiming at identifying interac-
tions which could occur between natural Smoke and Heat Exhaust
Vent Systems and sprinkler systems. The objectives are 3 ‐fold: (1) aReceived: 15 September 2016 Revised: 11 September 2017 Accepted: 10 October 2017
DOI: 10.1002/fam.2487
Fire and Materials . 2017;1 –8. Copyright © 2017 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/fam 1

better understanding of involved phenomena, (2) an evaluation of the
interaction effects (possible modification of the sprinkler activationtime, effects on visibility and temperatures), and (3) an evaluation ofthe numerical tool's ability to predict the smoke/sprinkler interactions.
The next step would be to provide an orientation toward the proposal
of design practices for the coexistence of both protection systemsbased on scientific arguments.
Real scale experiments were conducted at CNPP in order to
investigate interactions that may occur in a compartment where both ofthese systems are in use. A first series of 72 tests was conducted in an
8m×1 2m×4mt e s tr o o mw i t hs p r i n k l e r sa n dn a t u r a ls m o k ev e n t .
Thermocouples, pressure probes, opac imeters, and video acquisition were
used for the experimental characteriza tion. This first test campaign aimed
at gathering experimental data for a validation of the numerical simula-
tions with Fire Dynamics Simulator (FDS) code used in this study.
2|EXPERIMENTAL SETUP
2.1 |Compartment configuration
Tests were conducted in a 2 ‐pitch roof (13% slope) hall with a 100 ‐m2
concrete slab and a maximum heigh t of 4 m. East and North walls were
made of 20 ‐cm‐thick autoclaved aerated concrete, while South and West
walls, as well as roof, were made of single ‐skin steel cladding. Three win-
dows were located on the West wall and 2 on the South one with a1.3 m × 2.0 m door. A sketch of the compartment is reported in Figure 1.
2.2 |Fire source
The fire source consisted of a fuel oil spray burner. For all the tests pre-
sented in this paper, fuel was injected in the atomizing nozzle at a
steady fuel flow rate. The heat release rate (HRR) was estimatedthanks to Equation 1,
7
HRR ¼η_mfuelΔhc (1)
where η= 0.9 is the combustion yield, _mfuelthe fuel flow rate, and
Δhc= 42200 kJ/kg the heat of combustion. The use of this equation
is based on the assumption that all the injected fuel burns. To confirmthis hypothesis, the fraction of unburned fuel was estimated thanks to
a load cell and a metal tray under the spray collecting unburned fueldroplets. Results showed that less than 4% of fuel was lost.
In this study, 2 HRR were used: 400 and 800 kW. Spray ignition
was achieved thanks to a small propane pilot flame. A small metal plate
together with a metal lattice was used to protect the pilot flame fromthe sprinkler spray. For all the tests presented in this paper, the burner
was located close to the North wall. The fire growth was almost instan-
taneous as nominal fuel flow rate was reached in less than 5 seconds.
2.3 |Sprinkler protection
In order to be representative of a real sprinkler system, the installation wasdesigned according to the French standard APSAD R1.
8Twelve sprinklers
were installed with 3 m × 3 m spacing. Sprinkler heads were standard
pendant spray with a ResponseTime Index (RTI) of 108 ( m.s)1/2,a no r i f i c e
flow coefficient (or K ‐factor) equal to 81.2 L.m i n−1.bar−1/2, and an activa-
tion temperature of 68°C, as reported by the manufacturer. In addition to
automatic activation, a system using a metallic wire and a pulley allowed
us to manually break the glass bulb in some tests. The water was providedby a fire hydrant connected to CNPP main water supply at a nominal
pressure of 10 bars. A pressure regulator was used in order to decrease
pressure to our application range.
For this study, only the sprinkler no 6 (shown in Figure 1) was
used. The operating pressure ranged from 0.5 to 2 bar in order to fulfill
French recommendations regarding High Hazard Production facilities.
2.4 |Smoke and heat vent
A single 1 m2, 30 cm high, steel covered smoke vent was installed on
the West pan of the roof (as shown in Figure 1). Fresh air supply wasmade through the south door of the compartment. The vent was
operated manually from the outside of the compartment.
2.5 |Velocity measurement
Two differential pressure probes (McCaffrey) were mounted in the
horizontal section of the vent. Thanks to temperature measurementsand considering smoke as an ideal gas, the flow rate was evaluated
through the vent.
2.6 |Temperature measurement
Four trees of 11 K ‐type thermocouples (from 0.5 to 3.5 ‐m height)
were used in order to determine temperature distribution over thecompartment. Their positions are reported in Figure 1. Additionalmeasurements were done at various locations. Two thermocouples
were located in the vent (close to McCaffrey probes). Three others
were set in the section of the door at 100, 150, and 190 cm abovethe floor. Note that in the following sections, quantitative analyses
of temperatures are limited to thermocouples located far enough
from the activated sprinkler to avoid any uncertainty associated todroplet deposition. Finally, a thermocouple was located close to
the sprinkler no 6 to measure near ‐gas temperature before sprinkler
activation, as well as to detect sprinkler activation when wetted,considering the sharp and instantaneous temperature decrease
FIGURE 1 Sketch of the compartment and metrology location2 TRÉVISAN ET AL.

induced by the injected water (but avoiding any discussion on the
value itself after activation).
2.7 |Opacity
Homemade laser opacimeters were used, as presented in a previouswork done at LEMTA,
6in order to evaluate the absorption coefficient
of the gaseous medium along an optical path, thanks to Equation 2
I¼I0· exp −κLðȚ (2)
where I0and Istand for the signal measured by a photodiode before
and during the test, κis the extinction coefficient, and Lthe length of
the optical path between the laser diode (with light emission at635 nm) and the detector. The extinction coefficient is related to the
soot fraction, and various definitions are found in the literature. For
the present work, we used the formulation suggested in Mowreret al
9(which is also introduced in FDS, the code presently used for
numerical predictions and comparisons with our experimental results
as commented in the next section):
κ¼κmρYsoot (3)
where κmis the specific light extinction coefficient (for which the com-
monly used value 8700 m2/kg has been adopted here), ρis the gas density,
andYsootis the soot mass fraction. Two opacimeters were used during our
tests. The first one was located at the top of the air inlet and provided a
way to detect the smoke exhaust at this location. The second one was
at the vent, but results produced by this latter opacimeter were finallyabandoned, because of strong therma l stresses due to harsh conditions
possibly resulting in a misalignment between emitter and detector, which
could be interpreted as an erroneous decrease in transmittance.
2.8 |Additional metrology
The water flow rate was controlled using a flow meter. This flow was
also calculated thanks to a measurement of pressure at the sprinklerhead, with Equation 4.
Q¼kffiffiffipp(4)
where Qis the water flow rate in L/min, kthe sprinkler “K‐factor ”, and
pthe pressure at the sprinkler head. Atmospheric conditions wererecorded by a weather station. Initial temperature and relative
humidity inside the compartment were measured by hand ‐held
thermo ‐hygrometer. Video recording made by 2 cameras allowed us
to visually estimate the smoke layer height, thanks to 4 LED gauges.
A third camera was used outside the compartment to visualize the
smoke plume exiting through the vent.
3|EXPERIMENTAL RESULTS
Among the 72 tests conducted during the experimental campaign, 29
are commented in the present paper. They are presented in Table 1.
In all these tests, the fire was ignited 60 seconds after acquisition start.Nominal fuel flow rate, hence HRR, was reached in less than 5 seconds.
The vent was opened at t
ventand the sprinkler activation occurred at
tspk. The discussion is mainly based on temperature measurements
and smoke observations.
One of the arguments against vent opening prior sprinkler activation
is that, by design, the former will remove heat from the building, possiblydelaying or even preventing activation of the latter. To investigate thisassumption, tests were carried out where the vent was opened at various
instants before a reference sprinkle r activation time. For both HRR values,
tests were first conducted without vent, to evaluate this reference time(tests 1 to 7 for the HRR of 400 kW also conducted for repeatability inves-
tigation and tests 23 to 24 for the HRR set to 800 kW). Note that “No
vent”only refers to the Smoke and Heat Exhaust Vent Systems in the
cases when it was not opened, but the inlet door at the south wall was
always kept open during all tests, allowing exchanges with the outside.
Then, tests were done opening the ve nt prior to the expected activation
time of the sprinkler (tests 8 to 12 for the HRR set to 400 kW and tests25 to 27 for the HHR equal to 800 kW). Finally, tests 13 to 22 for the
HRR set to 400 kW (respectively, tests 28 and 29 at 800 kW) were done
to evaluate the successive cooling effects by both the sprinkler and thevent, this latter being opened before or after the sprinkler activation.
3.1 |Evaluation of a reference sprinkler activation
time.
The preliminary tests (nos 1 to 7) conducted on sprinkler no 6 to eval-
uate a reference sprinkler activation time, without vent, were also used
TABLE 1 List of the tests conducted to investigate the sprinkler/vent interactions
No Test Type of Sprinkler Activation / Aim of the Test HRR, kW tvent,s tspk,s
1–7 Automatic activation. Repeatability test for evaluation of the sprinkler
activation time.400 No vent Measured
8–12 Automatic activation. Effect of vent opening before sprinkler activation,
tventvaried by 60 ‐s time step400 300 to 540 Measured
13 Manual activation. Comparison of the respective cooling effects due to
sprinkler and vent, tventvaried by 30 ‐s time step400 No vent 240
14–22 400 120 –360 240
23–24 Automatic activation. Sprinkler activation time without vent 800 No vent Measured
25–27 Automatic activation. Effect of vent opening before sprinkler activation,
tventvaried by 60 ‐s time step800 60 to 180 Measured
28 Manual activation. Comparison of the respective cooling effects due to
sprinkler and vent, vent open before sprinkler activation800 180 240
29 Manual activation. Comparison of the respective cooling effects due to
sprinkler and vent, vent open after sprinkler activation800 300 240TRÉVISAN ET AL. 3

to check the repeatability of the measurements. Results are reported in
Table 2 for HRR set to 400 kW, with the activation time measured forthe sprinkler and the gas temperature measured near the sprinkler atthe exact time of activation.
The evolution of gas temperature near sprinkler no 6 with time is
also reported in Figure 2. Despite slight differences in initial tempera-tures, a good repeatability of temperature evolution can be observed.
However, according toTable 2, activation time of the sprinkler can vary
from 568 seconds (508 seconds after ignition) to 673 seconds(613 seconds after ignition) while temperature only varies between
108.8°C and 115.1°C.
Plunge tests were conducted in order to determine if these dis-
crepancies can be related to the nature of the sprinkler link. Twogroups of 5 sprinklers were immersed in heated oil until breakage of
glass bulb. The standard deviation of the activation time for this test
was found to be 13 seconds. This deviation cannot be the main sourceof discrepancies observed in our tests. Another possible explanation
could be related to the heat transfer between the air flow of the ceiling
jet and the glass bulb of the sprinkler. The sprinkler link temperature isassumed to be governed by Equation 5
10:
dTl
dt¼ffiffiffiffiffi ffi
ujjp
RTITg−Tl ðȚ −C1
RTITl−Tm ðȚ (5)
where Tlis the link temperature, Tgthe gas temperature close to the
sprinkler, and Tmthe sprinkler mount temperature. RTI is the Response
Time Index, C1an empirical constant, and uis the gas velocity of the
ceiling jet impacting the link. Changes in this term would cause modifi-cation of the evolution of the link temperature, hence activation time
of the sprinkler. In order to evaluate this velocity, a McCaffrey probe
was set close to the sprinkler, while care was taken to not disruptthe flow around the link. It appeared that for the 400 ‐kW fire source,
the velocity of the ceiling jet in this particular position was not high
and stable enough to allow full confidence in probe measurements.
However, an order of magnitude around 1 to 2 m/s was observed.Simulations of the sprinkler link temperature were performed based
on relation 5, for different uvelocities and with the gas temperature
registered during one of the evaluation tests (test no 3). It was foundthat this velocity has a strong impact on the link temperature
evolution, a slight variation of uresulting in a significant delay in the
time required to reach a given temperature threshold. For example,the theoretical value of 68°C for the sprinkler activation was reached
20 seconds earlier when the uvelocity was varied from 1 to 2 m/s. A
full study on the activation repeatability is beyond the scope of thepresent work, but this delay in addition to the uncertainty determined
from plunge tests may explain a significant part of the uncertainty
observed in the present tests.
With such a large range of activation time, using the mean value as
reference time would make no sense. Consequently, we decided to
choose the earliest time ( t=568 seconds) as a reference for the HRR
set at 400 kW. With a similar series of preliminary tests, the reference
time for sprinkler activation with the HRR equal to 800 kW was set to
200 seconds approximately.
3.2 |Influence of vent opening on sprinkler
activation time for moderate HRR (400 kW)
As shown in Table 1, tests 8 to 12 were done with the vent opened
prior to the reference activation time of the sprinkler with a 60 ‐s incre-
ment, starting at 300 s. Results are reported in Figure 3. When vent
was open at 300, 360, 420, or 480 seconds, temperature of gas near
the sprinkler no 6 decreased to reach a steady value around 120 sec-onds later. In these tests, sprinkler activation never occurred before960 seconds (end of acquisition). During the test with vent opening
at 540 seconds, the sprinkler activation happened at 505 seconds.
TABLE 2 Activation time of sprinkler n°6 and near ‐gas temperature
Test No tspk,s Tspk,° C
1 568 112.1
2 620 114.6
3 673 113.4
4 573 111.75 670 115.0
6 584 115.1
7 664 108.8
FIGURE 2 Gas temperature near sprinkler no 6
FIGURE 3 Gas temperature near sprinkler no 6 for various tventand
moderate HRR (400 kW)4 TRÉVISAN ET AL.

Another frequent claim made against use of vent with sprinklers is
that the cooling action of the sprinklers is so effective that the benefitsof vent will not be realized. To investigate this claim, tests were con-ducted with manual activation of the sprinkler at t
spk= 240 seconds
(smoke layer height was close to 1.75 m) for various vent opening
times. In Figure 4, temperature evolution is reported at 350 cm onthe thermocouple tree close to smoke exhaust for various vent open-
ing times (positive or negative delays with 30 ‐second increments).
For these tests, HRR was set to 400 kW. When the vent was keptclosed, only a slight temperature decrease to a steady value of 90°C
was observed after sprinkler activation. Other cases can be sorted in
2 categories:
•Vent opening before sprinkler activation. For the –90,−60, and –
30‐second tests, early vent operation led to a decrease of temper-
ature values, with a higher rate as vent opening occurred close tothe sprinkler activation. For the –120‐second test, vent operation
slowed temperature elevation until 240 seconds. At this time, and
for all tests, sprinkler was activated, and temperature quicklydecreased to reach a steady value approximately 90 seconds later.
The same temperature decrease was observed for simultaneous
activation of sprinkler and smoke exhaust.
•Vent opening after sprinkler activation. In these tests, a first slight
temperature decrease caused by sprinkler was observed. After
vent opening, the temperature quickly decreased at the same rate
regardless of opening time and reaches quite the same steadyvalue as in previous tests.
The height of the smoke ‐free layer and the visibility were esti-
mated qualitatively, based on observations with cameras and thanks
to a 2.5 ‐m high graduated scale, located at the center of the hall (while
an opacimeter was simultaneously used for a quantitative validation ofthe soot extracted at the door inlet, as will be discussed in the sectiondevoted to the numerical simulations). Without vent opening, the
smoke ‐free layer was logically observed to decrease with time. At time
t=110 seconds, the air/smoke interface reached the top of the grad-
uated scale (smoke ‐free layer height equal to 2.50 m). Then, the smokeinterface was evaluated at 1.75 m above the floor at 180 seconds, at
1.35 m at 240 seconds (time of sprinkler activation), before a continu-ous decrease until the hall was fully filled with smoke at 360 seconds(graduated scale no more visible). When the vent was opened prior
to the sprinkler (at time t=180 seconds), the smoke ‐free layer height
first re ‐increased up to 2.10 m, before decreasing once again when the
sprinkler was activated, finally stabilizing around 1.30 m above the
floor. When the vent was opened after the sprinkler activation (at time
t=300 seconds), the re ‐increase of the smoke ‐free layer height was
observed later, but it was still significant, finally reaching the same
height around 1.30 m. This means that the vent opening had the same
positive influence on the smoke ‐free layer whatever the scenario of
opening before or after sprinkler activation.
3.3 |Influence of vent opening on sprinkler
activation time for higher HRR (800 kW)
Similar tests were conducted with an 800 ‐kW fire source. Vent was
first opened before the evaluated reference time of sprinkler activation
(approximately 200 s based on preliminary tests no 23 –24). The vent
was opened after 60, 120, and 180 seconds, respectively (tests no25–27). Temperatures measured for the smoke near sprinkler no 6
are reported for all tests in Figure 5. Sprinkler activation always hap-
pened, regardless of vent opening time. Moreover, sprinkler activation
occurred at similar or even earlier time with vent opening. Hence, itappears that the assumption of break delay or default due to vent
opening does not hold for realistic fire HRR.
These tests were re ‐conducted with the same 800 ‐kW HRR fire
source and with vent opening occurring 60 seconds before and after
sprinkler activation, set by an arbitrary manual activation at 240 sec-
onds (tests no 28 –29). Temperatures are plotted as profiles at tree
no 3 (Figure 6). Solid lines and filled symbols are related to test withvent opening 60 seconds before sprinkler activation (180 seconds),
while dashed lines and empty symbols refer to vent opening 60 sec-
onds after sprinkler activation (300 seconds). The dashed ‐dotted line
at 175 cm represents the average human height. At t=60 seconds,
temperature is homogenous with height in the compartment, and the
burner is ignited. At t=180 seconds, the same thermal stratification
FIGURE 4 Temperature at 350 cm on tree close to smoke vent for
various tvent
FIGURE 5 Gas temperature near sprinkler no 6 for various tventand
HRR 800KW. Comparison with 2 cases with no vent openingTRÉVISAN ET AL. 5

can be observed in both tests, and the smoke layer height can be qual-
itatively evaluated to be 2.0 m. At t=240 seconds, we can observe
that the temperature in the higher part of the compartment decreasedin the test with early vent opening, while temperature in the lower part
did not change. Meanwhile, in test with late vent opening, temperature
in both higher and lower part increased. At 330 seconds, both temper-ature curves show a decrease in the upper part due to the vent opened
in both cases.
For comparison purpose, snapshots of video recording for both
tests are reported in Figure 7. Until first vent operation, visibility con-ditions are identical in both tests. At t=180 seconds, the smoke layer
height was visually estimated to 1.50 m (confirmed by the earlier
discussed evaluation method based on a camera observing the gradu-ated scale). After this time, we can observe that early vent openingimproved visibility level in the compartment (pictures on the left), even
after sprinkler activation. In the test with late vent operation, visibilityis totally lost 90 seconds after sprinkler operation.
4|NUMERICAL SIMULATIONS
Numerical simulations were carried out with FDS v.611using multiple
processors. Main numerical parameters were set as follows:
•Grid: Domain split in 7 meshes with 10 and 5 ‐cm cubic cell grids.
•Boundary conditions: Outside boundary conditions set to open
and domain extended over air inlet and smoke exhaust.
•Thermal properties of materials set as follows12:
•Concrete: λ= 1.57 W/(m.K), ρ= 2100 kg/m3, C = 1 kJ/(kg.K);
•Aerated concrete: λ= 0.12 W/(m.K), ρ= 350 kg/m3, C = 0.12 kJ/
(kg.K);
•Steel: λ= 50 W/(m.K), ρ= 7800 kg/m3, C = 0.45 kJ/(kg.K);
•Glass: λ= 1 W/(m.K), ρ= 2500 kg/m3, C = 0.75 kJ/(kg.K).
•Sprinkler properties: Rosin ‐Rammler function was used as droplet
distribution function (ddf) with 800 ‐μm mean droplet diameter.
Initial velocity of droplet was set to 10 m/s, and water flow was
prescribed to fit experimental data.
•Fire source: the spray burner was modeled by a burning surface
with prescribed HRRPUA matching experimental HRR calculatedthanks to Equation 1. Preliminary tests were conducted in order
to validate this choice.
First results of simulations are presented in Figures 8 and 9. In
Figure 8, measured and calculated temperatures are reported near
FIGURE 6 Temperature profiles at tree no 3 with early (solid) and late
(dashed) vent opening (HRR = 800 kW)
t = 60s
t = 180s
t = 240s
t = 330s
FIGURE 7 Evolution of visibility in the
compartment with early vent activation (left)and late vent activation (right) (HRR = 800 kW)6 TRÉVISAN ET AL.

sprinkler no 6 for the case with an HRR set to 400 kW and with the
sprinkler activated at 240 seconds. Until vent opening ( t=180 seconds),
experimental and numerical results are in good agreement. When ventis operated, FDS overestimates the cooling effect, leading to a
decrease of temperature while, experimentally, temperature was stabi-
lized. After sprinkler activation, both numerical and experimental tem-peratures drop to a steady value until the end of the test. Thediscrepancies between these values are especially due to the fact that
in our tests, this particular thermocouple is directly impacted by water,
leading to a temperature underestimation.
In Figure 9, temperatures at various heights on tree no 3 are
reported. On the lower part of the compartment, a good agreement
can be observed between numerical and experimental results. At175 cm, temperature is overestimated by the simulation, but its evolu-
tion is correctly reproduced. At 300 cm, temperature is also
overestimated by FDS until cooling effect due to sprinkler activation.The cooling of the upper smoke layer occurs later than what was
observed experimentally and with a slightly higher efficiency.
While temperature levels are mostly overestimated in our simula-
tions, the overall dynamics of smoke is well reproduced. Experimen-tally, the smoke reached the top of the inlet door approximately
60 seconds after burner ignition. At this particular location,comparisons between experimental opacity measurements and
numerical predictions are reported in Figure 10. The agreement is quitecorrect. The sudden increase in opacity is predicted at the right time.This confirms a satisfactory description of the smoke layer flow. How-
ever, a deviation in the opacity values is observed after 140 seconds,
with an absolute discrepancy stabilized around 0.15, still with the sameevolution for both curves. The experimental value is larger than the
numerical prediction, which could be partly explained by clogging, with
some soot deposition on the detector.
The experimental result allows an evaluation of the soot fraction by
inverting relation 3. The experiment at the instant of sprinkler activation
indicates an opacity close to 0.85, corresponding to a transmissivityequal to 0.15. Using relation 3 with an optical path between the emitterand the detector of the opacimeter equal to 1.1 m (the measurement
was conducted along the width of the door), the average extinction
coefficient characterizing the smoke exhaust at the door is 1.7 m
−1.
Inverting relation 4, this provides an estimated soot mass fraction
around 1.8.10−4kg/kg. The numerical simulation predicts an average
value around 1.1.10−4kg/kg at this location. This is the same order of
magnitude, but this significant discrepancy would require further inves-
tigations, both refining the experimental setup and the simulations.
Beside these comparisons between experimentations and numer-
ical simulations, the computational results can be used to provide data
hardly measurable. For instance, an evaluation of the soot flow
extracted at the vent was performed to see what part of the producedsoot is extracted and whether or not this might be affected by thesprinkler activation. Results show that the mass flow rate for soot is
close to 0.28 g/s for the 400 ‐kW HRR (respectively, 0.56 g/s for the
800‐kW case), which is actually related to the soot yield set to
0.03 kg/kg in the simulations. The vent opening results in an instanta-
neous soot mass flow at the exhaust, with a peak followed by a
decrease down to a stationary value reaching 85% of the producedsoot, weakly affected by the scenario (opening before or after sprinkler
activation). This means that the quality of the extraction is the same
whether the sprinkler is activated or not.
Moreover, an energy balance was added to the simulation as a
post‐processing step, following the method by Jenft et al
13in order
to evaluate the part of HRR really withdrawn by the vent and here
FIGURE 8 Temperature near sprinkler no 6
FIGURE 9 Temperature at Tc3 (350, 175, and 100 cm)
FIGURE 10 Opacity measurements and predictions at the top of the
doorTRÉVISAN ET AL. 7

again, whether or not this might be affected by the sprinkler activation.
This energy balance involves the heat produced by the fire, the changein internal energy of the compounds, the losses at walls, the heatexchanges with the droplets, and the losses at the various outflows,
in particular through the open vent (see Jenft et al
13for the details
of the formulation). Numerical results showed that the vent openinginduces a strong heat exhaust, with a peak followed by a decrease
toward a plateau between 25% and 28% of the heat produced, weakly
affected by the HRR and independent of the opening before or aftersprinkler activation. Of course, regarding the total energy extracted
from the hall and integrated over time, the earlier the opening occurs,
the larger the released energy is. This provides some answers to thequestions raised in the introduction regarding the possible influenceof the sprinkler activation on the vent efficiency.
5|CONCLUSIONS
A study on interactions between natural smoke, heat vent, and sprin-kler system was carried out experimentally and numerically. So far, 2
different aspects were studied: influence of earlier vent opening on
sprinkler activation time and effects of early and late vent openingon temperature field with controlled sprinkler activation.
Regarding the activation time of sprinkler, it appears that for real-
istic HRR value (800 kW), vent opening before sprinkler activationdoes not lead to break default and have little effect on delay.
Concerning the effect of vent opening on temperature field,
experiments with manual sprinkler activation coupled with variable
vent opening time were conducted. Both 400 and 800 ‐kW fire sources
were used. It was observed that early vent activation kept temperature
from increasing in the lower part of the compartment. On the same
time, it resulted in a decrease of temperature level in the upper part.
While spray fire with constant HRR is not representative of a real
fire event by many aspects, high repeatability of gathered data allows
us to build a strong validation case for our numerical simulations.Observations made during this first test campaign will be further inves-tigated in future work.
First simulations of experiments were presented. The general
trends are in a good agreement with experimental data. Dedicatedcomputations of the soot and heat extracted at the vent were per-
formed. They showed that the vent allowed the extraction of 85% of
the soot produced and around 25% of the heat released, not (orweakly) affected by the scenario of a vent opened before or after
sprinkler activation. However, additional sensitivity analyses and fur-
ther computations are still required in order to improve results accu-racy. A PDA characterization of sprinkler spray properties is also
planned to determine accurately the droplet distribution function of
our water spray as well as droplet initial velocity.
This is a first step of observation and understanding of phenom-
ena that may occur when sprinklers and smoke vents coexist. A second
campaign on a larger scale and based on our present results will beconducted. Tests will be carried under a 270 ‐m
2adjustable ‐height roof
(up to 12 m). Various fire sources will be used such as large pool, gasburner, or rack storages.
ACKNOWLEDGEMENT
The authors express their gratitude to the GIF, French trade associa-
tion for manufacturers and manufacturer/installers of fire resistance
and smoke exhaust equipment, member of the FFMI (French fire
equipment federation), which supports this research.
ORCID
Pascal Boulet
http://orcid.org/0000-0001-5151-7526
REFERENCES
1. Beyler CL, Cooper LY. Interaction of sprinklers with smoke and heat
vents. Fire Technol . 2001;37(1):9 ‐35.
2. Persson B, Ingason H. Modelling of Interaction Between Sprinklers and
Fire Vents: Present Knowledge . Swedish National Testing and Research
Institute; 1996.
3. Li KY, Huo R, Ji J, Ren BB. Experimental investigation on drag effect of
sprinkler spray to adjacent horizontal natural smoke venting. J Hazard
Mater . 2010;174(1):512 ‐521.
4. Li KY, Spearpoint MJ, Ji J, Huo R, Li YZ, Hu LH. A mathematical model
of the drag component of a sprinkler spray adjacent to horizontal
smoke vents. J Fire Prot Eng . 2010;20(1):27 ‐54.
5. Zhang CF, Chow WK, Huo R, Zhong W, Li YZ. Experimental studies on
stability of smoke layer with a sprinkler water spray. Exp Heat Transf .
2010;23(3):196 ‐216.
6. Morlon R, Boulet P, Parent G, et al. Study of de ‐stratification and
optical effects observed during smoke/mist interactions. Fire Technol .
2015;51(5):1231 ‐1248.
7. Babrauskas V. Heat release rates (Chap. 3). In: SFPE Handbook of Fire
Protection Engineering . 3rd ed. New England: SFPE; 2002:3 ‐37.
8. Référentiel APSAD ‐R1 Règle d'installation —extinction automatique à
eau type sprinkleur (2014), CNPP.
9. Mowrer FW. Handbook of Fire Protection Engineering Fifth Edition.
Chapter 33 —Enclosure smoke filling and fire ‐generated environmental
conditions . Springer; 2016.
10. Heskestad G, Bill RG. Quantification of thermal responsiveness
of automatic sprinklers including conduction effects. Fire Saf J .
1988;14(1):113 ‐125.
11. McGrattan KB, Forney GP. Fire Dynamics Simulator: User 's Manual .
US Department of Commerce, Technology Administration, National
Institute of Standards and Technology; 2014.
12. VDI Heat Atlas (2006), Springer ‐Verlag.
13. Jenft A, Collin A, Pianet G, Breton A, Muller A, Boulet P. Experimental
and numerical study of pool fire suppression using water mist. Fire Saf
J. 2014;67:1 ‐12.
How to cite this article: Trévisan N, Boulet P, Acem Z, et al.
Experimental and numerical study of interactions betweensprinklers and natural smoke vents. Fire and Materials .
2017;1 –8.
https://doi.org/10.1002/fam.24878 TRÉVISAN ET AL.