Fire Safety Journal 146 (2024) 104141

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Performance of various water-based fire suppression systems in tunnels

with longitudinal ventilation
Ying Zhen Li a, *, Haukur Ingason a, Magnus Arvidson a, Michael Försth a, b
a
Safety and Transport - Fire and Safety, RISE Research Institutes of Sweden, Borås, Sweden
b
Structural and Fire Engineering, Luleå University of Technology, Luleå, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Low pressure, medium pressure and high pressure water-based fire suppression systems were tested in a medium
Tunnel fire scale tunnel (scale 1:3). The primary objective was to investigate which of these systems are most effective in the
Fire suppression suppression or control of different types of tunnel fires. The default low, medium and high pressure systems refer
Heat release rate
to full scale water flow rates of 10 mm/min, 6.8 mm/min and 3.7 mm/min, respectively. Some other water
Water density
Operating pressure
densities were also tested to investigate the effects, as well as different ventilation velocities and activation
criteria. Several series of fire tests were conducted for different fire scenarios. The fire scenarios considered
included idle wood pallet fires, loosely packed wood crib fires, loosely packed wood and plastic crib fires, and
pool fires, with or without a top cover on the fuel load. Comparisons of the three default systems based on the
three parameters: heat release rate, energy released and possibility of fire spread, show that the performance of
the default low pressure system is usually the most effective based on the parameters studied. The default high
pressure system usually yields results less effective in comparison to the default low pressure system. The per­
formance of the default medium pressure system usually lies in between them. The high pressure system behaves
very differently in comparison to the others, in terms of tunnel ventilation velocity, water density, operating
pressure, and the presence of the top cover.

1. Introduction tunnel fire suppression tests. These tests include the Second Benelux
Tunnel tests [5], the IF tunnel tests in the UPTUN project during 2002
To meet the increasing needs of fast transportation, tunnels and and 2004 [6], the IF tunnel tests by Marioff Corporation Oy in 2004 [7],
underground structures are being built worldwide. Examples in Sweden the Hagerbach tests carried out in 2005 by Mawhinney [8] and Tuo­
are the North Link in Stockholm (opened in 2014) and the planned missaari [7], the San Pedro de Anes tunnel tests by Marioff Corporation
Stockholm Bypass (to be completed in 2034). Meanwhile, numerous Oy in 2006 [7,8], the Runehamar tunnel tests by SINTEF & Efectis
catastrophic tunnel fires have occurred over the past decades, causing Nederland BV during December 2007 and January 2008 [9], the San
huge life and economic losses. For example, in the Mont Blanc tunnel in Pedro de Anes tunnel fire tests conducted in the SOLIT project in 2008
1999, 34 vehicles caught fire, resulting in 39 deaths, 900 m long severely [10], and the SOLIT2 project during 2011 and 2012 [11], the San Pedro
damaged structure and 3 years’ closure [1]. de Anes tunnel tests by LTA Singapore in 2011 [12,13], and the Rune­
The need to improve tunnel fire safety levels is recognized by in­ hamar tunnel fire tests by Ingason et al., in 2013 [14]. Although these
ternational organizations and authorities and the use of water-based fire large scale tests varied in scope and performance, they have contributed
suppression systems or Fixed Fire Fighting System (FFFS) in road tunnels to improving the knowledge on suppression efficiency of FFFS. Since
attracts much interest. As an example, the Swedish Transport Adminis­ then there have been some additional large scale fire tests, e.g., the
tration has planned to use FFFS for the Stockholm Bypass and also for Runehamar tunnel fire tests comparing the performance of different
upgrading old road tunnels in Sweden [2]. In the past two decades, there large droplet nozzles by Ingason et al., in 2016 [3], the heptane pool fire
have been numerous large scale fire tests conducted with FFFS, mainly tests by Chang et al. [15], the diesel pool fire tests by Li et al. [16], and
to validate the performance of FFFSs prior to their installations in real the high pressure water mist tests in the Runehamar tunnel by Lei et al.,
tunnels [3]. Li and Ingason [4] gave a summary of previous large-scale in 2021 [17]. There have also been model-scale tunnel fire tests

* Corresponding author.
E-mail address: [email protected] (Y.Z. Li).

https://doi.org/10.1016/j.firesaf.2024.104141
Received 10 November 2023; Received in revised form 18 February 2024; Accepted 27 March 2024
Available online 29 March 2024
0379-7112/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

conducted to investigate the performance of FFFSs, e.g., the 1:23 scale not as clear as that of low pressure or high pressure FFFS. However, it is
tests by Ingason [18], the 1:10 scale tests with automatic sprinklers by Li reasonable to assume that the water density and the operating pressure
and Ingason [19], the intermediate-scale tunnel tests by Blanchard et al. used lie between the low pressure system and the high pressure system.
[20], the 1:4 scale water spray tests on design fires [21], the 1:4 scale In this study, the water density for the default medium pressure system
tests on toxic gas production [22] and the 1:3 scale automatic sprinkler refers to 6.85 mm/min, i.e., the mean value of the water densities for the
tests [23]. The focus of the literature review is on the suppression of fire default low and high pressure systems. The operating pressure is in a
development using water-based FFFSs in tunnels. Experiments with gas range of 5.2–35 bar. Note that the medium pressure system defined here
burners and numerical work are therefore not included. From these may refer to both the low pressure (12 bar or less) and intermediate
tests, it is recognized that FFFS can limit the fire growth and fire size, pressure water mist systems (12–35 bar) as per NFPA 750 [25].
and prevent the fire spread to adjacent vehicles, thus reducing the
required tunnel ventilation requirements. It also protects the tunnel 1.2. System configuration and activation mode
structure, and reduces the need for passive structural protection, thus
making significant construction and operational savings. FFFSs in tunnels are mostly divided into individual zones, each of
Depending on system operating pressure and water droplet size, which covers a tunnel length of 20–50 m. In case of fire, two or three
FFFSs are usually divided into low pressure and high pressure FFFSs [1], zones nearby the fire source will be activated, the so-called “deluge
while there are also low pressure water mist systems used in tunnels. mode” [1]. There has also been interest to use automatic water spray
For the sake of simplicity and to be more precise, in this study, FFFSs systems in tunnels. No zones are needed but the so-called “sprinkler
are classified into three types: low pressure FFFS, medium pressure FFFS mode” [1] is applied, i.e., the sprinklers are automatically activated in
and high pressure FFFS. Below the characteristics of the individual case of fire and hot smoke, usually with the aid of the embedded bulbs or
systems are briefly described in combination with the large-scale tests other heat sensitive elements. It has been found that they are suitable for
mentioned above. tunnels with low ventilation velocities, but not for tunnels with high
ventilation flows to avoid system failure, i.e., too many sprinklers are
1.1. Characteristics of various FFFSs activated [19,23]. The studies in Refs. [17,21] also show that fully
automatic sprinklers systems in tunnels are plausible for large fires and
All values on pressure and water density given in this section refer to the activation temperature is important as well as water density. In this
full scale. A low pressure FFFS refers to a normal water spray system. study, the deluge mode is used.
Low pressure FFFSs in buildings and warehouses have been used for over Most of the FFFSs have similar configurations in piping and nozzle
100 years. In buildings the systems are referred to as sprinklers as they placements in tunnels. Each nozzle covers a floor area of 8–16 m2, and
usually use sprinklers with individual thermal activation mechanism the spacing between nozzles is mostly between 3 m and 4 m. In this
whereas in tunnels open nozzles are activated in zones by an external study, an average coverage area of 12 m2 is assumed.
operator, hence the use of the different acronym (FFFS) for tunnels. For
ordinary and extra hazard occupancies, sprinklers are designed in 1.3. Research question
accordance with NFPA 13 [24] or other recognized standards. The water
density ranges from 6 mm/min to 16 mm/min, and the operating Despite many installations and tests, there is lack of general knowl­
pressure ranges from 0.5 bar to 5.2 bar. Japan introduced low pressure edge about how to design and optimize FFFS in tunnels with regard to
FFFS into its high-risk urban tunnels in 1963 and a water density of 6 pressure, water density and activation. The previous large-scale fire tests
mm/min is typically used. Australia has installed low pressure FFFSs in mostly focus on validating the performance of a specific system to be
road tunnels since 1992 [1] and the water density used is in the range of used in a given tunnel. No systematic study of the effects of different
7.5–10 mm/min. Sweden has installed its own large droplet low pres­ parameters has been conducted in the referred full scale tests. FFFS af­
sure FFFS in several tunnels such as the Stockholm Northern Link tun­ fects fire development by different suppression mechanisms, i.e., fuel
nels and the Götatunnel in Gothenburg, and the system is presently surface cooling, gas cooling, dilution, radiation attenuation and kinetic
being installed in the new Stockholm Bypass tunnel system [2]. The effects [1]. High pressure FFFS is efficient in suppressing compartment
typical water density used in Sweden is 10 mm/min. In the large-scale fires. However, Heskestad [26] concluded that it is much more difficult
tests with low pressure FFFSs mentioned above, the water densities to suppress fires with high pressure FFFS in spaces where appreciable
tested are in a range of 8–12.5 mm/min. The average water application oxygen depletion cannot take place. This conclusion is supported by
rate for low pressure systems is around 10 mm/min, and the typical tests by Arvidson [27]. The key mechanisms for low and high pressure
nozzle spacing for deluge systems is 3 m–5 m. The spray angle of the FFFSs in tunnels and how they affect system performance have not been
nozzles is typically between 120◦ and 160◦ . In most of those tests, clearly understood.
effective fire suppression was observed, i.e., after activation of FFFS the One key question that needs to be answered is: which type of system
fire development has been effectively reduced and the heat release rate performs most efficiently in suppression or control of different types of
(HRR) has been significantly lower than the maximum value obtained in tunnel fires. To answer this question, various FFFSs were tested in a
a corresponding free-burn test. In this study, the default low pressure medium scale tunnel to investigate the fire suppression performance in
system refers to a water density of 10 mm/min. different types of fire scenarios. In this study, the focus is on how effi­
High pressure FFFSs started to be used in the 1990s [4]. Referring to cient these different systems are in reducing the HRR and the energy
the definitions in NFPA 750 [25], high pressure water mist systems refer released, and preventing the fire spread to nearby vehicles. Effects on
to operating pressures of 34.5 bar or greater. In the past two decades, the gas temperatures and gas concentrations will be presented in separate
interest to use high pressure FFFSs has dramatically increased in Europe. publications.
In the large-scale tests with high pressure FFFSs mentioned above, the
systems usually have an operating pressure between 35 bar and 80 bar 2. Container tunnel fire tests
and a water density between 2 mm/min and 5 mm/min. The typical
coverage area for each nozzle is around 3.5 m × 3.5 m. The spray angle Carrying out a parametric study of numerous variables in a full scale
of the nozzles is mostly between 120◦ and 160◦ . In this study, the default tunnel is expensive and cumbersome. Small scale testing has also some
high pressure system refers to a water density of 3.7 mm/min. limitations considering the scaling of water sprays and evaluation of
The medium pressure system is developed aiming to take advantage system performance. The use of an intermediate size tunnel constructed
of both low pressure (effective fire suppression) and high pressure sys­ using regular shipping containers was considered the best option
tems (effective gas cooling). The definition of a medium pressure FFFS is available for this study. The Froude scaling was applied in the fire tests

2
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

performed. The scale ratio used in this study was 1:3. A summary of the In the tests with solid fuels (idle wood pallets and wood/plastic
scaling laws applied in the tests is given in Table 1. Details about the cribs), the fire sources were mostly placed on a steel platform and on one
Froude scaling can be found in the literature [1]. The length scaling ratio side of the tunnel cross section, as shown in Fig. 2, while in two tests, the
was 1:3 (1 length unit in model scale tunnel corresponds to 3 length fire source was placed along the centerline of the tunnel. The platform
units in full scale tunnel) which results in the following: time, velocity consisted of a 0.55 m high steel frame and two layers of 10 mm thick
and water density are square root of the length scale (1.73 times), water Promatect H boards on top of it. In these tests, 2 mm thick steel covers
operating pressure is the length scale (3 times), and HRR is 5/2 power of were used to simulate the HGVs. Side curtain HGVs with plates on tops
the length scale (15.6 times). Gas temperature and gas concentrations are commonly used due to the convenience in loading. There are also
are approximately the same in both scales. many HGVs having tarpaulins on both sides and the tops. The front and
back sides, however, are typically closed with plates. Therefore, in all
2.1. Test tunnel the tests except the pool fire tests, the front and back covers were placed.
In most of the tests, a top cover was also applied. Note that a completely
The container tunnel consisting of four standard shipping containers closed HGV, e.g., a steel container, is considered to have a lower fire
was 49.2 m long, 2.39 m high and 2.35 m wide, as shown in Fig. 1. The hazard, and thus not of key interest and not tested in this study. Two
tunnel ceiling and the upper parts of the tunnel walls (a height of 0.6 m) small liquid pans were used as the ignition source, each having di­
were protected with 10 mm thick Promatect H boards, within the range mensions of 100 mm (L) × 250 mm (W) × 30 mm (H). These two pans
of − 2 m and +4 m. Within the range of − 1.5 m and 1.5 m, the floor was were placed just behind the front cover, at the level of the lowest pallet
also protected with Promatect H boards to avoid damage to the plywood or crib on the upstream side (see Fig. 2). They were about 10 cm from the
floor. Outside the protected area, the walls and ceiling of the standard front cover. In the idle wood pallet fires, each pan was filled with 150 mL
containers were made of corrugated steel sheets. heptane. In the wood crib tests and wood and plastic crib tests, each pan
was filled with 300 mL heptane as it was more difficult to ignite the
wood crib. However, in one test 450 mL heptane was used.
2.2. Ventilation
In the pool fire tests, the pool was placed along the center line of the
tunnel, and at the floor level.
Two electric fans with a diameter of 0.71 m were placed outside of
the tunnel inlet to create a longitudinal air flow in the tunnel. The fans
2.3.1. Wood pallet fires
were placed at 0.32 m and 1.35 m above the tunnel floor, close to the
The scaling method developed for scaling time-resolved burning
centre of the tunnel cross section, see Fig. 1. A frequency regulator was
behaviors of wood pallet fires [28] was applied. In the tests, the full scale
adjusted to obtain the design values within 0.6 m/s – 2.1 m/s used in
European wood pallets (pine) were scaled down but the scaling laws are
most of the tests. Two perforated steel nets (70 % opening area) were
slightly modified, in the same way as in the literature [23]. The number
used to regulate the flow. In most of the tests, a velocity of 1.8 m/s was
of pallets in one direction is slightly reduced, while the dimensions in
applied but to further investigate the ventilation effects, a velocity of 1.2
that direction are raised accordingly to fulfil the scaling laws for ge­
m/s was also used in some scenarios. The corresponding full scale ve­
ometry, HRR, fire growth rate and energy content presented in the
locities are 3.1 m/s and 2.1 m/s, respectively.
literature [28]. In each of the idle wood pallet fire tests, 72 wood pallets
were used, consisting of 12 vertical layers and 6 piles longitudinally, as
2.3. Fire source shown in Fig. 3. The full scale fire load refers to 8 (L) × 2 (W) × 21 (H) =
420 wood pallets. The wood pallets are made of pine. The humidity of
The goods carried by heavy goods vehicles (HGVs) may vary wood pallets was mostly between 10 % and 11 %.
significantly. It is difficult to know which types of fuels will burn in an
incident. Different types of fire sources were tested, including idle wood 2.3.2. Wood crib fires
pallets, wood cribs, a mixture of wood and plastic cribs, and heptane In the tests, both wood cribs and wood/plastic cribs (mixture of wood
pools. The fire sources were placed at around 26.6 m from the left tunnel and plastic) were used as fire sources. The bottoms of the cribs were kept
portal, close to the longitudinal midpoint of the tunnel. The wood pallet at 40 mm above the platform to facilitate the placement of the ignition
fires refer to the situation where goods are densely packed with a small sources beneath. The humidity of wood pallets was mostly around 12 %.
porosity, while the crib fires refer to the situation where goods are One crib consisted of 56 square long sticks, each having dimensions
loosely packed with a large porosity. of 2.7 m (L) × 0.045 m (W) × 0.045 m (H), as shown in Fig. 4. The
spacing between two neighboring sticks was 0.09 m. The porosity was
Table 1 around 4.3 mm, indicating a weak influence of porosity on the burning
A list of scaling correlations for the model tunnel [1]. rate [29]. According to a previous study by Li et al. [30], in
Type of unit Scaling Equation number well-ventilated tunnels, there exists an upper limit for the heat release
rate in wood fires, which is around 195 kW/m2. As the exposed surface
Heat Release Rate Q̇ (kW) Q̇M /Q̇F = (lM /lF ) 5/2 (1)
area is known, the maximum HRR can be estimated to be 5.3 MW,
Velocity u (m/s) uM /uF = (lM /lF )1/2 (2)
Time t (s) (3)
corresponding to 83 MW in full scale.
tM /tF = (lM /lF )1/2
Energy content E (kJ) (4)
A steel stack was designed and used to place the sticks, as shown in
EM /EF = (lM /lF )3
Fuel mass m (kg)a (5) Fig. 5. The sticks were placed on the eight layers of horizontal frames at
mM /mF = (lM /lF )3
Temperature T (K) TM /TF = 1 (6) the locations specified in Fig. 4. Each horizontal frame had a diameter of
Gas concentration Ya YM /YF = 1 (7) around 10 mm.
Air pressure P (Pa)b PM /PF = lM /lF (8)
q̇″M /q̇″F = (lM /lF )1/2 (14)
Heat flux q̇″ (kW/m2)c 2.3.3. Wood and plastic crib fires
Water droplet size d (mm) dM /dF = (lM /lF )1/2 (15) The configuration of a wood and plastic crib was the same as that of a
Water density q̇″w (mm/min) q̇″w,M /q̇″w,F = (lM /lF )1/2 (16) wood crib. One wood and plastic crib had around 86% wood (48 sticks)
Water flow rate q̇w (l/min) q̇w,M /q̇w,F = (lM /lF )5/2 (17) and 14% plastic sticks (8 sticks). This ratio is close to the previous model
Water pressure Pw (bar)d PW,M /PW,F = lM /lF (18) scale tests with various cross sections [30] and the Runehamar tunnel
*Assume the ratio of heat of combustion ΔHc,M/ΔHc,F = 1. l is the length scale. fire test 1 conducted in 2003 [31]. Note that the red marked squares in
Index M is related to the model scale and index F to full scale (lM = 1 and lF = 3 in Fig. 4 (b) refer to the plastic sticks.
this study). The plastic used was high-density polyethylene (HDPE) with a

3
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 1. A schematic drawing of the medium scale test tunnel (Side view).

Fig. 2. Fuel arrangement for wood pallet fires in the tests. Dimensions are given in mm.

Fig. 3. Detailed drawing of the wood pallet [23]. Dimensions are given in mm.

Fig. 4. Configuration of a crib. Dimensions are given in mm.

density of approximately 1000 kg/m3. The aim was to place the plastic surface area, the maximum HRR is estimated to be 8.4 MW, corre­
sticks evenly within the crib, which usually was obtained but very sponding to 131 MW in full scale. This maximum HRR is in line with
cumbersome in labor. According to a previous study by Li et al. [30], in what was obtained in the Runehamar tests in 2003 [31].
well-ventilated tunnels, the heat release rate per unit area for poly­
ethylene is around 1000 kW/m2. Therefore, based on the exposed

4
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 5. Fuel arrangement for crib fires in the tests. Measures are given in mm.

2.3.4. Heptane pool fires other pressures for the same nozzle, with the aid of the correlation be­
The dimensions of the pool were 0.5 m (L) × 0.8 m (W) × 0.12 m (H). tween operating pressure and droplet size: dm ~ p1/3 [1].
Heptane was used as the fuel. In the tests with FFFS, 25 L of fuel were
used, while 35 L was used in the free burn test. 2.6. Instrumentation

2.4. Fire spread A total of 125 measurement sensors were used, including 0.25 mm
diameter thermocouples for gas and surface temperatures, plate ther­
In the tests with solid fuels, six wood samples were placed down­ mometers for incident heat fluxes, bi-directional probes for velocities,
stream of the main fire load at a height of around 1.6 m above floor, gas analyses for gas concentrations and lasers for visibilities, as shown in
acting as “targets” to investigate the possibility of fire spread, as shown Fig. 8. Between − 18 m (upstream) and +5 m (downstream), ceiling
in Fig. 2. Each target had dimensions of 0.045 m (W) × 0.045 m (L) × thermocouples were placed every 0.59 m. The focus of the paper is on
0.005 m (H). The target denoted “target 1” was closest to the fire source, the HRR and the other measured variables will be presented separately.
0.4 m from the downstream edge of the vehicle mock-up (end cover) or The most important measurement in this study is the HRR, calculated
1.8 m longitudinally from the fire source center. The furthest one was by the oxygen consumption method with CO/CO2 corrections based on
the one denoted “target 6”, positioned 3.8 m from the fire source center the data measured 16.7 m downstream of the fire (Pile B). At Pile B, the
(corresponds to 11.4 m in full scale). The distance between the centers of gas temperatures and velocities were measured at 0.24 m, 0.6 m, 1.2 m,
two neighboring targets was 0.4 m. Notice that all the targets were 1.67 m and 2.2 m above floor, and the gas concentrations were
placed within the water spray region. measured at 0.6 m, 1.2 m and 2.2 m above floor. While calculating the
mass flow rate and heat release rate, the cross section at Pile B was
2.5. Various FFFSs divided into five segments in the calculations of the mass flow rate and
heat release rate, following the methodology used for the previous
Three types of FFFSs were tested (see Fig. 6). The nozzle used for the Runehamar tunnel fire tests [32] and model scale fire suppression tests
high pressure system was a multi-orifice nozzle, and the other two [22]. The velocity measurements are in the centre of each segment. A
nozzles were single-orifice nozzles. Each system had two longitudinal flow profile study was conducted close to the downstream tunnel portal
rows of nozzles covering a spray region of around 9.4 m, corresponding using Testo 440 anemometer with closely evenly distributed 25 (5 × 5)
to 28.2 m in full scale, as shown in Fig. 7. Each row consisted of 8 measurement points across the cross section in order to validate and
nozzles. The spacing between nozzles was 1.175 m, corresponding to control the flow measurements. The longitudinal tunnel ventilation
3.525 m in full scale. All nozzle outlets were positioned 0.15 m beneath velocities (corresponding to ambient conditions), u0, were obtained
the ceiling. based on the calculated mass flow rates. The heat release rates were
A summary of the nozzles tested for various FFFSs is given in Table 2. estimated using the oxygen consumption method considering the pro­
In this study, the default low, medium and high pressure FFFSs refer to duction of CO and CO2. The total HRR was the summation of HRR from
water densities of 5.8 mm/min, 4 mm/min and 2.2 mm/min, respec­ these segments, determined by use of Eq. (1) [33]:
tively. The corresponding full scale values are 10 mm/min, 6.85 mm/
∑[ (
min and 3.7 mm/min, respectively. For comparisons, in some tests, the Q̇ = ΔHO2 φi − ΔHO2 , CO
water density of 4 mm/min was tested for both low pressure and me­ i
( )]
dium pressure systems, and the water density of 2.9 mm/min was tested ) 1 − φi XCO,i ṁi MO2 ( )
− ΔHO2 1 − X0,H2 O X0,O2 ,i (1)
for both medium pressure and high pressure systems. The volume me­ 2 XO2 ,i 1 + φi (α − 1) Ma
dian diameters measured at 0.5 m below the nozzle are also given for
information, but two data are predicted form the measured values at where the oxygen depletion factor, φ, is expressed as:

Fig. 6. Photos of the nozzles used in this study. Photo (a) was taken after the fire tests.

5
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 7. A schematic drawing of the FFFS set up. Dimensions are given in mm.

Table 2
Summary of nozzles tested for various FFFSs.
FFFS Nozzle Spray K-factor Narrowest hole Operating Water flow Water density Volume median
System angle diameter pressure, p rate diameter, dm
model full
scale scale
◦
l/(min⋅bar1/ mm bar l/min mm/min mm/ μm
2
) min

Low BETE MP156 120 8.8 3.97 0.39 5.5 4.0 6.85 880
0.82 8.0 5.8 10 861
Medium BETE WL1 120 2.35 2.08 2.9 4.0 2.9 5.0 429a
5.4 5.5 4.0 6.85 349
High Lechler ~130 0.88 0.5 11.5 3.0 2.2 3.7 181
502.448 20.5 4.0 2.9 5.0 149a
a
This diameter was calculated based on the measured one for the same nozzle with the aid of the correlation: dm ~ p1/3 [1].

Fig. 8. Instrumentation and measurements (Side view).

( )
X0,O2 (1 − XCO2 − XCO ) − XO2 1 − X0,CO2 carbon dioxide in the incoming air (ambient), and XO2 and XCO2 are the
φ= volume fractions of oxygen and carbon dioxide, respectively, as
(1 − XO2 − XCO2 − XCO )X0,O2
measured by a gas analyser (dry) at the measuring station downstream
and the mass flow rate of ith segment is: of the fire. The superscript 0 indicates the incoming fresh air flow and
the superscript i indicates the ith segment. X0,H2 O is the volume fraction
ṁi = Cd ρi ui Ai (2)
of water vapor in the incoming fresh air flow prior to combustion
The total mass flow rate is the sum of the mass flow rate of the five (ambient). The upstream water spray may increase the water vapor
segments: concentration of the incoming flow but its influence on the HRR is
∑ considered small as the ambient temperature is low.
ṁ = Cd ρi ui Ai = ρ0 u0 A (3) Note that due to the existence of boundary layers, the velocity
measured close to the wall is usually lower than that measured in the
i

In the above equation, Q̇ is the HRR (MW), ΔHO2 is heat released per centre of the tunnel cross section, although this effect is less for turbulent
unit mass of O2 consumed for complete combustion, ΔHO2 ,CO is heat flows which is the case in this study. In the tests, five velocity mea­
released per unit mass of O2 consumed for combustion of CO to CO2 surement points were vertically distributed, so the boundary effect has
(17.6 MJ/kg), ṁ is the mass flow rate of the ith layer, u is velocity, u0 is been partially reflected in the velocity measurement close to the ceiling
tunnel longitudinal velocity (tunnel average velocity at ambient condi­ and the floor. However, the sidewall effects should still be considered.
tions), Cd is the flow correction coefficient, M is molecular weight (kg/ Another issue is related to the placements of the bi-directional tubes and
kmol), α is a constant (approx. 1.105), X0,O2 is the volume fraction of other nearby accessories that are necessary for testing. For both reasons,
oxygen in the incoming air (ambient), X0,CO2 is the volume fraction of a flow profile survey was conducted to determine the flow correction

6
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

coefficient Cd in Eqs. (2) and (3), which was found to be approximately During each test, when no visible flame was observed from the
0.8. It is worth noting that this coefficient may be different for another portals and the observation window (where the side camera was), the
tunnel or another measurement design, depending on the flow profile measurement results were checked to confirm the fire extinguishment
and the measurement installations. For a new experimental set up, flow and the test was usually terminated after approx. 5 min.
calibration should always be done to improve the accuracy of velocity
and HRR measurements. In this work, for further validation, this flow 4. Results and discussion
correction factor was used in HRR measurements (see Eq. (1)) against
gas fires with known HRRs (ranging from 0.5 MW to 3 MW) and the The following analysis focuses on comparing the performance of the
good agreement further supports its use. The transportation time [1] default FFFSs in terms of reducing the HRR and the energy released, and
from the fire location to Pile B was taken into account for all the flow preventing the fire spread, while the influences of various influencing
parameters. The response time of the gas analyzers, i.e. mainly due to factors on the system performance are also investigated. The HRRs from
the transportation time from the measuring position inside the tunnel to the four test series with different fire sources are presented in sequence.
the measuring cell inside the instrument, was also considered and found The free-burn test is presented and discussed at first, followed by the
to be around 22 s. comparisons of the HRRs for various FFFSs without and with the top
At the measurement stations (Pile A and B), the water vapor was cover. Thereafter the influences of various factors, i.e., ventilation,
filtered before conducting gas analysis measurements, so its influence on water density and operating pressure and activation time, are analyzed.
the gas measurements was eliminated. However, the evaporated water Note that in this study, the default low, medium and high pressure FFFSs
vapor contained in the gas flow at the measurement station may affect refer to water densities of 5.8 mm/min, 4 mm/min and 2.2 mm/min,
the mass flow rate measurement. It is difficult to accurately assess its respectively. The corresponding full scale values are 10 mm/min, 6.8
influence unless the water vapor concentration is well measured. mm/min and 3.7 mm/min, respectively. Furthermore, the tunnel air
Despite this, it is possible to roughly estimate it. The key parameter is the velocity defaults to 1.8 m/s unless otherwise stated in the following
amount of water vapor that may be contained in the gas flow at the sections.
measurement station. The maximum possible value for the total amount The maximum HRRs from the tests are given in Table 3. For each fuel
of water vapor in the gas flow could be estimated using the lower value type, the maximum HRR is obtained in the free burn test without FFFS,
between the total water flow rate (assuming all water was evaporated) and it is 6.3 MW for the idle wood pallet fire, 4.5 MW for the wood crib
and the HRR divided by the heat of gasification of water (assuming all fire, 7.5 MW for the wood and plastic fire, and 1.4 MW for the heptane
heat released was consumed to produce water vapor). The average value pool fire. The proportion of energy that was released or consumed in
for the maximum water vapor volumetric concentration in these tests is each test, E/Etot, is also given in Table 3 where Etot refers to the total
estimated to be around 9.8%. However, the actual value is considered energy released in the corresponding free burn test. The energy released
much smaller, due to three reasons: (1) only a portion of water was in a test is calculated from the integration of the HRR curve of the test. It
evaporated and much water went to drain; (2) only a portion of the heat can be seen that the energy release data generally follow the same trend
released was consumed to produce water vapor and much heat was as the HRR, i.e., a higher HRR usually corresponds to a higher propor­
either lost to tunnel structure or left in the gas flow; and (3) a portion of tion of energy released in the test.
water vapor may condense on the way to the measuring station due to In the presence of FFFSs, at the activation (141 ◦ C), the HRR was
the temperature decay along the tunnel. To sum up, in these tests, as the within 0.3–0.4 MW (at approx. 4 min after ignition) in the wood pallet
amount of evaporated water in the tunnel at the measurement station fire tests, while in the crib tests it was around 0.4 MW without top cover
was considered to be much less than the total amount of the tunnel flow, (approx. 2.8 min), and 0.5 MW (approx. 3 min) with the top cover. This
uncertainty due to water vapor in the HRR estimation was deemed to be corresponds to a heat release rate ranging from 4.7 (0.3) MW to 7.8 (0.5)
small. However, if the ventilation velocity is low, the water vapor con­ MW in full scale. The corresponding activation times were mostly be­
centration at the measurement points may be significant, and thus it may tween 3 min and 5 min. The higher HRR at activation in the crib fire tests
need to be measured and considered in the calculation of the heat could be attributed to the loosely packed fuels where the fire plume can
release rate, see Refs. [33,34]. entrain more air before leaving the vehicle mock-up. In the crib tests, the
presence of the top cover also increases the activation HRR, which
3. Test procedure should be related to the enhanced entrainment since in the case of the
top cover two fire plumes towards both sides are formed. This effect,
A total of 32 fire suppression tests was conducted, including 14 idle however, is not so significant in the idle wood pallet tests where the fuels
wood pallets fires (tests series 1), 10 wood crib fires (test series 2), 4 are densely packed. In the wood crib tests with an activation tempera­
wood/plastic cribs (test series 3), and 4 pool fires (test series 4). Among ture of 300 ◦ C, the HRR at activation was about 0.8 MW (approx. 3.3
these, four free burn tests (marked with bold text) were conducted min), greater than that for the tests with an activation temperature of
without any FFFS to obtain reference data. A summary of these tests is 141 ◦ C, as can be expected. In the pool fire tests, the activation time was
given in Table 3, including some key results such as the fire spread re­ 3.5 min and the HRR was around 0.75 MW. The corresponding full scale
sults. The first number of a test no. refer to the test series, e.g., 201 refers values are around 12 MW, which is higher than what was stated earlier.
to the test 01 in test series 2. In the first two columns, the free burn tests Comparing these values with the maximum HRRs given in Table 3 shows
are marked in bold. The correlations between the water density and that the maximum HRRs are usually much higher than the activation
operating pressure can be found in Table 2. The longitudinal air velocity HRRs, indicating that the fire continues to develop to some extent after
in the tunnel was mostly set as 1.8 m/s but it was 1.2 m/s in some tests. the system activation. However, it can be noted that for the wood crib
The values are only nominal ones as they may vary within a range of fires without the top cover and the pool fires, the maximum HRRs are
±0.1 m/s. close to the activation HRRs, indicating that efficient and prompt fire
During the tests with solid fuels, the FFFSs were activated mostly suppression takes place after the system activation.
when the maximum gas temperature measured beneath the ceiling A summary of results related to the occurrence of fire spread based
reached a value of 141 ◦ C as a detection criterion, in accordance with the on visual observations of the wood samples (targets 1 to 6) after the fire
previous large scale tests [3,35] and model scale tests [21,22]. In some tests is given in Table 3. “No” indicates that the target was not affected
wood crib fire tests, a higher fire detection temperature of 300 ◦ C was by the fire. “Burnt” indicates that the target was completely burnt,
applied. In test series 4 with pool fires, the ceiling gas temperature at the however, if only part of the target was burnt, the approximate per­
early stage was not so high, and a fixed activation time of 3.5 min was centage of the burnt part based on visual observations is given behind.
used, corresponding to a gas temperature of around 100 ◦ C. “Charred” indicates that the target was only charred but not burnt. Note

7
 Y.Z. Li et al.
Table 3
Summary of test series 1–4 on fire suppression.
Test no. FFFS system Fire source Top cover Nominal velocity Water density Tdetection Q̇max E/Etot Fire spread to different targets (samples)
o
m/s mm/min C MW 1 (nearby fire) 2 3 4 5 6 (furthest)

101 Low Wood pallets Yes 1.8 4.0 141 3.4 68% Charred No No No No No
102 Low Wood pallets Yes 1.8 5.8 141 2.6 68% No No No No No No
103 Low Wood pallets No 1.8 5.8 141 0.8 16% No No No No No No
104a Low Wood pallets Yes 1.8 5.8 141 2.9 69% No No No No No No
105 Low Wood pallets Yes 1.2 5.8 141 2.4 54% No No No No No No
106 High Wood pallets Yes 1.8 2.9 141 3.1 73% Burnt~80% Charred No No No No
107 High Wood pallets No 1.8 2.2 141 3.7 81% Burnt~10% No No No No No
108 High Wood pallets Yes 1.8 2.2 141 2.7 60% Burnt~10% No No No No No
110 High Wood pallets Yes 1.2 2.2 141 3.1 74% Burnt~80% Burnt~60% Charred Charred Charred No
111 Medium Wood pallets Yes 1.8 2.9 141 2.7 71% Charred No No No No No
112 Medium Wood pallets Yes 1.8 4.0 141 2.7 65% Charred No No No No No
113 Medium Wood pallets No 1.8 4.0 141 1.4 36% No No No No No No
115 Medium Wood pallets Yes 1.2 4.0 141 1.8 38% No No No No No No
116 Free burn Wood pallets Yes 1.8 0 – 6.3 100% Burnt Burnt Burnt Burnt Burnt Burnt
200a Low Wood crib Yes 1.8 5.8 141 0.8 21% No No No No No No
201 Low Wood crib Yes 1.8 5.8 141 0.9 7% No No No No No No
202 Low Wood crib No 1.8 5.8 141 0.5 3% No No No No No No
8

203 Low Wood crib Yes 1.8 5.8 300 1.1c 11%c No No No No No No
204 High Wood crib Yes 1.8 2.2 141 2.8 79% No No No No No No
205 High Wood crib No 1.8 2.2 141 0.6 3% No No No No No No
206b High Wood crib Yes 1.8 2.2 300 4.0 98% Charred No No No No No
207 Medium Wood crib Yes 1.8 4.0 141 1.0 38% No No No No No No
208 Medium Wood crib No 1.8 4.0 141 0.6 2% No No No No No No
210 Free burn Wood crib Yes 1.8 0.0 – 4.5 100% Burnt Burnt Burnt Burnt Burnt Burnt
301 Low Wood and HDPE Yes 1.8 5.8 141 2.4 55% No No No No No No
302 High Wood and HDPE Yes 1.8 2.2 141 5.9 68% Burnt~70% Burnt~40% Burnt~30% Charred Charred Charred
303 Medium Wood and HDPE Yes 1.8 4.0 141 4.9 61% No No No No No No
304 Free burn Wood and HDPE Yes 1.8 0.0 – 7.5 100% Burnt Burnt Burnt Burnt Burnt Burnt
401 Low Pool fire No 1.8 5.8 FTd 0.8(1.1e) ~100% NA
402 High Pool fire No 1.8 2.2 FTd 0.8 ~100%
403 Medium Pool fire No 1.8 4.0 FTd 0.9 ~100%
404 Free burn Pool fire No 1.8 0 – 1.4 100%
a
Centered fire source.
b
Centered fire source and 1.5 times the amount of ignition fuel.

Fire Safety Journal 146 (2024) 104141

c
Estimated based on test 201.
d
FT: Fixed Time.
e
During the period of malfunction, HRR increases to 1.1 MW.
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

that in the pool fire tests, the fire spread was not tested due to the of the low pressure system, the HRR increases slightly to its peak of 0.8
different placement of the fire source. It is clearly shown that in the free MW at 14 min, and the fire is almost extinguished after 25 min. By
burn tests with solid fuels, all the targets were burnt. In comparison, in contrast, after the high pressure system is activated, the fire continues to
the presence of the FFFS, the targets were usually not burnt. In these increase until it reaches its peak of 3.7 MW at 33 min. For the medium
tests, the low pressure system performed well in terms of preventing fire pressure system, the peak HRR of 1.4 MW is reached at 23 min. Clearly,
spread, and only in test 101 with the lower water density of 4 mm/min in terms of reducing the HRR, the low pressure system performs best, the
the first target was charred. The medium pressure system also performed medium pressure system performs moderately well, and the high pres­
relatively well in preventing fire spread, and only in two tests the first sure system performs worst. The influence on the gas temperatures
target was charred. By contrast, the high pressure system performed less shows a similar trend and it will be presented in a separate paper.
effective in preventing the fire spread. In the majority of the tests with The results indicate that the system performance, in terms of effi­
the high pressure system (5 out of 8 tests), one or more targets were ciency in reducing the HRR, is improved as the water flow rate increases
partially burnt. In the following sections, more detailed comparisons of and the operating pressure decreases. This should be attributed to three
fire spread will be made together with the HRR and the proportion of factors. Firstly, larger droplets from a lower pressure nozzle can better
energy released. penetrate the fire plume. Secondly, a higher water flow rate cools down
the fuel surfaces more effectively. Thirdly, the influence of forced air
flow on the trajectories of larger droplets is less in comparison to very
4.1. Wood pallet fires
fine droplets which may be carried away by the air flow for a long dis­
tance before hitting the fire load or tunnel floor, as shown by Rein et al.
4.1.1. Free-burn test
[36].
The HRR curve from the free burn wood pallet fire test at a longi­
From Table 3, the same trend can be found for the proportion of
tudinal velocity of 1.8 m/s with the top cover is shown in Fig. 9 (the
energy released in the tests (16%, 36% and 81% for the low, medium
black line). The fire develops steadily up until 16.5 min (when the HRR
and high pressure systems, respectively). Furthermore, there was no fire
reaches around 4 MW). At 17.5 min, the HRR reaches 5 MW (78 MW in
spread to the targets in the tests with the low and medium pressure
full scale), and after 0.1 min (17.6 min) the HRR suddenly rises to its
systems, however, the first target was partially burnt in the test with the
peak of 6.3 MW (98 MW in full scale). This should be related to the
high pressure system. It should be noted that the fire spread is affected
collapse of the first wood pile on the upstream side, towards the side
not only by the amount of water delivered to the targets, but also by the
wall nearby, as observed in the test, which leads to a sudden increase in
heat exposed to the targets, which depends on the HRR, the ventilation
the total exposed fuel surface area and thus an immediate increase in the
velocity, and the gas cooling and radiation attenuation effects of the
HRR. This phenomenon was also noticed in the previous tests [23,28]. In
water droplets. The results of fire spread should be considered as the
general, the peak HRR correlates well with the previous free burn test in
combined effect of these two factors.
1:4 scale where the obtained peak HRR corresponds to 96 MW in full
scale [21]. It is worth mentioning that in the 1:4 scale tests, besides the
4.1.3. Comparison of default FFFSs with top cover
50 pallets as the main fire load, 10 pallets were used as the target to
The performances of the three default FFFSs in the idle wood pallet
evaluate possible fire spread and they were also involved in burning in
fires with the top cover at 1.8 m/s are compared in Fig. 10. The acti­
that free burn test. However, in the current study, no wood pile was used
vation HRR is within 0.3–0.4 MW at approx. 4 min. The maximum HRRs
as the target and therefore the fire load was slightly lower than that in
for the default low, medium and high pressure systems are 2.6 MW, 2.7
the 1:4 scale tests.
MW and 2.7 MW, respectively. They are rather close to each other, and
only show an insignificant decrease of HRR with the increasing water
4.1.2. Comparison of default FFFSs without top cover
density. The fire curves are also similar. The characteristics of the low
The performances of the three default FFFSs in the idle wood pallet
pressure system resulted in improved results compared to others at the
fires without the top cover are compared in Fig. 9. The activation HRR is
early stage (before 12 min), which could be related to the strong
within 0.3–0.4 MW (approx. 4 min). It is shown in Fig. 9 that the fire is
effectively suppressed by the low-pressure system. After the activation

Fig. 9. HRRs in wood pallet tests with various FFFSs and without top cover (uo Fig. 10. HRRs in fire suppression tests with top cover (uo = 1.8 m/s). Data from
= 1.8 m/s). Data from the free-burn test is plotted for comparison. the free-burn test is plotted for comparison.

9
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

penetration capacity of the large droplets. However, between 12 min 1.2 m/s, the default low and medium systems produce similar results for
and 26 min, the medium pressure system performs better. After 26 min, the first 30 min (except for a short period of around 20 min), but for the
the HRR curves for the low and medium pressure systems are close to low pressure system, the fire redevelops after 30 min although the sec­
each other. Between 12 min and 20 min, the HRRs for the low and high ond peak HRR is lower. By contrast, the high pressure system performs
pressure systems are almost the same. considerably less effectively. The HRR continues to increase until it
The results from these tests with the top cover as presented in Fig. 10 reaches its peak of 3.1 MW at 32 min. This also resulted in large portions
are very different to the results without the top cover presented in Fig. 9, of the first two targets (about 80% and 60%) being burnt and three more
which justifies the influence of the top cover. One may probably expect targets being charred in the test with the high pressure system, whereas
that with the top cover, fire suppression is much more difficult to ach­ there was no fire spread for the low and medium pressure systems, as
ieve as the water cannot be directly discharged to a large portion of the shown in Table 3. The proportion of energy released in the test is 54%,
fuels beneath the top cover. This appears to be true for the low and 38% and 74% for the low, medium and high pressure systems, respec­
medium pressure systems. However, it is not the case for the high tively. The higher proportion observed for the low pressure system as
pressure system which performs slightly better with the top cover. After compared to the medium pressure system is mainly attributed to the
20 min, the HRR without the top cover is higher, with a peak of 3.7 MW, redevelopment of the fire after 30 min.
compared to 2.7 MW with the top cover. This indicates that the key
suppression mechanisms of the high pressure system differ from those 4.1.4. Influence of ventilation velocity for individual default systems
for the low and medium pressure systems. The performance of the high The effects of the longitudinal ventilation velocity on the perfor­
pressure system relies more on gas cooling and dilution, thus requiring mance of individual default systems are shown in Fig. 12. For the low
finer droplets, whereas the performance of the low pressure system re­ pressure system, the maximum HRRs at the two velocities (1.8 m/s and
lies more on fuel surface cooling, thus requiring a higher water density 1.2 m/s) are close to each other, but the lower velocity seems to have a
and larger droplets. The key suppression mechanism of the medium positive effect on the system performance. At 1.2 m/s, the fire reaches
system appears to be more similar to the low pressure system. These the maximum HRR at around 19 min and starts to decay afterwards, and
results are very much in line with the large scale tests carried out by the corresponding time at 1.8 m/s is around 35 min. For the medium
Arvidson [27] on cargo in a ferry setup in a fire laboratory (no longi­ pressure system, the maximum HRR at 1.8 m/s is much higher. Similar
tudinal flow) using different top cover configurations, pressures and to the low pressure system, the fire decays earlier at the lower velocity.
flow rates. From Table 3, less energy is released at the lower velocity for both the
As shown in Table 3, the proportions of energy released in these tests low and medium pressure systems, i.e., the proportion of energy
are similar (68%, 65% and 60% for the low, medium and high pressure released is 54% at 1.2 m/s compared to 68% at 1.8 m/s for the low
systems, respectively). However, there is a clear difference in the fire pressure system, and 38% at 1.2 m/s compared to 65% at 1.8 m/s for the
spread between these tests. There was no fire spread for the low pressure medium pressure system. To sum up, a decrease in velocity from 1.8 m/s
system. The first target was charred for the medium pressure system, and to 1.2 m/s improves the performance of the low and medium pressure
it was partially burnt for the high pressure system. This indicates that the systems. This could be related to the slower fire growth at the lower
risk of fire spread increases from the default low pressure system, the velocity [1], thus allowing more time for the suppression systems to take
default medium pressure system to the default high pressure system. effect.
This seems to be intimately related to the water density and not so much By contrast, the results for both HRR and energy released are
to the radiation attenuation as one could anticipate. different for the high pressure system. The performance at 1.2 m/s is not
For comparison, the results for the three default FFFSs at a longitu­ as effective compared to that at 1.8 m/s. The maximum HRR at 1.2 m/s
dinal ventilation velocity of 1.2 m/s are plotted in Fig. 11. The activation is 3.1 MW, compared to 2.7 MW at 1.8 m/s. As expected, more energy is
HRR is about 0.3 MW at approx. 3.2 min. This value is slightly less than released in the test with the lower velocity for the high pressure system.
that at 1.8 m/s, as the temperature is inversely proportional to the ve­ Specifically, the proportion of energy released is 74% at 1.2 m/s
locity [37]. As no free burn test was performed at 1.2 m/s, the free-burn compared to 59% at 1.8 m/s, as shown in Table 3. Regarding fire spread,
HRR curve at 1.8 m/s is plotted as a reference. At the lower velocity of the results for the lower velocity are also worse, with more targets being
burnt or charred at 1.2 m/s. One probable reason could be that a higher
velocity increases the longitudinal travel distance of the fine droplets.
This indicates that droplets from nozzles positioned further upstream of
the fire can have a positive influence on the fire size, thereby increasing
the effects of gas cooling and dilution, and possibly surface cooling.

4.1.5. Influence of water density for individual systems

Water density is a key parameter for a FFFS. The required water
density in an engineering design is usually related to the fire load and its
configuration. In this section, results from the tests with varying water
densities are presented for individual systems. It should be noted that,
for a given system, an increase of the water density also increases the
operating pressure, thereby resulting in finer water droplets. Therefore,
this section investigates the combined effects for individual systems.
Fig. 13 shows how different water densities affect the performances
of the individual systems. For the low pressure system, increasing the
water density from 4 mm/min to 5.8 mm/min reduces the maximum
HRR from 3.4 MW to 2.6 MW, whereas the energy released in the tests is
very similar, as shown in Table 3. In both tests, there was no fire spread
to the targets. For the medium pressure system, the increase of water
density from 2.9 mm/min to 4 mm/min delays the early fire develop­
ment between 10 min and 30 min, although the maximum HRRs are
Fig. 11. HRRs in fire suppression tests with top cover (uo = 1.2 m/s). Data from both around 2.7 MW. The energy released in the test with 2.9 mm/min is
the free-burn test (1.8 m/s) is plotted for comparison. slightly higher, as shown in Table 3. In both tests, the first target was

10
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 12. Influence of the ventilation velocity on the system performance.

charred. For the low and medium systems, as expected, improved per­ this study is to find an optimal system, it is of interest to know how the
formance is found at higher water densities. This should be related to the operating pressure solely affects the system performance. Therefore, the
improved ability to penetrate fire plumes and the greater cooling effect results from different systems with same water densities are directly
on fuel surfaces. compared below.
For the high pressure system, the increase of water density from 2.2 Note that the low pressure test 101 and the medium pressure test 112
mm/min to 2.9 mm/min shows an insignificant influence on the fire had the same water density of 4 mm/min, but different operating
suppression performance. As a matter of fact, it can be noticed that the pressures. The influence of operating pressure on the fire suppression
high pressure system with the higher water density even performs performance can thus be reflected in Fig. 14(a). The increase of the
slightly worse, i.e., both the maximum HRR and the energy released are pressure appears to be beneficial to the fire suppression performance.
slightly higher. Moreover, from Table 3, at 2.9 mm/min, the first target The maximum HRRs are 3.4 MW for the low pressure system and 2.7
was burnt approximately 80 %, compared to 10% at 2.2 mm/min, and MW for the medium pressure system with the same water density of 4
the second target was charred while it was not at 2.2 mm/min. mm/min. However, it should be noted that the corresponding nozzle
The slightly reduced performance at the higher water density could operating pressure for the low pressure system is around 0.4 bar, which
be related to the weaker penetration capability of finer droplets and the is close to its lower pressure limit. Table 3 shows that the energy released
greater influence of ventilation on the trajectories of water droplets. in these tests is very similar, and the first target was charred in both
Note that for a given system, a higher water density refers to a higher tests.
operating pressure, and thus finer droplets. The results indicate that the Furthermore, the results for the medium and high pressure systems
positive effect of the increased water density was counteracted by the with the same water density of 2.9 mm/min but different operating
negative effect of finer droplets for the high pressure system. In other pressures (2.9 bar and 20.5 bar respectively) are compared in Fig. 14(b).
words, too small water droplets are not as beneficial to the system The increase of the pressure from 2.9 bar to 20.5 bar reduces the system
performance for these wood pallet fires as with larger droplets. The main performance at the early stage (before 20 min). The medium pressure
reason is the ability of larger droplets to penetrate through the fire system with 2.9 bar slows down the initial fire development, although
plume down to the fuel surface where the water droplets cool the fuel the results are quite similar after 20 min. This trend is different to the
surface. comparison of the low and medium pressure systems shown in Fig. 14
(a). It can be seen from Table 3 that the energy released in these two tests
4.1.6. Comparing different FFFSs with same density but different operating is almost the same. However, the possibility of fire spread differs. In test
pressures 106 with the high pressure system at 20.5 bar, approximately 80 % of
In the last section, the influence of water density on individual sys­ the first target was burnt and the second target was charred, whereas in
tems was investigated, however, the operating pressures were varied test 111 with the medium pressure system at 2.9 bar, only the first target
accordingly for a given system. Therefore, the combined effects of water was charred.
density and operating pressure were investigated. As the main task of This indicates that there may exist an optimal operating pressure that

11
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 13. Influence of water density on the system performance (1.8 m/s).

lies between the low pressure and the high pressure system. The physical shown in Fig. 16 (the black line). The fire develops very rapidly after
explanation is that such a system takes advantage of both strong pene­ ignition and reaches the maximum HRR of 4.5 MW at 4.1 min, after
tration of the droplets through the fire plume and efficient cooling which it starts to decay. Note that the maximum HRR is slightly lower
during the transportation towards the fuel surface and when it hits the than the estimate of 5.3 MW presented earlier. The reason could be that
fuel surface. Due to the relatively larger difference between the perfor­ in the previous estimate, the upper limit for the heat release rate per unit
mances of the low and medium pressure systems as shown in Fig. 14(a) area of 195 kW/m2 was used. The results from this free burn test give a
compared to that between the medium and high pressure systems as value of 165 kW/m2 instead.
shown in Fig. 14(b), it is probable that the optimal pressure lies between
the operating pressures for the low and medium pressure systems. This, 4.2.2. Comparison of default FFFSs without top cover
however, does not mean that the default medium pressure system Fig. 16 shows the HRRs for the three default FFFSs in the wood crib
generally performs better than the default low pressure system, as the fires without the top cover. The activation HRR is around 0.4 MW at
conclusion here is made for a given water density. As is valid for all scale approx. 2.8 min. All of the default systems perform very well. The dif­
modelling, a confirmation of these results in a full scale tunnel would be ferences are minor, although the maximum HRR slightly decreases with
beneficial, but this parametric study indicates how to design such full the increasing water density. In all of these tests, the fire HRR at the
scale tests in a tunnel. activation is around 0.4 MW. After activation, it takes around 15 s for
the suppression systems to take full effect, and thereafter the fire decays
4.1.7. Side and centered wood pallet fires rapidly. Clearly, the results are rather different to those for the densely
In one idle wood pallet fire test (test 104), the fire source was placed packed wood pallet fires (Fig. 9). This indicates that an effective sup­
at the centre line of the tunnel (centered fire source). The results are pression can be more easily achieved for uncovered fires with loosely
compared with those from the corresponding test with the default side packed fuels. This should be related to the fact that it is much easier for
placement (test 102) in Fig. 15. The HRRs are similar, but the fire water droplets to reach the inner side of the loosely packed fire load. As
development for the side placed fire source is slightly delayed. This may shown in Tables 3 and in these tests, the energy released is similar (2%–
be due to the lower velocity on the side close to the wall, as it is known 3%) and the fire had no impact on the targets.
that the fire growth becomes slower at a lower velocity [1]. However,
Table 3 shows that in both tests, the energy released is almost the same 4.2.3. Comparison of default FFFSs with top cover
and the fire had no impact on the targets. The results from the wood crib fires with the top cover are shown in
Fig. 17. The activation HRR is around 0.5 MW at approx. 3 min. The
4.2. Wood crib fires maximum HRRs for the default low, medium, and high pressure systems
are 0.9 MW, 1 MW and 2.8 MW, respectively. Overall, the default low
4.2.1. Free-burn test pressure system performs best, and the fire is completely extinguished
The HRR curve from the free burn wood crib fire test at 1.8 m/s is after around 7 min. The fire with the medium pressure system is

12
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 16. Comparison of HRRs for various FFFSs in the wood crib fires without
top cover.

Fig. 14. Influence of operating pressure for different FFFSs with same density.

Fig. 17. Comparison of HRRs for various FFFSs in the wood crib fires with
top cover.

Despite this, the fire had no impact on the targets.

The results shown in Fig. 17 are very different to those for the wood
pallet fires with the top cover (Fig. 10) where the fires were not effec­
tively suppressed using the default systems. This further confirms that an
effective suppression can be more easily achieved for loosely packed
fuels using the low pressure and medium pressure systems, as observed
from the tests without the top cover. But this is not the case for the high
pressure system, which does not effectively suppress the fire. This could
Fig. 15. Comparison of results of side and centered fires.
be related to the weakened dilution effect due to the relatively better
ventilation inside the loosely packed wood crib, compared to the densely
effectively suppressed but not extinguished after the system activation. packed wood pallets.
It continues to burn for around 30 min at a relatively low level (less than
1 MW). For the high pressure system, the fire continues to burn intensely 4.2.4. Side and centered wood crib fires for low pressure system
after the system activation, and reaches a maximum HRR of 2.8 MW at In test 200 with the low pressure system, the wood crib was placed
7.6 min, significantly higher compared to the other systems. This is also along the centerline of the tunnel. The results are compared to test 201
reflected in Table 3, where the proportion of energy released is 7%, 38% where the fire source was placed on the side of the tunnel, as shown in
and 79% for the low, medium and high pressure systems, respectively. Fig. 18. Both fires are effectively suppressed. The maximum HRRs are

13
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 18. Comparison of HRRs for centered and side wood cribs fires.

close to each other, while the duration of the centered fire is longer, thus
resulting in a greater proportion of energy released in the test.
Furthermore, in both tests, there was no fire spread to the targets, as
shown in Table 3.
Overall, the different fire source locations (centered or side) have
limited influences on the performances of the low pressure system, for
both wood pallet fires and wood crib fires. In most of the tests, the fire
source was placed on one side, as can be expected in most fire scenarios
where the driver stops the incident vehicle on its lane.

4.2.5. Influence of activation criterion for low and high pressure systems
Different activation temperatures (Tac = 141 C or 300 C) were used
◦ ◦

in some tests with the low and high pressure systems. In the tests with an Fig. 19. Influence of activation temperature in wood crib fires with top cover
activation temperature of 300 ◦ C, the HRR at the activation is approx­ for low pressure system.
imately 0.8 MW (approx. 3.3 min).
Unfortunately, in test 203 with the activation temperature of 300 ◦ C,
the gas analyzers malfunctioned. The HRR curve of test 201 is shown in
Fig. 19(a), and the measured maximum gas temperatures beneath the
tunnel ceiling from test 201 and test 203 are compared in Fig. 19(b). The
temperatures were measured by thin thermocouples and radiation was
not corrected but its effect was considered small due to the small size of
the thermocouples. From Fig. 19(b) it can be observed that the measured
gas temperatures decrease rapidly after activation and at around 7 min
the temperatures are close to ambient. This indicates that both fires were
effectively suppressed by the low pressure system for both activation
temperatures and is partly confirmed by the HRR curve of test 201
shown in Fig. 19(a). It is known that without a FFFS or prior to its
activation, the heat release rate is proportional to the ceiling excess gas
temperature [37]. By analogy to the data of test 201, a rough estimation
gives a maximum heat release rate of 1.1 MW for test 203, compared to
0.9 MW in test 201. From Table 3, the proportion of energy released in
the tests is insignificant (7% in test 201 and 11% in test 203), and there
was no fire spread to the targets in both tests. To sum up, the use of the
higher activation temperature of 300 ◦ C has an insignificant effect on the
performance of the low pressure system under the conditions tested. The
default low pressure system is able to suppress the covered wood crib
fires efficiently even with a late activation. This, however, may not be
the case for the other systems. Fig. 20. Influence of activation time in wood crib fires with top cover for high
The comparisons of the HRR curves for the high pressure system with pressure system.
the two different activation temperatures are shown in Fig. 20. It should
be noted that in test 206 with the activation temperature of 300 ◦ C, the by default as in test 204, and the amount of heptane used for ignition
fire source was placed at the center line of the tunnel, instead of the side was 1.5 times the default amount. Therefore, they are not completely
comparable. However, from the results for the low pressure system, it

14
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

was observed that the two different fire source location could have rapid decay. The maximum HRR is also slightly lower than the previous
insignificant impacts on the fire suppression performance. Furthermore, estimate of 8.4 MW. This should be related to the use of upper limit for
the slight increase in the amount of ignition fuels in test 206 (1.5 times wood as stated in Section 4.2. If 165 kW/m2 is used for heat release rate
that of test 204) may only have an influence on the initial development per unit area of wood, the heat release rate becomes 7.7 MW, much
and its effect is considered small after the system activation (HRR = 0.8 closer to the measured value of 7.5 MW. The contribution from the
MW). HDPE to the HRR is rather significant. Although the volume portion of
It can be seen in Fig. 20 that the HRR curve of test 206 with the HDPE is only 14%, its contribution to the total HRR is estimated to be as
activation temperature of 300 ◦ C is very close to the free-burn test. The high as about 50 %.
maximum HRRs are 2.8 MW and 4 MW for activation temperatures of
141 ◦ C and 300 ◦ C, respectively. This indicates that in case of a late 4.3.2. Comparison of default FFFSs with top cover
activation for the high pressure system, the suppression becomes much The results for the three default FFFSs with the top cover at 1.8 m/s is
more difficult, resulting in a higher maximum HRR. Moreover, the fire shown in Fig. 21. The activation HRR is around 0.5 MW at approx. 3
growth rate (the slope of the HRR curve) with the late activation is much min. Clearly, the default low pressure system performs much better than
steeper than that of test 204. One reason could be the slightly larger the other systems, and the maximum HRR was 2.4 MW. By contrast, the
ignition source used in test 206. However, the ignition source burnt out HRR curve for the high pressure system is close to a free burn test, and it
at around 3 min (HRR = 1 MW), so its influence on the continued fire reaches the maximum value of 5.9 MW at 6.6 min. The medium pressure
development involving FFFS is considered insignificant. As shown in system does not perform well although it is slightly better than the high
Table 3, the proportion of energy released in these tests is significant pressure system, and the fire reaches the maximum HRR of 4.9 MW at
(79% in test 204 and 98% in test 206). The first target was charred in test 9.3 min. The results for the energy released show less differences. From
206 with the activation temperature of 300 ◦ C but not in test 204. Table 3, the proportion of energy released is 55%, 61%, and 68% for the
Conclusively, delayed system activations usually result in higher default low, medium and high pressure systems, respectively. For the
maximum HRRs and more fuels consumed. Therefore, activation time default low and medium pressure systems, the fires had no impact on the
plays a key role in the system performance. However, the default low targets, however, for the default high pressure system, the targets 1 to 3
pressure system effectively suppressed the wood crib fire even with a were partially burnt and the other three were charred.
late activation (i.e., a higher activation temperature), and thus the in­ In general, the results are consistent with those for the wood crib
fluence of late activation is limited. These results are in line with the fires. Due to the presence of the plastics (HDPE), complete suppression
large scale tests by Ingason et al. in 2013 [14] where the delayed acti­ of the fire becomes more difficult as for plastics a higher water density is
vation time with a low pressure system was successively increased from required for effective suppression, compared to wood [38]. The water
2 min to 4 min, and finally up to 8 min, and the activation heat release densities used for the default medium and high pressure systems seem to
rate increased from 10 MW to 20 MW, but the system did not have any be insufficient. Even for the low pressure system, the fire decays to
noticeable problem in controlling or suppressing the fire. The significant around 0.5 MW at 18 min but it redevelops and reaches a second peak at
differences in the performance of various systems indicate that to around 27 min. From the visual observations during the test, the up­
effectively suppress loosely packed wood crib fires, adequate water stream part of the fuel was almost burnt out at 18 min.
densities are required. To sum up, for both loosely packed wood crib fires, and wood and
HDPE crib fires, the suppression performance seems to be proportional
to the applied water density, and adequate water densities are required
4.3. Wood and HDPE crib fires with the top cover for effective suppression. The lower performance of the high pressure
system is likely to be related to the weakened dilution effect due to the
4.3.1. Free-burn test relatively better ventilation inside the loosely packed fuels, as explained
The results from the free burn wood and HDPE crib fire test are earlier.
shown in Fig. 21 (the black line). The fire develops very rapidly, similar
to the wood crib fire. The HRR reaches its peak value of 7.5 MW at 5.6
4.4. Pool fires without the top cover
min, and then it dropped to around 6 MW for about 4 min before the
4.4.1. Free-burn test
The HRRs from the heptane pool fire tests with various FFFSs as well
as the free burn test is shown in Fig. 22 (the black line). The maximum
HRR in the free burn test is around 0.9 MW before 13 min, but it rises to
about 1.4 MW within a very short period. This phenomenon has also
been observed in previous tests, e.g. Ref. [30]. The reason could be that
at the final stage of a pool fire, the fuel layer becomes thin, and due to
enhanced heat transfer the whole layer of fuel is heated up to boil [30].

4.4.2. Comparison of default FFFSs

A comparison of the HRRs in the tests without the top cover is shown
in Fig. 22, as well as the free burn test data. In all the FFFS tests, at the
activation time (3.5 min), the HRR is around 0.75 MW. Note that due to
a mistake, 35 L of fuel was used in the free burn test 404 while the
volume was 25 L in the other tests. Despite this, the HRR curve from the
free burn test is very close to those from the other tests. This indicates
that the influence of the fuel amount on the initial fire development is
insignificant for such thick fuel pools.
For the low and medium pressure systems, after the activation, the
HRR decreases immediately to around 0.3 MW. Note that in the low
pressure test 401, at 10.5 min, a technical problem occurred with the
Fig. 21. HRRs in wood and plastic crib tests with various FFFSs and the top water supply, but the operating conditions were reestablished at 12.3
cover. Data from the free-burn test is plotted for comparison. min. In other words, during this 1.8 min period, the suppression system

15
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

Fig. 22. HRRs in heptane pool fire tests with various FFFSs. Data from the free-burn test is plotted for comparison.

was not functioning, and no water was discharged. This can be seen from performance of the high pressure system cannot be well explained by
the HRR curve where the HRR rises to a higher level at around 12 min, Heskestad’s theory in an open environment. The reason could be that in
close to the free-burn HRR. But shortly after the water supply is restored, the test with the high pressure system, the influence of ventilation on the
the HRR drops to the same level before the system malfunction. The time trajectories of water droplets cannot be ignored, different to an open
for the suppression system to make full effect (period of the HRR decay) environment. Significant amounts of water droplets may have been
is approximately 30 s. This is much shorter than the time for the fire to carried away by the air flow before reaching the pan, and thus the
restore its burning (period of the HRR increase), which is around 100 s. suppression effect could be much less than the other systems. As stated
This phenomenon is positive from the fire protection point of view. by Heskestad [26], it is much more difficult to suppress fires using
From Fig. 22 it can be observed that all of the FFFSs cannot extin­ fine-spray or misting nozzles in spaces where appreciable oxygen
guish the fire. For the low and medium pressure systems, the HRR in­ depletion cannot take place.
creases slightly with the time after the system activation. This could be By visual observations during the tests, no fuel was spilled out of the
related to the continuous heating of the pan, surroundings, the fuel, and pan. After each of the tests, no fuel but some water was left in the pan.
the water beneath the fuel. The sudden increase in the HRR at the final The calculated energy released in the tests is also consistent with the
stage as can be found in the free burn test, however, does not occur for amounts of fuel used. It can therefore be concluded that all the fuels
the low and medium pressure systems. This should be attributed to the were burnt out even in the presence of FFFS.
fact that the thin fuel layer attached to the bottom of the pan as in the
free burn test was not formed as there was always water left after these 5. Conclusion
tests due to the water sprays continuously discharged to the pan. The
HRR at the relatively stable level is around 0.45 MW. This indicates a Fire suppression performance of various FFFSs, in terms of reducing
reduction of 50% compared to the free burn HRR of 0.9 MW at the early the HRR and the total energy released, and preventing the fire spread to
stage or a reduction of 68% compared to the free burn HRR of 1.4 MW at nearby vehicles, was investigated by carrying out several series of me­
the final stage. The similar performance of the low and medium systems dium scale (1:3) tunnel fire tests. Three FFFSs were selected to represent
could be explained by Heskestad’s theory on extinction of gas and liquid typical systems used in road tunnels, namely a low pressure system, a
pool fires with water spray [26] in an open environment. The medium medium pressure system and a high pressure system, referring to a water
pressure system has a lower flow rate but a higher operating pressure density of 10 mm/min, 6.8 mm/min and 3.7 mm/min, respectively in
and momentum. This results in a smaller effective nozzle diameter and full scale. Other water densities were also tested to explore its effects, as
thus a lower critical water flow rate to extinguish a fire [26]. well as the ventilation velocity and the activation criterion. Various fire
The HRR for the high pressure system is close to the free burn test. In scenarios were considered, i.e., wood pallet fires simulating densely
contrast to the low and medium pressure systems, an increase of around packed fuels, wood crib fires and wood and plastic crib fires simulating
20% after 12 min can be observed, probably due to the lower water loosely packed fuels, and heptane pool fires, with or without the top
density applied for the high pressure system. This increase, however, is cover.
much lower than the increase of about 50 % at the final state of the free Comparisons of the three default systems based on the HRR, the total
burn test. Prior to 12 min, the HRR is about 0.75 MW, but afterwards it energy released and the possibility of fire spread to nearby targets show
rises to about 0.9 MW. This indicates a reduction of 20% compared to that the performance of the default low pressure system is usually the
the free burn HRR of 0.9 MW at the early stage or a reduction of 36% most effective in comparison to others in the densely packed fuel fires
compared to the free burn HRR of 1.4 MW at the final stage. The (without the top cover) and in all the loosely packed fuel fires (with or

16
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

without the top cover). For the densely packed fuel fires with the top 6. Limitations and future work
cover, the three default system performs similarly but the low pressure
system performs better at the early stage (before 12 min), and the me­ This work focuses on comparing the performances of different FFFSs
dium pressure system performs better at the middle stage of the fire in terms of reducing the HRR and the total energy released, and pre­
(between 12 min and 26 min). For the heptane pool fires, the low venting the fire spread to nearby targets. The influence of the tested
pressure and medium pressure systems perform similarly but signifi­ systems on gas temperature, gas concentration and visibility attenuation
cantly better than the high pressure system, and for both systems, the will be covered in a separate paper.
maximum HRR is reduced by 50%–68% of that in the free burn test, Most of the tests were carried out using the three default FFFSs. The
compared to 20%–36% for the high pressure system. Conclusively, in all default water densities for the medium and high pressure systems are
the fire scenarios tested, the default low pressure system performs well close to or slightly greater than the commonly used values. It is expected
and is usually the most effective, and the default high pressure system that this may cause overestimations of the performances of the medium
usually yields results less effective, using HRR, energy released and and high pressure systems in these tests.
possibility of fire spread as measures. The performance of the default The FFFS design is simple, i.e., two rows of ceiling mounted nozzles
medium pressure system usually lies in between them, but in the idle with downward sprays and an equal spacing for all the systems. In
wood pallet fires with the top cover, the performance during a certain practice, a different design may affect the results, e.g., using side
period is slightly better than the low pressure system. spraying nozzles like the ones used in the Runehamar tunnel fire tests [3,
The high pressure system performs very differently in many aspects 14], having a lower tunnel clearance or a lower positioning of the
in comparison to the low and medium pressure systems. These aspects nozzles.
include tunnel ventilation velocity, water density, operating pressure, Nowadays, alternative fuel vehicles are used in almost every type of
and the presence of the top cover. In both the idle wood pallet and wood transportation, e.g., car, bus, HGV and train locomotive, with various
crib fires, the influence of the fire source location on the suppression types of energy carriers: liquid fuels, liquefied fuels, compressed gases,
performance appears to be insignificant, but the top cover plays a key and electricity [39]. Use of FFFS in tunnels could also benefit the
role in fire suppression. Without the top cover, the default low and increasing use of such vehicles, as a FFFS could suppress fires and may
medium pressure systems perform more efficiently than the scenario also cool down gaseous or liquefied tanks and battery packs, thus having
with the top cover. However, the default high pressure system performs great potential to minimize the risks of tank rupture and thermal
slightly better in the wood pallet fire test with the top cover. In the wood runaway of batteries. The research on this topic is clearly needed.
pallet fires, for a given system, a reduction of the longitudinal ventila­
tion velocity, or an increase in the water density improves the perfor­ CRediT authorship contribution statement
mance of the low and medium pressure system but has no or even
negative effects on the performance of the high pressure system. This Ying Zhen Li: Conceptualization, Data curation, Formal analysis,
indicates that the suppression mechanisms of the high pressure system Funding acquisition, Investigation, Methodology, Project administra­
are different to the others, which does not rely on forceful penetration tion, Resources, Writing – original draft. Haukur Ingason: Funding
through the fire plume and direct surface cooling, but more on, e.g., gas acquisition, Methodology, Resources, Writing – review & editing.
cooling and dilution. Magnus Arvidson: Methodology, Writing – review & editing. Michael
Comparisons of performance of different systems with same water Försth: Funding acquisition, Methodology, Writing – review & editing.
densities but different operating pressures in the idle wood pallet fires
with the top cover show that an increase of the operating pressure from
the low pressure system level to the medium system level improves the Declaration of competing interest
fire suppression performance. But it is on the opposite side when the
operating pressure is increased from the medium pressure system level The authors declare the following financial interests/personal re­
to the high pressure system level. This indicates that, for the covered lationships which may be considered as potential competing interests:
fires, there appears to be an optimal pressure between the low and high Y.Z. Li and H. Ingason are serving as guest editors for a special issue.
pressure systems, and probably between the low and medium pressure However, they are not involved in the editorial review or the decision to
system levels due to the corresponding larger differences. In other publish this article.
words, a slight increase in the operating pressure of a low pressure Other authors declare that they have no known competing financial
system whilst applying the same water density may probably improve interests or personal relationships that could have appeared to influence
the system performance in the densely packed fires with the top cover. the work reported in this paper.
The system may perform equally well in the other scenarios due to the
same water density applied. This provides a possible way to further Data availability
improve the performance of the low pressure system.
In terms of preventing the fire spread to nearby targets, the low Data will be made available on request.
pressure system demonstrated good efficiency, and only in one test at
the lowest water density of 4 mm/min was the first target charred. The Acknowledgement
medium pressure system also performs relatively well in preventing fire
spread, with the first target being charred in only two tests. By contrast, The work was financially supported by the Swedish Research Council
the high pressure system was not as efficient in preventing the fire Formas (2019-00521), which is gratefully acknowledged. The authors
spread. In the majority of the tests with the high pressure system, one or would also like to express their gratitude to Prof. Patrick van Hees at
more targets were partially burnt. Note that the results of fire spread Lund University for his valuable input in the test planning process.
should be considered as the combined effect of the amount of water Thanks also to our colleague Joel Blom, and other technicians at RISE for
delivered to the targets, and the heat exposed to the targets, which de­ the great assistance in conducting the tests, and the Södra Älvsborg’s
pends on the HRR, the ventilation velocity, and the gas cooling and Rescue Service for the support on site.
radiation attenuation effects of the water droplets. Nevertheless, there
appears to be a trend that a higher water density and larger droplet size
References
facilitates the prevention of fire spread to downstream objects or vehi­
cles, and the dependence of fire spread on the system pressure seems to [1] H. Ingason, Y.Z. Li, A. Lönnermark, Tunnel Fire Dynamics, Springer, New York,
be a secondary factor in these tests. 2015.

17
Y.Z. Li et al. Fire Safety Journal 146 (2024) 104141

[2] U. Lundström, Development of the Swedish road tunnel safety concept, in: [20] E. Blanchard, P. Boulet, P. Fromy, S. Desanghere, P. Carlotti, J.P. Vantelon, J.
Proceedings from the Tenth International Symposium on Tunnel Safety and P. Garo, Experimental and numerical study of the interaction between water mist
Security, 2023, pp. 45–55. Stavanger, Norway. and fire in an intermediate test tunnel, Fire Technol. 50 (2014) 565–587.
[3] H. Ingason, Y.Z. Li, Large scale tunnel fire tests with different types of large droplet [21] Y.Z. Li, H. Ingason, Parametric study of design fires for tunnels with water-based
fixed fire fighting systems, Fire Saf. J. 107 (2019) 29–43. fire suppression systems, Fire Saf. J. 120 (2021) 103107.
[4] Y.Z. Li, H. Ingason, Overview of research on fire safety in underground road and [22] Y.Z. Li, H. Ingason, Influence of fire suppression on combustion products in tunnel
railway tunnels, Tunn. Undergr. Space Technol. 81 (2018) 568–589. fires, Fire Saf. J. 97 (2018) 96–110.
[5] T. Lemaire, Y. Kenyon, Large Scale Fire Tests in the Second Benelux Tunnel, vol. [23] H. Ingason, Y.Z. Li, M. Arvidson, L. Jiang, Fire tests with automatic sprinklers in an
42, Fire Technology, 2006, pp. 329–350. intermediate scale tunnel, Fire Saf. J. 129 (2022) 103567.
[6] Development of new innovative technologies, UPTUN Work Package 2, 2006. [24] NFPA 13 - Standard for the Installation of Sprinkler Systems, National Fire
[7] M. Tuomisaari, Full Scale fire testing for raod tunnel applications - evaluation of Protection Association, 2023.
acceptable fire protection performance, in: A. Lönnermark, H. Ingason (Eds.), Third [25] NFPA 750 - Standard on Water Mist Fire Protection Systems, National Fire
International Symposium on Tunnel Safety and Security, 2008, pp. 181–193. Protection Association, 2023.
Stockholm. [26] G. Heskestad, Extinction of gas and liquid pool fires with water spray, Fire Saf. J.
[8] J.R. Mawhinney, Evaluating the performance of water mist systems in road 38 (2003) 301–317.
tunnels, in: IV Congreso Bienal Apci Ingenieria de Pci, 2007. Madrid. [27] M. Arvidson, Large-Scale Water Spray and Water Mist Fire Suppression System
[9] A.D. Lemaire, V.J.A. Meeussen, Effects of Water Mist on Real Large Tunnel Fires: Tests for the Protection of Ro–Ro Cargo Decks on Ships, vol. 50, Fire Technology,
Experimental Determination of BLEVE-Risk and Tenability during Growth and 2014, pp. 589–610.
Suppression, Efectis Nederland BV, 2008. [28] Y.Z. Li, H. Ingason, Scaling of wood pallet fires, Fire Saf. J. 88 (2017) 96–103.
[10] S. Kratzmeir, M. Lakkonen, Road Tunnel Protection by water mist systems - [29] P.A. Croce, Y. Xin, Scale modeling of quasi-steady wood crib fires in enclosures,
implementation of full scale fire test results into a real project, in: Third Fire Saf. J. 40 (2005) 245–266.
International Symposium on Tunnel Safety and Security, SP Technical Research [30] Y.Z. Li, C.G. Fan, H. Ingason, A. Lönnermark, J. Ji, Effect of cross section and
Institute of Sweden, Stockholm, 2008. ventilation on heat release rates in tunnel fires, Tunn. Undergr. Space Technol. 51
[11] SOLIT, Engineering Guidance for a comprehensive Evaluation of Tunnels with (2016) 414–423.
fixed fire fighting systems Scientific report of the SOLIT2 research project, in: [31] H. Ingason, Y.Z. Li, A. Lönnermark, Runehamar tunnel fire tests, Fire Saf. J. 71
Annex 3 - Engineering Guidance for Fixed Fire Fighting Systems in Tunnels, 2012. (2015) 134–149.
The SOLIT2 consor-tium. [32] H. Ingason, A. Lönnermark, Heat release rates from heavy goods vehicle trailers in
[12] M.K. Cheong, W.O. Cheong, K.W. Leong, A.D. Lemaire, L.M. Noordijk, Heat Release tunnels, Fire Saf. J. 40 (2005) 646–668.
Rates of Heavy Goods Vehicle Fire in Tunnels with Fire Suppression System, Fire [33] M. Janssens, W.J. Parker, Oxygen consumption calorimetry, in: V. Babrauskas, T.
Technology, 2013. J. Grayson (Eds.), Heat Release in Fires, E & FN Spon, London, UK, 1992,
[13] M.K. Cheong, W.O. Cheong, K.W. Leong, A.D. Lemaire, L.M. Noordijk, F. Tarada, pp. 31–59.
Heat release rates of heavy goods vehicle fires in tunnels, in: 15th International [34] B.Z. Dlugogorski, J.R. Mawhinney, V.H. Duc, The measurement of heat release
Symposium on Aerodynamics, Ventilation & Fire in Tunnels, BHR Group, rates by oxygen consumption calorimetry in fires under suppression, in: Fire Safety
Barcelona, Spain, 2013. Science - Proceedings of the Fourth International Symposium, Ottawa, Canada:
[14] H. Ingason, Y.Z. Li, G. Appel, U. Lundström, C. Becker, Large scale tunnel fire tests International Association for Fire Safety Science (IAFSS), 1994.
with large droplet water-based fixed fire fighting system, Fire Technol. 52 (5) [35] H. Ingason, Y.Z. Li, G. Appel, U. Lundström, C. Becker, Large Scale Tunnel Fire
(2016) 1539–1558. Tests with Large Droplet Water-Based Fixed Fire Fighting System, vol. 52, Fire
[15] C. Hue-Pei, H. San-Ping, C. Chao-Shi, C. Shen-Wen, Performance of a spray system Technology, 2016, pp. 1539–1558.
in a full-scale tunnel fire test, Tunn. Undergr. Space Technol. 67 (2017) 167–174. [36] G. Rein, R. Carvel, J.L. Torero, Approximate trajectories of droplets from water
[16] J. Li, Y.F. Li, Q. Bi, Y. Li, W.K. Chow, C.H. Cheng, C.W. To, C.L. Chow, Performance mist suppression systems in tunnels, in: Proceedings from the Third International
evaluation on fixed water-based firefighting system in suppressing large fire in Symposium on Tunnel Safety and Security, SP Technical Research Institute of
urban tunnels, Tunn. Undergr. Space Technol. 84 (2019) 56–69. Sweden, Stockholm, Sweden, 2008.
[17] L. Jiang, R.H. Mostad, T. Li, K.S. Arsava, Experimental study of the performance of [37] Y.Z. Li, B. Lei, H. Ingason, The maximum temperature of buoyancy-driven smoke
a water mist system on fires in a full-scale tunnel, in: Proceedings from the Tenth flow beneath the ceiling in tunnel fires, Fire Saf. J. 46 (4) (2011) 204–210.
International Symposium on Tunnel Safety and Security, RISE Research Institutes [38] H.-Z. Yu, J.L. Lee, H.-C. Kung, Suppression of rack-storage fires by water, in: Fire
of Sweden, Stavanger, 2023. Safety Science - Proceedings of the Fourth International Symposium, 1994.
[18] H. Ingason, Model scale tunnel tests with water spray, Fire Saf. J. 43 (7) (2008) [39] Y.Z. Li, Study of fire and explosion hazards of alternative fuel vehicles in tunnels,
512–528. Fire Saf. J. 110 (2019) 102871.
[19] Y.Z. Li, H. Ingason, Model scale tunnel fire tests with automatic sprinkler, Fire Saf.
J. 61 (2013) 298–313.

18