Tunnelling and Underground Space Technology 95 (2020) 103116

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Optimal distance between jet fans used to extinguish metropolitan tunnel T

fires: A case study using fire dynamic simulator modeling
⁎,1 ⁎
Miaocheng Wenga,b,c,d, , Imad Obadia,1, Fei Wanga, Fang Liua,b,c,d, Chunhui Liaoa,b,c,d,
a
Faculty of Urban Construction and Environmental Engineering, Chongqing University, Chongqing 400045, PR China
b
Key Laboratory of Three Gorges Reservoir Region’s Eco-Environment, Ministry of Education, Chongqing 400045, PR China
c
National Centre for International Research of Low-carbon and Green Building, Chongqing 400045, PR China
d
Joint International Research Laboratory of Green Building and Built Environment, Ministry of Education, Chongqing 400045, PR China

A R T I C LE I N FO A B S T R A C T

Keywords: Several disastrous tunnel fires have occurred in recent decades; safe evacuation, minimization of damage, and
Extra-long tunnel reliable ventilation have received a great deal of attention. Here, we use the Fire Dynamic Simulator (FDS)
FDS software to optimize jet fan spacing in an extra-long tunnel: the Zhong Liangshan tunnel of Metropolitan Line 1
Jet fan in Chongqing, China. The model tunnels were 500, 550, and 600 m in length, with cross-sectional areas of
Temperature
30.72 m2; the jet fans were elevated 5.60 m from the floor at different spacings (100, 150, and 200 m). A 5-MW
Tunnel fire
Velocity
fire burned at 200, 175, and 150 m from the tunnel entrance. We simulated five scenarios using FDS software
and compared airflow velocity and temperature curves. We also briefly investigated the effect of the length of an
extra-long tunnel. The numerical results revealed that the air temperature always remained well below the 60 °C
that the human body can withstand during evacuation necessitated by a fire; the air velocity remained nearly
constant. We built a 1:15-scale model to validate the simulation; the simulation and experimental curves were in
good agreement.

1. Introduction transverse mechanical ventilation is seldom used today. Tunnel venti-

lation fans (TVFs) should be used to introduce fresh air in one direction
Multiple considerations are involved when seeking to prevent fires if the length of the tunnel is greater than 305 m, as directed by National
in tunnels; both tunnel design and fire-fighting measures save lives. In Fire Protection Association (NFPA) code 130-2017 (National Fire
the past, several accidental tunnel fires have caused great loss of life Protection Association (NFPA), 2017). However, to ensure longitudinal
and serious damage to the tunnels. Table 1 lists the numbers of ca- ventilation, jet fans are extremely useful in unidirectional traffic tun-
sualties; 85% were attributable to smoke inhalation (Weng et al., 2015). nels; the fans are normally placed on the roof of the tunnel to control
The longer the tunnel, the greater the risk of poor access for fire- smoke longitudinally. The air velocity must be sufficiently high to
fighters and inefficient evacuation of passengers. The safety of both prevent smoke spread in the flow direction. Thus, jet fans are used to
humans and the environment are at stake. Damage to tunnels has major drive air under the roof sufficiently rapidly to allow the smoke to reach
economic and environmental consequences. Ventilation is critical for an exit portal, greatly reducing the heat of the fire. When a fire erupts in
tunnel safety; it is essential to remove smoke and reduce the tempera- a tunnel, such a system should control the spread of smoke from the fire
ture during evacuation (Colella et al., 2011; Eftekharian et al., 2014). and maintain a layer of clean air at the bottom of the tunnel, to allow
Modern tunnels usually feature several ventilation systems, varying in for smooth evacuation and ensure the absence of a back layer caused by
characteristics and configurations and by traffic direction (Li and Chow, smoke movement against the dominant flow. Back-layering is con-
2003). Both transverse and longitudinal ventilation are common. sidered to occur if the flow velocity is higher than a threshold, termed
Transverse mechanical ventilation allows independent regulation of the the critical velocity (Oka and Atkinson, 1995; Thomas, 1958; Vauquelin
air supply and exhaust along the entire tunnel length via two separate and Wu, 2006). In terms of the air supply, the choice of a ‘critical ve-
plenums, and is appropriate for long tunnels with bidirectional traffic, locity’ or a ‘low velocity’ supply remains controversial; the latter is the
but is associated with high capital and operational costs; thus, minimum velocity at which smoke is prevented from spreading against

⁎
Corresponding authors at: Faculty of Urban Construction and Environmental Engineering, Chongqing University, Chongqing 400045, PR China.
E-mail addresses: [email protected] (M. Weng), [email protected] (C. Liao).
1
Both authors contributed equally to this manuscript and should be considered co-first authors.

https://doi.org/10.1016/j.tust.2019.103116
Received 15 September 2018; Received in revised form 11 July 2019; Accepted 10 September 2019
0886-7798/ © 2019 Elsevier Ltd. All rights reserved.
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Table 1
Casualties of tunnel fires.
Case No. of casualties Tunnel Length Country Year Reference

1 289 Baku subway 2.2 km Azerbaijan 1995 (Beard and Carvel, 2005; Hedenfalk, 1998)
2 198 Daegu subway 25.9 km South Korea 2003 (Hong, 2004)
3 155 Kaprun funicular tunnel 3.3 km Austria 2000 (Beard and Carvel, 2005; Meyer, 2003)
4 39 Mont-Blanc tunnel 11.6 km France/Italy 1999 (Abraham and Dérobert, 2003; Vuilleumier et al., 2002)
5 12 Tauern tunnel 6.4 km Austria 1999 (Leitner, 2001)
6 11 Gotthard tunnel 16.9 km Switzerland 2001 (Martin et al., 2005)
7 9 Viamala tunnel 750 m Switzerland 2006 (Beard and Carvel, 2005; Beard, 2009)
8 5 Gleinalm tunnel 8.2 km Austria 2001 (Beard and Carvel, 2005)
9 5 Eiksund tunnel 7.7 km Norway 2009 (Beard and Carvel, 2005)
10 3 Burnley tunnel 3.4 km Australia 2007 (Maevski, 2011)
11 2 Martino tunnel 4.8 km Italy 2005 (Beard and Carvel, 2005)
12 2 Fréjus tunnel 12.9 km France/Italy 2005 (Beard and Carvel, 2005)
13 2 Rotsethhorn tunnel 1.2 km Norway 2000 (Beard and Carvel, 2005)
14 1 Eidsvoll tunnel 1.2 km Norway 2006 (Beard and Carvel, 2005)
15 1 Baregg tunnel 1.1 km Switzerland 2004 (Beard and Carvel, 2005)

Fig. 1. (a) A photograph of the Zhong Liangshan Tunnel entrance. (b) Schematic of the dimensions of the tunnel cross-section showing the relationship between the
real tunnel and the model built in the FDS.

2
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 2. Schematic of tunnel models showing the spacings between the jet fans. (a) The distance from the portal to the fire source varied. (b) The distance from the
portal to the fire source was constant (not to scale).

the longitudinal tunnel ventilation flow. Typically, maximum critical or two-dimensional mathematical models, or via full-scale experiments.
velocities are 2.5–3.0 m/s (Oka and Atkinson, 1995; Thomas, 1958; However, full-scale tunnel fire tests are difficult, costly, and oper-
Vauquelin and Wu, 2006; Wu and Bakar, 2000). ationally constraining. Thus, numerical computational fluid dynamics
Previously, tunnel longitudinal ventilation was designed using one- (CFD) modeling is increasingly being used (Bailey et al., 2002; Grant

3
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Table 2 close to the wall, but was reduced when it passed through the tunnel
The scenarios modeled by the distance between the fire and the open portal. axis.
Case Model Jet fan Fire position Distance from the open Se et al. (2012) investigated three important aspects of fires in
length distance portal to the fire source longitudinally ventilated tunnels that are directly relevant here: (1)
their tendencies (orientations); (2) their nature; (3) and their effect on
Case 1 500 m 100 m X = 200 m Varied
airflow velocity. Also, the CFD method was used to explore temperature
Case 2 500 m 150 m X = 175 m
Case 3 500 m 200 m X = 150 m
distributions at various positions when the fans were activated. The
Case 4 550 m 150 m X = 200 m Fixed results revealed that: (1) changes in fire source dimensions or orienta-
Case 5 600 m 200 m X = 200 m tion did not affect the airflow structure; (2) the upstream velocity was
(surprisingly) reduced; and (3) when active fans were spaced
at ≥ 200 m, the correlation between the distance of the fans from the
and Drysdale, 1997; Hwang and Edwards, 2005; Li and Chow, 2003; fire source and the upstream velocity was very marked.
Wu and Bakar, 2000). Such modeling, which is relatively simple, yields Few studies have explored the performance of jet fans in extra-long
highly accurate predictions of very complex velocity and temperature tunnels. Previous works using fans at different spacings and angles fo-
patterns in the event of fire, even if the geometrical characteristics of cused on those installed on the tunnel roof or side walls. Hence, it was
space and time boundaries are complex. The CFD method is ideal for necessary to further investigate optimal jet fan spacing. Also, only a few
the design and enhancement of ventilation systems. Myrvang and reports using the Fire Dynamic Simulator (FDS) have specifically
Khawaja (2018) used CFD to analyze how air develops inside a tunnel, evaluated the performance of tunnel ventilation systems on such jet
from the inlet to the outlet, for a given number of fans. The placement fans. Thus, we conducted a detailed numerical investigation of air ve-
of the fans was approximately the same as in a real tunnel. The entire locity and temperature distribution profiles associated with differently
model was 800 m in length, and the fans were placed at 250, 400, and positioned jet fans in an extra-long tunnel. The Zhong Liangshan tunnel
550 m from the inlet. Furthermore, they built a small experimental of Metro Line 1 of Chongqing city, China was chosen as a case study of
model representing a miniature model of the tunnel. The results of their the optimal distance between jet fans. The velocity and temperature
experiments with the miniature model and CFD simulations were in distributions were analyzed via FDS simulation.
good agreement with real-case scenarios. Their results also indicated
that the air velocity inside the tunnel does not increase linearly with the 2. Simulation
number of active ventilation stages. Betta et al. (2010) performed a
numerical study on an alternative jet fan, termed the Banana Jet (Witt 2.1. Fire dynamics simulator
and Schütze, 2006); the fan features an inlet/outlet region inclined to
the pitch angle (α ≠ 0°) . The fluid performance of the fan was compared Today, many jets are tested using numerical CFD techniques and
to those of conventional ventilation systems in various scenarios (e.g., a models that consider multiple variables (Król and Król, 2018). The FDS
traffic jam and a total lack of vehicles). CFD was used to simulate is a predictive tool used for computational analysis, specifically to
ventilation in a unidirectional tunnel 800 m in length; longitudinal compare the performance of ventilation systems and analyze fluid flow
ventilation featured four jet fans installed 5.60 m above the floor. The and other complex thermal phenomena. It allows engineers to analyze
first fans were located 100 m distant from the entrance and exit portals, and simulate flow problems of various complexity, and the results are
and the other fans were spaced at 200 m. In both cases, use of an op- rapid, reliable, and accurate. Physical modeling has been expanded to
timal pitch angle (αopt ) improved safety performance compared to that analysis of air flow, air velocity, heat transfer, and other features. Re-
of conventional systems (α = 0°). cent expansions in computational power for modeling via CFD and, in
Martegani et al. (1997) studied the effects of separation, pitch, and particular, large eddy simulation (LES) via codes incorporated in the
swirl on the performance of single and coupled jet fans in a tunnel, in an FDS (as updated by the National Institute of Standards and Technology;
effort to optimize performance by varying the pitch angle. An experi- NIST) (McGrattan et al., 2013a), have rendered the FDS very reliable
mental 1:10-scale tunnel was used to explore performance. Momentum and user-friendly. The FDS has many applications in fire safety (Zhong
tests were performed using seven eccentricities (from the tunnel wall to et al., 2017) and is widely used to model relevant physical phenomena
the tunnel center). The jets were tested at inclinations of 0, 2, 4, and 6°, and fire hazards (Grigoriadis et al., 2018; Ryder et al., 2004). The FDS
at all eccentricities. The increase in jet momentum created by coupling Technical Reference Manual (McGrattan et al., 2013b) includes a brief
two fans enhanced efficiency only when the two jet axes were maxi- overview of the validation work performed to date. However, any
mally separated. A recirculation zone was evident downstream of each overview obviously lacks detailed information about verification. Fur-
fan; the efficiency decreased when the jet contribution to tunnel air ther FDS verification has been performed since the publication of the
velocity was large. Technical Reference Manual, because the FDS has undergone numerous
Wang et al. (2010) examined the behaviors of jet fans, and the ef- verifications and enhancements in the time since it was first developed
fects of pressure distribution on such behavior in curved tunnels and on in 2000. Many examples of validation have been presented at scientific
tunnel ventilation, in various tunnel cross-sections. As the jet fans were conferences and in published works, which are collated at http://fire.
moved further away from the ceiling, the airflow from the fans pro- nist.gov/fds/ (Ji et al., 2015; Zhong et al., 2013).
pagated faster. When fans were close to other fans or next to the tunnel Various studies on the use of jet fans to ventilate tunnels and nu-
walls, the increased pressure affected jet development over 90–120 m. merical models of such ventilation have appeared, and many authors
Zhang et al. (2017) studied how to optimize the arrangement of jet fans have compared FDS data to experimental measurements (Chen et al.,
near a cross passage at the bottom of a shaft in an extra-long tunnel to 2013; Hu et al., 2009, 2014; Hwang and Edwards, 2005; Kim et al.,
help push out polluted air. Their results indicated that it is reasonable 2008; Roh et al., 2007; Tilley et al., 2011; VK et al., 2007; Yan et al.,
to install a fan 5 m away from the axis between the shaft and the ex- 2009). LES has been incorporated into the FDS to investigate the effects
haust passage. The layout for the jet fan was underneath the airflow, of temperature and longitudinal forced velocity on tunnel fires, and has
which was close to the cross passage. The jet should not be too far away validated the FDS approach.
from the cross passage, and it is better to install it in as high a place as Lee and Ryou (2006) used the FDS to explore the effects of aspect
possible. Mutama and Hall (1996) sought to optimize jet fan perfor- ratio on smoke movement, in terms of both the velocity and tempera-
mance using a wind tunnel. The initial pressure drop in the tunnel in- ture distribution of tunnel fires, and compared the findings with ex-
creased over a considerable period of time when the jet fan was moved perimental data. When the aspect ratio increased, the temperature near
toward the wall. Recirculation was observed when the jet fan lay very the fire decreased, and the velocity distribution was approximately 3%

4
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 3. Schematic of features over distances of 500, 600, and 700 m, used to compare the effects of the physical domain (not to scale).

Table 3
Mesh types and sizes.
Case Zone type Grid size/m Total grid number Length of grid/m

Case 1 A 0.10 115,200 15

B 0.20 139,200 145
81,600 85
134,400 140
C 0.20 129,600 45
115,200 40
100,800 35

Case 2 A 0.10 115,200 15

B 0.20 86,400 90
91,200 95
129,600 135
C 0.20 86,400 30
100,800 35

Case 3 A 0.10 115,200 15

B 0.20 38,400 40
43,200 45
48,000 50
C 0.20 43,200 15
57,600 20

Case 4 A 0.10 115,200 15

B 0.20 115,200 120
129,600 135
C 0.20 86,400 30
100,800 35
115,200 40
129,600 45

Case 5 A 0.10 115,200 15

B 0.20 76,800 80
86,400 90
91,200 95
100,800 105
C 0.20 86,400 30
100,800 35

2.2. Physical model

2.2.1. The tunnel

Fig. 1(a) presents the tunnel used in this case study (Zhong Liang-
shan): it is a straight, two-lane one-directional tunnel 4.3 km in length
Fig. 4. Effects of the physical length at distances of 500, 600, and 700 m. (a)
and 60.43 m2 in cross-section; the maximum cross-sectional height and
Velocity distribution curves. (b) Temperature distribution curves.
width, respectively, are 7.40 and 9.60 m. Fig. 1(b) presents a schematic
of the tunnel cross-section.
greater than the experimental result. Hu et al. (2006) calculated the
highest near-fire smoke temperatures under the tunnel roof using the
FDS, and found that the results were in good agreement with the ex- 2.2.2. The calculation model
perimental data. Se et al. (2012) used CFD to explore the behaviors and Considerable computational time would be required if the model
nature of fires in longitudinally ventilated tunnels, and their effects on were to reflect the shape of the real tunnel. From a practical viewpoint,
airflow velocity and temperature distribution at various positions the simpler the mesh (i.e., a cubic mesh), the more accurate the data.
where jet fans were located. All studies reported good agreement be- The real tunnel and the model built in the FDS program are related in
tween FDS-predicted and experimental data. terms of their area, A, and the hydraulic diameter, H̄. The hydraulic
diameter of the tunnel, H̄, is defined as the ratio of four times the cross-
sectional area to the tunnel wetted perimeter (H̄=4A/χ). We simplified

5
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 5. The mesh divisions and zone types of the different domains (not to scale).

Table 4
Scale factors for the model tunnel.
Parameter Scaling Equation no.

Geometry (m) Xm/Xf = Lm/Lf = λL (3)

Temperature (K) Tm = Tf (4)
Velocity (m s−1) vm = vf = (Lm = Lf)1/2 = λL1/2 (5)
Volume flow (m3 s−1) Vm/Vf = (Lm/Lf)5/2 = λL5/2 (6)
Heat release rate (HRR) (W) Qm/Qf = (Lm/Lf)5/2 = λL5/2 (7)

the FDS simulation using rectangular cross-sections, which enabled us

to take the cost of the numerical computation into consideration when
investigating the effect of varying the hydraulic diameter and the area,
which are similar to the real values shown in Fig. 1(b), similar to the
work done by Weng et al. (2015). Symmetry was considered; thus, we
modeled the right half of the tunnel (the tunnel cross-section is sym-
metrical), as described below. The physical domains were of (arbitrary) Fig. 6. Photograph of the 1:15-scale model tunnel used in the experiment.
lengths 500, 550, and 600 m for cases 1, 2, and 3; case 4; and case 5,
respectively. However, we showed that our assumptions were valid by simulate an extra-long tunnel ventilated (along the x-coordinate) with
comparing the effects of physical domain length (Section 2.2.3) in a jet fans at different positions. The width and height of the tunnel cross-
one-directional extra-long tunnel equipped with three jet fans placed section were 6.4 and 4.8 m (y- and z co-ordinates), respectively. The
(in cases 1–3) at various distances from the fire and (in cases 4 and 5) at three jet fans were installed 5.60 m above the floor at different spacings
fixed distances from the fire; the distance required for complete stabi- (100, 150, and 200 m) that are often employed by tunnel designers, and
lization of the axial static and dynamic pressures was approximately were manufacturer-rated to afford a volumetric flow of 12.70 m3/s at a
90–120 m (Wang et al., 2010). discharge flow velocity of 40.90 m/s. The fan ducts were 2.0 m in
As shown in Fig. 2(a) and (b) and Table 2, we used five cases to length, and were simulated as basic flow devices with pressure

6
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 7. Schematic of the experimental layout. (a) Model dimensions with jet fan distances and the thermocouple arrangement. (b) Cross-section (not to scale).

increases across them (reflecting the outflow air velocity); the cross- 2.4. Boundary conditions and other simulation details.
sections of the inlets and outlets were 0.31 m2. A 5-MW fire was placed
at 200, 175, or 150 m (cases 1, 2, and 3, respectively) from the tunnel FDS version 6.5 was used to predict buoyancy-driven flow when the
entrance; or at 200 m (cases 4 and 5; a fixed distance). The obstruction boundary conditions at the tunnel entrance and exit portals were set to
caused by the fire was arbitrarily assumed to lie equidistant between “Open.” Both tunnel portals were assumed to be at zero pressure; the
the first and second jet fan; the fixed fuel source area and burner surface inside and outside temperatures were set to 20.0 °C; and the reaction
thickness were 2.0 m × 2.0 m and 0.2 m, respectively. was that of heptane. All walls were assumed to be smooth. The “Inert”
default FDS boundary condition featured a fixed temperature (“TMP”)
and an emissivity of 0.9. Radiative and convective heat transfer were
2.2.3. Influence of length on the physical model calculated, and the tunnel ceiling and floor roughnesses were set to
As explained above (Section 2.2.2), we did not model the entire 0.01 m. The fire source surface of the FDS burner was a rectangular
tunnel because of the large computational cost and complexity of the block of 2 m ((length) × 2 m (width) × 0.2 m (height), placed 200, 175,
full CFD calculation. We verified the effect of the real physical length and 150 m from the tunnel entrance in cases 1, 2 and 3, and 200 m
only. We retained the model boundary conditions and grid size distant in cases 4 and 5. In previous studies, the HRR of a metropolitan
(Sections 2.3 and 2.4) but extended the lengths to 500, 600, and 700 m tunnel fire was usually set to 5–10 MW (Weng et al., 2015). The air
(Fig. 3). We explored whether the reduction in computational time velocity was set to a value typical for a jet fan; the volume flow rate was
compromised our ability to predict the full CFD accurately in terms of 12.70 m3/s. The inlets and outlets were part of a heating, ventilation,
length. and air-conditioning (HVAC) system; an FDS surface was used to re-
Fig. 4(a) and (b) show the effects of physical length on the tem- present vents connecting the HVAC system to the computational mesh.
perature and velocity regions. The horizontal velocity and temperature All simulations used the LES turbulence model. The computational time
curves of all three cases were very similar. Thus, the presumptions of was adequate, at 300 s (Musto and Rotondo, 2014). To measure tem-
the physical model afford an adequate balance between the reliability perature and velocity distributions within the tunnel, thermocouples
of the results and the computational cost. and slices were installed every 1.0 m; the distance between the floor
and the center line of the jet fans was 5.6 m. FDS simulations were
executed using a dual-processor PC with a 2.4-GHz 24-core CPU and
2.3. Meshes
64 GB of RAM.

In FDS simulations, the most sensitive numerical parameter is the

3. Experimental setup
mesh (grid cell) size, which affects the accuracy, computational time,
and costs of tunnel fire simulations (Weng et al., 2015).
3.1. Scaling law
In the FDS user guide (McGrattan et al., 2013a), the non-dimen-
sional expression D∗/ δx is used to assess mesh resolution, where δx is the
The scaling law used was the commonly employed Froude Law;
nominal size grid (in m); the characteristic length of D∗is expressed as
scaling laws have been well-reviewed (Ingason, 2007; Ingason and Li,
follows:
2010). The tunnel was constructed at a geometric 1:15-scale. The FDS
Q 2 and experimental data are compared in Section 4.3. The scaling cor-
D∗ = ( )3
ρ∞ cp T∞ g 0.5 (1) relations between the model and a real tunnel are shown in Table 4
(where L is the length scale, the M index refers to the model scale, and
where Q is the fire heat release rate (HRR) (W), ρ∞is the ambient the F index to the full scale). A 1:15-reduced model tunnel was built
density (kg m−3), cp is the specific heat capacity (J kg−1 K−1), T∞ is the applying the Froude scaling method. This ensured geometrical, kine-
ambient temperature (K), and g is the gravitational acceleration con- matic, and dynamic similitude. The reduction scale preserved the
stant (m s−2). Interestingly, the user guide suggests that D∗/ δx should lie layout of the real tunnel and the turbulence features thereof. Buoyancy
in the range 4–16. Thus, a very fine mesh for a 5-MW fire should be forces dominated the turbulent flow after Froude number modeling,
between 0.114 and 0.456 m. FDS divides the model space into a large because the Reynolds number was large.
number of rectangular grids and calculates the temperature, velocity,
and other factors for each grid. Model accuracy is exquisitely dependent 3.2. Model tunnel and experimental installation
on grid resolution; smaller grids afford more accurate modeling out-
comes (Ryder et al., 2004). We built an experimental tunnel at a scale ratio of 1:15. The tunnel
Mesh sizes of 0.10 and 0.20 m were used to enhance simulation was 11 m in length, 0.32 m in width, and 0.48 m in height, and com-
efficiency while minimizing computational time (Li and Chow, 2003). prised 11 tunnel units each 1 m in length. The relationship between the
The domain was discretized into three zones: zone A included the jet model and a real tunnel can be computed using Eqs. (3)–(7). The tunnel
fans, zone B included the upper slices of the modeled domains, and zone was made of a welded stainless steel plate and other non-combustible
C included the lower slices. The mesh divisions and the zone types of materials (to prevent thermal damage). The floor, ceiling, and one of
the three domains are listed in Table 3 and shown in Fig. 5. the walls were made of 8-mm-thick fireproof board; the other (front)

7
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 8. Velocity distribution contours at various distances between the fire and the open portal. (a) Case 1 (100 m), (b) Case 2 (150 m), (c) Case 3 (200 m). Velocity
distribution contours at a constant distance from the fire to the open portal. (d) Case 4 (150 m); (e) Case 5 (200 m).

wall was made of 8-mm-thick fire-resistant window glaze to allow ob- 3.5 m from the tunnel entrance; liquefied petroleum gas served as fuel
servation of experimental phenomena. The joints of the fireproof board because this affords a steady HRR.
were bonded with tinfoil to ensure that the setup was airtight (Fig. 6). To measure ceiling temperatures, thermocouples were fixed 0.07 m
Several three-jet fans of diameter 0.07 m and length 0.14 m were below the ceiling along the centerline. In total, 36 hot-wire thermo-
installed in the model, with horizontal spacings of about 4.0 m, located couples 1 mm in diameter were placed at 0.30-m intervals and con-
0.11 m under the ceiling (Fig. 7). Fan velocity was measured before the nected to an Agilent 34980A data acquisition computer. Signals gen-
fire test using an anemometer: the axial velocity was 2.20 m/s. The erated by the output were transmitted to a PC via an RS-232 cable and
initial temperature was set to 30 °C. The simulation was run for up to saved using the SWEMA program. A schematic side-view of the 1:15-
300 s and the last 50 s of data were averaged. The fire source was placed scale model tunnel showing the thermocouple arrangement is shown in

8
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

tunnel depends on the difference in airspeed at the outlet compared to

the inlet. Fig. 9(a) and (b) present the velocity curves of simulations
along the tunnel axis; the data are the averages obtained on the center
jet fan line (z = 5.6 m, y = 2.4 m) for cases 1, 2, and 3 (100, 150, and
200 m jet spacing) respectively, when the distance from the fire to the
open portal varied. The velocity trends remained nearly constant, as
shown in Fig. 9 (a).
Fig. 9(b) compares the simulated velocity curves along the tunnel
axis of the model with the averages on the center lines of the jet fans
(z = 5.6 m, y = 2.4 m) for cases 1, 4, and 5 (100, 150, and 200 m jet
spacings) respectively, with the distance from the fire to the open portal
held constant. The velocity curves were all rather similar, as shown in
Fig. 9(b), because the airflow was stabilized. However, the velocity
curves of cases 3 and 5 are economically optimal. The fewer the number
of jet fans, the lower the power consumption and the costs of pur-
chasing, installing, and maintaining the fans, as shown in the literature
and as found in real-world situations. Increasing fan spacing increases
the cross-sectional effects of jet action compared to what is seen when
fans are closely spaced (Wang et al., 2010). Fan horizontal spacing
affects velocity profiles, indicating jet development and pulse exchange.
The significant increase in flow rate was attributable to amplification of
the momentum generated by the fans; the flow rate was gradually re-
duced by stress friction until the flow approached the next fan.

4.2. Temperature profiles

Fig. 10(a, b, c) and (d, e) show the temperature contours when fans
were spaced differently and a fire erupted. Different colors represent
various temperatures between T = 20–150 °C. Regions of high tem-
perature were restricted to volumes of only a few cubic meters directly
above the fires. Free-burning fires were associated with tempera-
tures ≥ 130 °C in their ascending plumes until the plumes reached the
ceiling. At this point, the plumes predominantly flowed horizontally as
the jet fans started up, promptly decreasing the temperature to 45 °C in
most cases, and even to < 30 °C near the jet fan inlets. Fig. 10(c) and
(e) present cases 3 and 5, respectively, showing that regions remote
from the fire are at much lower temperatures than those near the fire, as
shown in Fig. 10(a, b, and d) for cases 1, 2, and 4, respectively. In other
Fig. 9. Velocity distribution curves along the × co-ordinate direction
words, temperatures near the fire are significantly higher than those of
(z = 5.6 m, y = 2.4 m). (a) Various distances between the fire and the open
areas far from the fire. Fig. 11(a) compares the simulated temperatures
portal. (b) Constant distances from the fire to the open portal.
along the modeled tunnel axis; the data are averages derived from the
central lines of the jet fans (z = 5.6 m, y = 2.4 m) for cases 1, 2, and 3
Fig. 7. (100-, 150-, and 200-m fan spacings), respectively; the distances from
the fire to the open portal varied. All temperature curves are essentially
4. Results and discussion identical; the temperature never exceeded the that the human body can
withstand (Xu et al., 2017; Zhang et al., 2018).
We considered geometry, mesh analysis, physical modeling, Fig. 11(b) compares the simulated temperatures along the model
boundary conditions, and tunnel length when obtaining numerical tunnel axis with the average temperatures on the center line of the jet
data, to better compare velocity and temperature distributions. The fans (z = 5.6 m, y = 2.4 m) for cases 1, 4, and 5 (100-, 150-, and 200-m
contour range was set as shown in Figs. 8 and 10, and was sufficiently fan spacings), respectively, when the distance from the fire to the open
high to capture deterministic phenomena. The velocity and tempera- portal was constant. Notably, the temperature curve of case 3 did not
ture contours varied among the zones. The curves obtained are shown exceed the limit of 60 °C, as shown in Fig. 11(b). However, the differ-
in Figs. 9 and 11. ences were negligible for the other cases. The temperature curves did
not change significantly and met all safety requirements. The air tem-
4.1. Velocity profiles peratures would allow rescue workers to approach the fire and evacuate
passengers.
Fig. 8(a–c) and (d) and (e) show, respectively, the velocity contours
and air distributions at different jet fan spacings. Different colors refer 4.3. Comparison of experimental and numerical data
to velocities of 2–32 m/s. The velocities near jet outlet nozzles were
high, and then decreased to about 10.0 m/s. The outlet velocities A numerical simulation requires experimental validation. To verify
ranged widely. Moreover, extensive air stagnation at the inlet nearest the performance of the model, we developed another small model using
the portal was about 8 m/s, caused by the blowing of air into the tunnel FDS and matched it to an experimental model tunnel with the same
was evident in cases 3 and 5; even lower velocities (about 6 m/s) were cross-section, length, scale, and boundary conditions. Fig. 12(a) and (b)
evident in cases 1, 2, and 4. The air velocities near the jet inlets were compare the simulated and experimental data on the horizontal tem-
slightly higher when the fans were spaced at 200 m (cases 3 and 5), perature distribution and the velocity generated by the jet fan (sym-
reducing the temperature and thus saving energy. The airflow in the metrically with other fans, i.e., the distance between them is equal. As

9
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 10. Temperature distribution contours at various distances between the fire and the open portal. (a) Case 1 (100 m), (b) Case 2 (150 m), (c) Case 3 (200 m).
Temperature distribution contours at constant distances between the fire and the open portal. (d) Case 4 (150 m); (e) Case 5 (200 m).

shown in Fig. 12(a), the predicted and experimental upstream and temperature associated with a clearly defined situation after definition
downstream temperatures were in excellent agreement, but some de- of an appropriate grid resolution and acceptable boundary conditions.
viations in temperatures upstream of the fire were apparent. The up- Fig. 12(b) compares the results of FDS modeling and the small-scale
stream temperature in the small-scale tunnel was close to the ambient velocity distribution curves generated by the jet fans from the outlet
temperature. There are two possible explanations for this: first, slight stream (in the absence of fire); the data were in good agreement. In
inconsistencies in gas temperature may be in play; second, some smoke general, the numerical and model-scale test data on temperature and
may spread through the insulation covering the tunnel walls and roof. velocity distribution were consistent. Finally, the comparison of nu-
In contrast, the FDS curve commences at about 30 °C and then rises merical predictions to experimental data demonstrated that the nu-
slightly, perhaps because heat accumulates at these positions, raising merical scenario employed was appropriate.
the local temperature. Thus, the FDS results accurately predict the

10
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

Fig. 12. Comparison of the simulation results with those of the experiment. (a)
Temperature distribution curves. (b) Velocity distribution curves generated by
Fig. 11. Temperature distribution curves along the x co-ordinate direction jet fan.
(z = 5.6 m, y = 2.4 m). (a) Various distances between the fire and the open
portal. (b) Constant distance between the fire and the open portal. Foundation of China (Grant No. 51608076), the Chongqing Science and
Technology Commission (Grant No. cstc2016shmszx30016), the
5. Conclusion Chongqing Science and Technology Planning Project (Grant No. 2015-
1-34), the Fundamental Research Fund for Central Universities (Grant
To ensure a high level of safety and safe evacuation when a fire No. 106112015CDJXY210008), the Graduate Scientific Research and
breaks out in a long tunnel, we used numerical models to study the Innovation Foundation of Chongqing, China (Grant No. CYB16006),
effects of jet fans at different spacings in both the temperature and and the 111 Project No. B13041.
velocity domains. The distances tested were 100, 150, and 200 m. We The author expresses sincere thanks to Dr. Long Xing Yu and Dr.
compared five scenarios in terms of velocity and temperature beha- Shengzhong Zhao for their valuable assistance with the FDS modeling.
viors. Numerically, the velocity distribution was approximately con-
stant for all cases tested, and the temperatures all remained below the References
level that the human body can bear (60 °C). Firefighters will be safe.
Economic considerations suggest that cases 3 and 5 (200 m) should be Abraham, O., Dérobert, X., 2003. Non-destructive testing of fired tunnel walls: the Mont-
preferred. We experimentally tested this finding and verified the tem- Blanc Tunnel case study. NDT & E Int. 36, 411–418.
Bailey, J., Forney, G.P., Tatem, P., Jones, W.W., 2002. Development and validation of
perature and velocity curves using data from a reduced-scale model. corridor flow submodel for CFAST. J. Fire. Prot. Eng. 12, 139–161.
The numerical results agreed relatively well with the experimental re- Beard, A., Carvel, R., 2005. The Handbook of Tunnel Fire Safety. Thomas Telford London.
sults. Every metropolitan tunnel has unique characteristics, so various Beard, A.N., 2009. Fire safety in tunnels. Fire Saf. J. 44, 276–278.
Betta, V., Cascetta, F., Musto, M., Rotondo, G., 2010. Fluid dynamic performances of
factors must be considered when spacing jet fans in a very long tunnel. traditional and alternative jet fans in tunnel longitudinal ventilation systems. Tunn.
Our results may indicate that variable fan spacing is preferable; we will Undergr. Space Technol. 25, 415–422.
explore this topic in the near future. Chen, L., Hu, L., Tang, W., Yi, L., 2013. Studies on buoyancy driven two-directional smoke
flow layering length with combination of point extraction and longitudinal ventila-
tion in tunnel fires. Fire Saf. J. 59, 94–101.
Colella, F., Rein, G., Borchiellini, R., Torero, J.L., 2011. A novel multiscale methodology
Acknowledgment for simulating tunnel ventilation flows during fires. Fire Technol. 47, 221–253.
Eftekharian, E., Dastan, A., Abouali, O., Meigolinedjad, J., Ahmadi, G., 2014. A numerical
investigation into the performance of two types of jet fans in ventilation of an urban
This work was supported by the National Natural Science

11
M. Weng, et al. Tunnelling and Underground Space Technology 95 (2020) 103116

tunnel under traffic jam condition. Tunn. Undergr. Space Technol. 44, 56–67. Myrvang, T., Khawaja, H., 2018. Validation of air ventilation in tunnels, using experi-
Grant, G.B., Drysdale, D., 1997. Estimating heat release rates from large-scale tunnel fires. ments and computational fluid dynamics.
Fire Saf. Sci. 5, 1213–1224. National Fire Protection Association (NFPA), 2017. NFPA 130 Standard for Fixed
Grigoriadis, D.G., Geurts, B.J., Kuerten, H., Fröhlich, J., Armenio, V., 2018. Direct and Guideway Transit and Passenger Rail Systems, 2017 Edition. National Fire Protection
Large-Eddy Simulation X. Springer. Association, I Batterymarch Park, P.O. Box 9101.
Hedenfalk, J., 1998. Lessons from the Baku subway fire. In: 3rd International Conference Oka, Y., Atkinson, G.T., 1995. Control of smoke flow in tunnel fires. Fire Saf. J. 25,
on Safety in Road & Rail Tunnels. Nice World Road Association, Fire & Smoke in Road 305–322.
and Rail Tunnels. PIARC (Permanent International Association of Road Congress) Roh, J.S., Ryou, H.S., Kim, D.H., Jung, W.S., Jang, Y.J., 2007. Critical velocity and
Committee on Road Tunnels. burning rate in pool fire during longitudinal ventilation. Tunn. Undergr. Space
Hong, W.-H., 2004. The progress and controlling situation of Daegu Subway fire disaster. Technol. 22, 262–271.
Fire Saf. Sci. 6 s-5–1. Ryder, N.L., Sutula, J.A., Schemel, C.F., Hamer, A.J., Van Brunt, V., 2004. Consequence
Hu, L., Chen, L., Tang, W., 2014. A global model on temperature profile of buoyant ceiling modeling using the fire dynamics simulator. J. Hazard. Mater. 115, 149–154.
gas flow in a channel with combining mass and heat loss due to ceiling extraction and Se, C.M., Lee, E.W., Lai, A.C., 2012. Impact of location of jet fan on airflow structure in
longitudinal forced air flow. Int. J. Heat Mass Transf. 79, 885–892. tunnel fire. Tunn. Undergr. Space Technol. 27, 30–40.
Hu, L., Huo, R., Peng, W., Chow, W., Yang, R., 2006. On the maximum smoke tem- Thomas, P., 1958. The movement of buoyant fluid against a stream and the venting of
perature under the ceiling in tunnel fires. Tunn. Undergr. Space Technol. 21, underground fires. Fire. Saf. Sci. 351 -1–1.
650–655. Tilley, N., Rauwoens, P., Merci, B., 2011. Verification of the accuracy of CFD simulations
Hu, L., Huo, R., Yang, D., 2009. Large eddy simulation of fire-induced buoyancy driven in small-scale tunnel and atrium fire configurations. Fire Saf. J. 46, 186–193.
plume dispersion in an urban street canyon under perpendicular wind flow. J. Vauquelin, O., Wu, Y., 2006. Influence of tunnel width on longitudinal smoke control.
Hazard. Mater. 166, 394–406. Fire Saf. J. 41, 420–426.
Hwang, C., Edwards, J., 2005. The critical ventilation velocity in tunnel fires—a com- Vk, S., Tam, L., Lao, S., Choi, H., 2007. Numerical Simulation of Atrium Fire using Two
puter simulation. Fire Saf. J. 40, 213–244. CFD Tools, Computational Methods in Engineering & Science: Proceedings of
Ingason, H., 2007. Model scale railcar fire tests. Fire Saf. J. 42, 271–282. Enhancement and Promotion of Computational Methods in Engineering and Science
Ingason, H., Li, Y.Z., 2010. Model scale tunnel fire tests with longitudinal ventilation. Fire X“. Springer Science & Business Media, Sanya, China.
Saf. J. 45, 371–384. Vuilleumier, F., Weatherill, A., Crausaz, B., 2002. Safety aspects of railway and road
Ji, J., Wan, H., Li, K., Han, J., Sun, J., 2015. A numerical study on upstream maximum tunnel: example of the Lötschberg railway tunnel and Mont-Blanc road tunnel. Tunn.
temperature in inclined urban road tunnel fires. Int. J. Heat Mass Transf. 88, Undergr. Space Technol. 17, 153–158.
516–526. Wang, F., Wang, M., He, S., Zhang, J., Deng, Y., 2010. Computational study of effects of
Kim, E., Woycheese, J.P., Dembsey, N.A., 2008. Fire dynamics simulator (version 4.0) jet fans on the ventilation of a highway curved tunnel. Tunn. Undergr. Space Technol.
simulation for tunnel fire scenarios with forced, transient, longitudinal ventilation 25, 382–390.
flows. Fire Technol. 44, 137–166. Weng, M.-C., Lu, X.-L., Liu, F., Shi, X.-P., Yu, L.-X., 2015. Prediction of backlayering
Król, A., Król, M., 2018. Study on numerical modeling of jet fans. Tunn. Undergr. Space length and critical velocity in metro tunnel fires. Tunn. Undergr. Space Technol. 47,
Technol. 73, 222–235. 64–72.
Lee, S.R., Ryou, H.S., 2006. A numerical study on smoke movement in longitudinal Witt, K., Schütze, J., 2006. Bend it like a Banana. Proceedings BHR Group, 12th
ventilation tunnel fires for different aspect ratio. Build. Environ. 41, 719–725. International Symposium on Aerodynamics and Ventilation of Vehicle Tunnels.
Leitner, A., 2001. The fire catastrophe in the Tauern Tunnel: experience and conclusions Wu, Y., Bakar, M.A., 2000. Control of smoke flow in tunnel fires using longitudinal
for the Austrian guidelines. Tunn. Undergr. Space Technol. 16, 217–223. ventilation systems–a study of the critical velocity. Fire Saf. J. 35, 363–390.
Li, J.S., Chow, W., 2003. Numerical studies on performance evaluation of tunnel venti- Xu, Z.S., You, W., Kong, J., Cao, H.H., Zhou, C., 2017. A study of fire smoke spreading and
lation safety systems. Tunn. Undergr. Space Technol. 18, 435–452. control in emergency rescue stations of extra-long railway tunnels. J. Loss Prev.
Maevski, I.Y., 2011. Design fires in road tunnels. Transp. Res. Board. Process Ind. 49, 155–161.
Martegani, A., Pavesi, G., Barbetta, C., 1997. The influence of separation, inclination, and Yan, T., MingHeng, S., YanFeng, G., JiaPeng, H., 2009. Full-scale experimental study on
swirl on single and coupled jet fans installation efficiency. In: BHR Group Conference smoke flow in natural ventilation road tunnel fires with shafts. Tunn. Undergr. Space
Series Publication. Mechanical Engineering Publications Limited, pp. 43–56. Technol. 24, 627–633.
Martin, J.-C., Delemont, O., Calisti, C., 2005. Tunnel fire investigation II: The St Gotthard Zhang, H., Sun, J., Lin, F., Wang, L., 2017. Optimization on energy saving ventilation of
Tunnel fire, 24 October 2001. gallery-type combined construction shaft exhaust in extra long tunnel. Procedia Eng.
McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., Overholt, K., 205, 1777–1784.
2013a. Fire dynamics simulator, user’s guide. NIST special publication 1019, 20. Zhang, L., Yan, Z.-Z., Li, Z.-H., Wang, X.-M., Han, X.-F., Jiang, J.-C., 2018. Study on the
McGrattan, K.B., McDermott, R.J., Weinschenk, C.G., Forney, G.P., 2013b. Fire dynamics effect of the jet speed of air curtain on smoke control in tunnel. Procedia Eng. 211,
simulator, technical reference guide. 1026–1033.
Meyer, H.J., 2003. The Kaprun cable car fire disaster—aspects of forensic organisation Zhong, W., Duanmu, W., Wang, T., Liang, T., 2017. A study of the critical velocity of
following a mass fatality with 155 victims. Forensic Sci. Int. 138, 1–7. smoke bifurcation flow in tunnel with longitudinal ventilation. Fire Technol. 53,
Musto, M., Rotondo, G., 2014. Numerical comparison of performance between traditional 873–891.
and alternative jet fans in tiled tunnel in emergency ventilation. Tunn. Undergr. Zhong, W., Fan, C., Ji, J., Yang, J., 2013. Influence of longitudinal wind on natural
Space Technol. 42, 52–58. ventilation with vertical shaft in a road tunnel fire. Int. J. Heat Mass Transf. 57,
Mutama, K., Hall, A., 1996. The experimental investigation of jet fan aerodynamics using 671–678.
wind tunnel modeling. J. Fluids Eng. 118, 322–328.

12