Downloaded from https://iranpaper.

ir
https://www.tarjomano.com/order

Tunnelling and Underground Space Technology 81 (2018) 494–505

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Full-scale immersed tunnel fire experimental research on smoke flow T

patterns
⁎
P. Xua,b, , S.P. Jiangc, R.J. Xinga, J.Q. Tana
a
School of Civil Engineering, Chongqing Jiaotong University, Chongqing 400074, China
b
State Key Laboratory Cultivation Base for Bridge and Tunnel Engineering in Mountain Areas, Chongqing Jiaotong University, Chongqing 400074, China
c
China Merchants Chongqing Communications Technology Research & Design Institute Co., LTD, Chongqing 400067, China

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

Keywords: To research the smoke flow patterns and their effects on personnel evacuation during a tunnel fire, a full-scale
Tunnel fire model of an immersed 150 m tunnel was constructed, taking the Hong Kong–Zhuhai–Macao Bridge immersed
Smoke layer height tunnel in China as the prototype. A series of fire tests, including pool fires and vehicle fires, were developed to
Smoke spread velocity simulate a real fire scenario. The characteristic parameters describing the smoke flow were measured, including
CO concentration
smoke layer height, smoke spread velocity, and CO concentration. Their development behaviors and main in-
Full-scale experiment
fluence factors were explored, and finally smoke flow patterns were examined. The results showed that smoke
stratification appeared, which was measured by smoke layer height. The smoke layer height distribution along
the tunnel was found to obey a quartic polynomial trend (goodness of fit > 95%). From this, the safety distances
were predicted to be 20 m in windy conditions and 80–90 m in still (no wind) conditions. Thus, the longitudinal
velocity plays a leading role. Around the fire source or in still conditions, the buoyancy of the smoke is the
critical factor driving smoke flow; the influence of other factors is no more than ± 20%. However, other con-
ditions depended on the ventilation state. CO concentration decreased sharply with increasing distance from the
fire source. At a distance of 60 m, CO concentration decayed by 50%.

1. Introduction and thermophysical properties during the burning process in a tunnel is

The Statistical bulletin of transportation industry development in 2016
(17 April 2017) stated that by the end of 2016 the number of national
highway tunnels in China, including 815 extra-long tunnels
(L > 3000 m, total length 3,622,700 m) and 3520 long tunnels
(1000 m ≤ L < 3000 m, total length 6,045,500 m), increased to
15,181 tunnels with a total length of 14,039,700 m. The number of
tunnels increased by 1175 with an increased length of 1,355,800 m
compared to the previous year. Tunnel fires are one of the common
potential risks which restrain the development of tunnels on a longer
and larger scale, when more tunnels bring convenience to transporta-
tion. Fires in the Mont Blanc tunnel, the Tauern tunnel, and the St.
Gotthard tunnel were typical cases of serious fire disasters around the
world (Vuilleumier et al., 2002). In recent years in China, fires in the
New Qidaoliang tunnel in Gansu and the Yanhou tunnel in Shanxi both
caused serious explosions and casualties (Yang, 2014). The accident
investigation and analysis indicated that the main cause of casualties
was toxic substances released by the incomplete combustion of fuels
(Babrauskas et al., 1998, 1992). Thus, research on smoke flow patterns
the basis and key point for solving tunnel safety problems (Yang et al.,
2012; Hu et al., 2007; Lee and Ryou, 2006; Colella et al., 2009; Li et al.,
2011). Previous studies focusing on the four main characteristic para-
meters of smoke flow have produced a great deal of information, in-
cluding (1) the maximum temperature under the tunnel ceiling
(Kurioka et al., 2003; Li et al., 2011; Liu et al., 2016; Tang et al., 2017);
in particular the maximum smoke temperature model proposed by
Kurioka et al. (2003); (2) longitudinal temperature decay (Kashef et al.,
2013; Gong et al., 2016; Ji et al., 2016; Liu et al., 2017; Meng et al.,
2017); in particular the effects of tunnel geometry and fire location on
ceiling temperature decay studied by Kashef et al. (2013); (3) smoke
back-layering length (Hu et al., 2008; Li et al., 2010; Gannouni and
Maad, 2016; Fan and Yang, 2017); with a non-dimensional model to
predict the back-layering flow length proposed by Li et al. (2010); (4)
critical ventilation velocity (Tsai et al., 2010; Li et al., 2010; Tang et al.,
2013; Chow et al., 2015); in which the critical velocity was measured in
small-scale experiments and predicted by numerical simulations. In this
paper, specific values (smoke spread velocity, smoke layer height, and
CO concentration) were investigated in a realistic tunnel-fire scenario.

⁎
Corresponding author at: Chongqing Jiaotong University, No. 66 Xuefu Rd., Nan’an Dist., Chongqing 400074, China.
E-mail address: xu_pai@126.com (P. Xu).

https://doi.org/10.1016/j.tust.2018.08.009
Received 10 October 2017; Received in revised form 8 June 2018; Accepted 16 August 2018
0886-7798/ © 2018 Elsevier Ltd. All rights reserved.
 Downloaded from https://iranpaper.ir
P. Xu et al.
It provided data which is the basis for further research on tunnel smoke
flow behavior. Compared with small-scale models and numerical si-
mulation models, a full-scale tunnel model is a more realistic reflection
of an actual tunnel fire and the smoke flow state, because errors due to
reduced scale and establishing numerical models vanish in full-scale
experiments (Tian et al., 2017; Liu et al., 2016; Hu et al., 2013; Ji et al.,
2012; Liu et al., 2017; Tang et al., 2017; Hu et al., 2013; Ji et al., 2015).
In Europe, the United States, and other developed countries, a series of
full-scale fire experiments have been carried out. The larger tests in-
cluded the EUREKA 499 test programs (1990–1993) conducted in
abandoned tunnels in Norway, Germany, and Finland, with an em-
phasis on smoke temperature, heat conduction, smoke flow rate, smoke
concentration and rescue strategy. Tests on smoke control effects were
conducted under different ventilation modes in the Memorial Tunnel in
America (2001) (Buvik, 2006). The UPTUN project in Europe con-
ducted in abandoned Runehamar tunnels in Norway emphasized the
smoke spread behavior of burning container trucks, and the selection of
fire-fighting devices (2003) (Ingason and Lonnermark, 2005;
Lonnermark and Ingason, 2005; Ingason and Li, 2015; Ingason et al.,
2012a, 2012b). The team at the University of Science and Technology
of China conducted a series of tests in the Yangzong tunnel, Da-
fengyakou tunnel, and Yuanjiang No. 1 tunnel in Yunnan to investigate
the temperature damping behavior of the smoke in the longitudinal
direction. The prediction model regarding critical velocity and smoke
back-layer length were discussed (2006) (Hu, 2006; Hu et al., 2005).
Fire tests, fire source calibration, and alarm system tests were carried
out in a tunnel in Shanghai, China, to obtain the key parameters, in-
cluding smoke temperature, smoke concentration, general visibility and
the visibility of exit signs, and accuracy of fire detection. The structure
and fire-fighting devices were identical to those in existing maximum
shield tunnels (2009).
The immersed-tube tunnel of the Hong Kong – Zhuhai – Macao
Bridge is the longest immersed highway tunnel and the only deeply
buried tunnel in the world, with a length of 5664 m. It is installed 40 m
underwater, and is designed for a 120-year service life. Such an im-
portant and complex project, there is a big difference between existing
full-scale tests and our study regarding the structural form and escape
environment of the tunnel and how to ensure personnel safety in the
event of a fire. The above questions were mainly solved, with the focus
being on the criteria for personnel evacuation and provides reference
material for the effective management of smoke and evacuation in the
initial stage of a tunnel fire. To approach this, the scenario of a real
tunnel fire was simulated by constructing a full-scale tunnel model in
which both pool fires and burning vehicles were the fire sources. Based
on this, the smoke spread process, the thermal parameters of the smoke,
the behavior of the smoke flow, and its main factors were determined. A
prediction of behavior of the thermal parameters for each working
condition was proposed.
2. Experimental setup and conditions
2.1. Full-scale experimental tunnel
A full-scale experimental tunnel (150 m × 14.5 m × 7.1 m) com-
prising a tunnel structure and smoke extraction duct (Xu, 2014) was
built to simulate the immersed tube tunnel of the Hong Kong – Zhuhai –
Macao Bridge tunnel complex. Schematic diagrams are shown in Figs. 1
and 2. The model cross-section dimensions are shown in Table 1.
To determine smoke movement behavior most accurately in the test,
adjustable resistance grilles were installed at both ends of the tunnel
(Fig. 3) to compensate for the short experimental tunnel length by in-
creasing local resistance to air flow in the model. The experimental
results thus approximate the actual situation. The resistance grilles
were composed of soft curtains 19 cm wide, based on equivalent friction
theory, as follows:
495
https://www.tarjomano.com/order
Tunnelling and Underground Space Technology 81 (2018) 494–505
Fig. 1. Cross-section schematic of the experimental model of an immersed tube
tunnel.
λ m Lm
ξg = ,
dm (1)
where ξg is the loss coefficient of the resistance grilles; λm is the friction
resistance coefficient; Lm is the length; and dm is the equivalent dia-
meter of the experimental tunnel. The design parameters of the re-
sistance grilles are listed in Table 2.
2.2. Experimental apparatus and instrumentation
2.2.1. Smoke layer height measurement
The most intuitive and accurate way to obtain the smoke layer
height, which is one of the most important parameters describing the
behavior of the smoke flow, is a visual method. To approach this, the
location system for observing smoke spread was coordinate lines
painted on the side wall of the tunnel using a bright yellow fire-re-
tardant coating. As shown in Fig. 4, vertical “1 m lines” 10 cm wide
were drawn every 1 m, and “0.5 m lines” 5 cm wide were drawn
midway between them.
To reduce visual errors, and to check the result, the smoke flow
pattern was recorded at set time intervals by cameras installed in the
side wall of the tunnel.
2.2.2. Smoke spread velocity measurement
A longitudinal airflow was generated by a jet fan located at the
centerline of the tunnel. The specification of the jet fan was chosen to
be Φ1000, 30 kW. A portable anemograph was used to measure the air
speed generated by the jet fan and the speed at which the smoke spread.
To ensure that combustion had reached the stable stage, measurements
were conducted 3 min after the tests began. Fig. 5 shows the arrange-
ment of monitoring points and fan position when the fire source was at
different locations. The arrangement in the actual situation is shown in
Fig. 6, where the monitoring points are denoted by solid dots and the
hollow dots indicate the position of the jet fan.
2.2.3. Measurement of CO concentration in smoke
As shown in Fig. 7, smoke analyzers were installed in the experi-
mental tunnel to detect CO movement in the personnel evacuation
environment. To study the CO concentration longitudinally distributed
at the inner and lower edges of the smoke layer, the smoke analyzers, in
a arrangement consistent with the anemograph, were installed at dif-
ferent heights of the same cross-section, as well as at different long-
itudinal locations.
2.2.4. Fuel weighing system
The fire heat release rate was determined by the weight-loss
method, whose accuracy depends on the accurate determination of the
fuel quality. Two types of weighbridge were selected in accordance
with the quality of the fuel, as shown in Table 3.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

Fig. 2. Full-scale immersed tunnel model.

Table 1 Table 2
Model tunnel cross-section dimensions. Design parameters of resistance grilles.
L1 L2 L3 L4 L5 L6 L7 Area Perimeter Simulated length Loss coefficient Effective area Horizontal spacing of
(m) ratio (%) soft curtains (m)
2
0.7 m 11.3 m 2.8 m 3.5 m 7.1 m 4.4 m 14.5 m 97.8 m 41.2 m
1000 2.411 81 0.81
1500 3.715 51 0.20
2000 5.018 47 0.17
2.2.5. Radiant heat flux measurement system
3000 6.322 40 0.13
Total heat flux sensors were set up 8 m, 18 m, and 48 m from the fire
source and 1.5 m above the ground to monitor the total heat flux during
tests. Captec radiation heat flux sensors were selected Table 4). The following vehicle collision. The fire heat release rate (HRR) is basically
sensors comprised a thin foil heat-flow sensor that indicates the heat constant during the stable stage of combustion, so that a liquid fire
flux when the temperature changes, and a circulating water-cooled heat source may be regarded as stable. Gasoline was selected as the fire
receiver. The sensors were connected by copper wire to the acquisition source in the preliminary test, as in most similar tests in China and
module and the data was transmitted by an RS485/232 data acquisition overseas. Gasoline fires quickly reach the stable combustion stage,
system (Windmill Software Ltd, Manchester, England) and recorded by which is sustained for a period of time and produces a moderate
computer. amount of black smoke and CO after it is ignited. This meets the re-
The errors of each measurement system are shown in Table 5. quirements of the smoke test for fuel. Abandoned cars and buses were
used as fire sources to simulate real tunnel fire scenarios as closely as
possible. It was considered unnecessary to repeat the experiment due to
2.3. Fire source calibration and experimental conditions
the high cost of the burning object and because repetition of this kind of
experiment was not significant enough to warrant it.
Road tunnel fires are mainly caused by vehicles and their load.
The HRR is usually measured by either the weight-loss method or
Vehicle combustion is mainly due to liquid fuels, so a liquid fire source
radiation method. Both methods were adopted in this experiment for
and vehicle fire source were selected for the tests. Gasoline and diesel
corroboration. The methods are as follows:
oil fire sources simulated oil surface burning caused by fuel leakage

Fig. 3. Experimental tunnel with resistance grilles.

496
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

Fig. 4. Coordinate lines in test tunnel.

• Weight loss method:

Q̇ = ṁ ′ ′A × χ × ΔHg ,

where Q̇ is the fire heat release rate (MW); ṁ ′ ′ is the combustion rate
(kg/m2 s); A is the combustion area (m2); ΔHg is the complete com-
bustion calorific value (diesel oil 42 MJ/kg; gasoline 45 MJ/kg); and χ
is the combustion efficiency (0.3–0.9).
The HRR is obtained by measuring ṁ ′ ′ by a fuel weighing system.

• Radiation method:
qstr = C1·PRHR / R2 ,

where qstr is the unit radiation flux R meters from the fire source (W/
m2); PRHR is the total heat release rate: PRHR = χr ·Q̇ (W); χr is the ra-
diation efficiency (0.2–0.6); and C1 is a proportionality constant:
4πR2qstr
C1 = 1 4π , then Q̇ = χr
.
The HRR is obtained by measuring qstr by a radiant heat flux mea- Fig. 6. Measured velocity field.
surement system.
The results of fire source calibration tests are shown in Table 6.
The experiments were carried out in the model tunnel in the fol-
lowing ambient conditions: temperature 20 ± 3 °C; pressure

Fig. 5. Schematic diagram of arrangement of velocity monitoring points when the fire source was located (a) midway along the tunnel; (b) 15 m from the tunnel
portal.

497
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

Fig. 7. Arrangement of smoke analyzers.

Table 3
Fuel weighing system.
Number Range/accuracy Dimensions Quantity Picture

1 3 t/1 kg 5 m × 2.5 m 1

2 500 kg/100 g 1m×1m 1

Table 4
Instrumentation in radiant heat flux measurement system.
Name Range Specifications Number Picture

Captec radiation heat flux sensor ± 500 kW/m 2

HS-30B 4

Table 5 behavior (Xu, 2016). A summary of the test conditions is given in

System errors. Table 7.
Measurement Fuel Smoke Velocity CO concentration Radiation
system weight layer
height 3. Experimental results and discussion

Errors (%) ±5 ±5 ± 2.5 ± 3–8 ±5 3.1. General observations

3.1.1. Smoke stratification

101 ± 5 kPa; and humidity 25 ± 15%. The fuel type, the HRR and the
Fires in a tunnel are usually accompanied by large quantities of
longitudinal velocity were the main factors affecting smoke spread. The
smoke due to the relatively narrow dimensions. The comparative lack
representative conditions are helpful in analyzing the smoke spread
of oxygen results in incomplete combustion and produces a great

Table 6
Results of calibration tests.
Fuel type Gasoline Gasoline Diesel oil Car Bus

Oil pan dimensions (m × m × m) 1.0 × 1.0 × 0.1 1.5 × 1.5 × 0.1 1.5 × 1.5 × 0.1 – –
Amount of fuel (L) 60 60 60 – –
Burning time (s) 7′27 8′55″ 13′00″ ≈25′ ≈40′
HRR (MW) 2.2–2.8 4.9–5.6 3–5 ≈5 ≈15

498
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

Table 7
Test working conditions.
Test Fuel type HRR Oil pan dimensions Air velocity (m/
(MW) (m × m × m) s)

T1 Gasoline 25 3.0 × 3.0 × 0.1 ≈0

T2 Gasoline 25 3.0 × 3.0 × 0.1 3.0
T3 Gasoline 50 3.0 × 4.5 × 0.1 ≈0
T4 Car ≈5 – 1.5
T5 Bus ≈15 – ≈0
T6 Gasoline 5 1.5 × 1.5 × 0.1 ≈0
T7 Gasoline 5 1.5 × 1.5 × 0.1 1.5
T8 Diesel oil 5 1.5 × 1.5 × 0.1 ≈0 ceiling jet
T9 Diesel oil 5 1.5 × 1.5 × 0.1 1.5

(a)

smoke layer boundary

hydraulic jump
(a)

(b)

front boundary lower boundary

(b)
Fig. 8. Smoke stratification in pool fire test: (a) 50 MW fire; (b) 50 MW fire
where stratification is damaged.
number of solid particles. The density of high-temperature smoke is less
than the surrounding cold air, forming updrafts. The smoke rises to the
tunnel ceiling, and then moves along the vault away from the fire
source as the cold air below it pours into the fire source. Consequently,
the hot smoke layer and the cold air layer form separate streams of air,
termed smoke stratification.
Fig. 8 shows the experiment which displays (a) the smoke stratifi-
cation and (b) the damaged smoke stratification during a 50 MW pool
fire. A distinct boundary is clearly seen between the layers, where the
buoyancy of the smoke is balanced by gravity, and the cold air layer
(without ventilation) below it (Fig. 8a). However, the temperature of
the smoke falls as it spreads along the tunnel, gradually weakening its
buoyancy and causing it to sink and eventually become mixed with the
surrounding air as its weight exceeds its buoyancy and the smoke layer
is destroyed (Fig. 8b). Conversely, in ventilation conditions, smoke
stratification is destroyed by the shear stress produced by the long-
itudinal flow of air, so that the smoke is quickly mixed with cold air.
Hence the smoke stratification disappears as the boundary vanishes, as
(c)
Fig. 9. Smoke spread process in the test Fig. 9. Smoke spread process during the
test. (a) Formation stage of the smoke plume and ceiling jet. (b) Smoke
spreading from the radial direction to the longitudinal direction. (c) Smoke
spreading in the longitudinal direction.
shown in Fig. 8b. As a result, the tunnel space rapidly fills with smoke.
3.1.2. Smoke spread process
Observation of the smoke spread process in the fire test (Fig. 9)
showed that, after the gasoline was ignited, the ceiling jet formed when
the smoke plume rose vertically to the tunnel vault then moved hor-
izontally (Fig. 9a). Fig. 9b shows the smoke spreading along the tunnel,
being confined by the tunnel wall in the radial direction. During the
smoke spread process, the ‘hydraulic jump’ phenomenon was observed,
where the plume energy was suddenly lost and a large volume of cold
air was entrained in the smoke, converting the kinetic energy of the
spreading smoke into potential energy. Fig. 9c shows smoke spreading
in the longitudinal direction. The front and lower boundaries of the
smoke are extremely clear.

499
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

25MW(3m/s) experimental date of the car

25MW(0m/s) experimental date of the bus
50MW(0m/s) polynomial fit of the car
polynomial fit of 25MW(3m/s)
5.5
polynomial fit of the bus
polynomial fit of 25MW(0m/s)
polynomial fit of 50MW(0m/s)
4.5

4.0 5.0
H25MW(L)=-4.58×10-7L4+9.97×10-5L3-0.0079L2+0.25L+1.88
2 Hcar(L)=-6.11×10-7L4+9.83×10-5L3-0.006L2+0.14L+4.1
3.5 R =97%
R2=98%

H (m)
3.0 4.5
H50MW(L)=-3.53×10-7L4+4.84×10-5L3-0.003L2+0.1L+2.05
H (m)

2.5 R2=98%

2.0 4.0
Hbus(L)=-2.41×10-7L4+4.02×10-5L3-0.0025L2+0.06L+4.05
1.5
H25MW(3)(L)=-5.87×10-7L4+0.1×10-5L3-0.0069L2+0.19L+0.2 R2=95%
2
1.0 R =97%
3.5
0 10 20 30 40 50 60
0 10 20 30 40 50 60 70
L (m) L (m)

Fig. 10. Longitudinal distribution of smoke layer height. Fig. 12. Distribution of smoke layer height in vehicle fires.

Fig. 11. Smoke layer height in vehicle fire tests: (a) car test; (b) bus test.

500
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

2.50 gasoline 0m/s 4.5 m to 3.5 m, an average drop of 22%. Therefore, longitudinal air
gasoline 1.5m/s velocity has a far greater effect than HRR on the height of the smoke
diesel oil 0m/s layer.
2.00 The burning tests T4 and T5 on vehicles is shown in Fig. 11. The
1.77 variation of smoke layer height in the longitudinal direction of the
tunnel is shown in Fig. 12. The HRR of the bus fire was approximately
1.50
15 MW, and approximately 5 MW for the car fire. The smoke layer of
1.21 the bus was obviously thicker than for the car.
V (m/s)

1.15
1.00 Figs. 10 and 12 show that the smoke layer height obeyed a quartic
polynomial trend in the longitudinal direction of the tunnel (goodness
of fit > 95%). This trend indicates that the smoke layer height gra-
0.50 dually decreases as the longitudinal distance from the fire source in-
creases. This occurs because both the temperature and the buoyancy
0.00
decrease gradually as the smoke spreads with the result that the degree
of mixing of the smoke and cold air gradually increases, increasing the
thickness of the smoke layer. Limited by the length of the experimental
-100 0 100 200 300 400 500 600 700 800
tunnel, the above relationship indicates that the safe distance for per-
T (s) sonnel is 97 m in T1, 20 m in T2, 75.5 m in T3, 87.5 m in T4 and 94.5 m
Fig. 13. Smoke spread velocity at 60 m from the fire source. in T5. Since the ventilation air flow was the most destructive to the
smoke layer, as discussed, the minimum safe distance occurred in the
ventilation conditions. Therefore, in order to extend the safe evacuation
3.2. Smoke layer height
distance and time for personnel in bi-directional traffic or where the
road is blocked, the longitudinal velocity should be reduced to as close
Because of the smoke stratification, as long as personnel stay in the
to zero as possible when a fire occurs in a tunnel.
lower cold air, they will not be directly exposed to the high-temperature
and toxic substances of the smoke. Then the possibility of escaping is
3.3. Smoke spread velocity
increased, along with a greater visibility distance. Therefore, it is im-
portant that the smoke layer is not destroyed, and that it remains intact
The velocity at which the smoke spreads is the characteristic para-
for as long as possible. Smoke stratification was measured in terms of
meter of the smoke flow state, and reflects the smoke movement along
the smoke layer height.
the tunnel. Personnel are safe to leave as long as the evacuation speed is
In tests T1–T3, during the stable combustion stage, Fig. 10 shows
at least greater than the smoke spread velocity.
the variation of the height of the smoke layer along the tunnel. The
Tests T6–T8 were the gasoline fire and diesel oil fire tests with the
experiment showed that the longitudinal velocity and the HRR were the
same HRR, in windy and windless conditions. The variation of the
two main factors affecting smoke layer height. When the longitudinal
smoke spread velocity over time at a height of 5.5 m at 60 m from the
velocity is close to zero, the smoke remains stratified. In these condi-
fire source is shown in Fig. 13. This shows that the velocity rose to
tions, even if the HRR reaches as high as 50 MW, the smoke layer does
smoke-spread velocity from ambient velocity as the fuel burned for
not fall as low as the average height of personnel within a certain area
about 90 s. It took 90 s for the smoke front to arrive at the location 60 m
around the fire source. The smoke layer is thicker with greater HRR,
from the fire source, so the smoke front velocity was about 0.67 m/s.
mainly due to the differences both in the amount and temperature of
This velocity did not change over time, obviously, indicating that the
the smoke. When this equilibrium is broken by the flow of ventilation
smoke velocity was essentially stable during the spreading process,
air, they mix, reducing the smoke temperature. This causes the smoke
even if the longitudinal velocity of the ventilation air stream exists. In
to descend and finally diffuse across the whole tunnel, threatening
tests T6–T8 the smoke spread at a velocity of 1.21 m/s, 1.77 m/s and
personnel safety. Overall, the smoke layer height decreased from 4.5 m
1.15 m/s respectively, and was barely affected by the different fuels.
to 2 m due to the increased longitudinal velocity, with an average drop
For better observation, diesel oil was used in the other tests. When the
of 55%. Increase in HRR causes the smoke layer height to decrease from
longitudinal velocity of the ventilation air was close to zero, the

2.00 30m 2.00 30m

90m 90m
1.70
1.63
1.50 1.50

1.12
V (m/s)

1.00 1.09 1.00

0.50 0.50

0.00 0.00

-100 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
T (s) T (s)
(b)
Fig. 14. Smoke spread velocity at different distances from fire source (5 MW diesel oil fire) when longitudinal ventilation air velocity was (a) 0 m/s; (b) 1.5 m/s.

501
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

3.5 3.5
car 45m bus 45m
3.0 car 60m 3.0 bus 60m
2.5 2.5

2.0 2.0
V (m/s)

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0

-100 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
T (s) T (s)
Fig. 15. Smoke spread velocity at different locations for vehicle fires.

120 gasoline 60m 120 diesel oil 60m

gasoline 120m diesel oil 120m
100 100

80 80
CO (ppm)

60 60

40 40

20 20

0 0

0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
T (s) T (s)
Fig. 16. CO concentration vs. time (air velocity = 1.5 m/s).

100 100
60m 60m
120m 120m
80 80

60 60
CO (ppm)

40 40

20 20

0 0

0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
T (s) T (s)
(b)
Fig. 17. CO concentration at ventilation air velocities of (a) 0 m/s; and (b) 1.5 m/s.

velocity of smoke spread rose to 1.1–1.2 m/s from ambient velocity,
increasing at a rate of 110–120%. As the longitudinal velocity of the
ventilation air reached 1.5 m/s, the smoke spread velocity increased to
1.7 m/s from the ambient velocity, with an increase rate of only 15%.
This illustrates that smoke flow depends on the buoyancy of the smoke
in windless condition and on wind pressure in windy conditions.
The smoke spread velocity at different distances from the fire source
was observed in tests T8 and T9 for ventilation velocities of 0 m/s and
1.5 m/s, respectively (Fig. 14). It was concluded from the experiment
that the smoke spread velocity was lower further from the fire source.

502
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

P. Xu et al. Tunnelling and Underground Space Technology 81 (2018) 494–505

120 cold air.

5.5m
4.5m (2) High-temperature smoke from a fire has a buoyancy force due to its
100 reduced density.
(3) Thermal expansion of the smoke was caused by the high tempera-
80 ture of the fire.
(4) The wind pressure increased the smoke flow.
CO (ppm)

60
The effect of the above factors depends on the conditions inside the
tunnel. Usually, the smoke flow depends on the buoyancy of the smoke
40
near the fire source, while the external factors play the most important
role in smoke flow at locations far from the fire source.
20

3.4. CO concentration in smoke

0
0 200 400
T (s)
Fig. 18. CO concentration at different heights.
Closer to the fire, the effect of the buoyancy of the smoke was more
obvious. At longitudinal velocity of the ventilation air of 0 m/s, the
smoke spread velocities were 1.12 m/s at a point 30 m from the fire
source, and 1.09 m/s at a point 90 m from the fire source. When the
longitudinal velocity of the ventilation air was 1.5 m/s, the smoke
spread velocities were 1.7 m/s and 1.63 m/s, respectively. Thus, the
differences in smoke velocity at different distances from the fire source
were not obvious in this test, perhaps due to the limited tunnel length.
Otherwise, this phenomenon does not seem to be related to the long-
itudinal velocity of the ventilation air.
In tests T4 and T5, as the real vehicles were burning, the smoke
spread velocity vs. time are plotted in Fig. 15. This shows that the
smoke spread velocity of the car fire was different from that of the bus
fire. The most obvious difference was that the smoke spread velocity for
the car fire remained constant, while that for the bus fire decreased. The
reason is that the smoke produced by burning interior fittings in the bus
raised the smoke spread velocity temporarily, then remained at ap-
proximately 1.5 m/s once the interior fittings had been completely
burned. The smoke spread velocity of the burning car was steady at
1.5 m/s. The other distributions were basically consistent with those of
the pool fires.
In summary, the factors driving high-temperature smoke flow
mainly include:
(1) Stack effect is caused by the pressure difference induced by tem-
perature difference between the high-temperature smoke and the
The smoke contained toxic and harmful substances, corrosive
600 800
components and solid particles. Toxicological post-mortem reports
show that CO is the main cause of death in fires (Alarie, 2002;
Hietaniemi et al., 1999). Thus a study of CO transport properties fol-
lowing a fire was particularly important in the present work. The CO
produced by combustion in the tunnel is mixed with fresh air and
gradually diffused and was transported similarly to smoke movement,
mainly by CO molecular diffusion in air and by convection. This ex-
periment defined the space-time distribution of the CO in the tunnel.
T7 and T9 were the tests on a gasoline fire and a diesel oil fire with
the same HRR in windy conditions. The variation of CO concentration
over time at a height of 5.5 m at different locations downstream of the
fire source is shown in Fig. 16. It is clear that the CO concentration from
both fuels gradually increased to a maximum value, then fell rapidly
over time. Because the CO was generated mainly from incomplete
oxidation because of insufficient oxygen in the tunnel, the main stages
for forming the observed concentration trend were: (a) in the initial
stage of the fire, sufficient oxygen was available for complete com-
bustion to CO2 and the production of CO was small. (b) As the com-
bustion intensified, the available oxygen was depleted and the CO yield
increased continuously until the oxygen was exhausted, and the CO
concentration reached its highest point. (c) Eventually, the CO con-
centration decayed gradually until the fuel was exhausted. The oxygen
concentration is inversely proportional to the CO concentration. In
addition, the CO concentration generated by burning gasoline is greater
than for diesel oil, due to the difference in fuel component.
The influence of fuel type on CO concentration was much less than
the influence of location. It is clear in the two figures that the CO
concentration 60 m from the fire source was 50% higher than at 120 m
for two reasons: as the smoke spread, heat was transferred from the
smoke to the tunnel wall; and the smoke entrained cold air. The smoke
temperature gradually decreased and the smoke sank, diluting the CO

500 500

400 400
Bus 30m Car 30m
Bus 60m Car 60m
300 300
CO (ppm)

200
200

100
100

0
0
-500 0 500 1000 1500 2000 2500 3000 3500 4000 0 500 1000 1500 2000 2500 3000 3500 4000
T (s) T (s)
Fig. 19. CO concentrations for burning vehicles.

503
 Downloaded from https://iranpaper.ir
P. Xu et al.
concentration. In addition, the rate of increase of CO concentration
120 m from fire source was slower than at 60 m, since CO was con-
stantly being transported to the more distant location. A lag time of 75 s
before CO concentration increase was noted, both for the gasoline fire
and the diesel oil fire, so that CO spread was independent of fuel type.
Tests T8 and T9 were conducted on the diesel oil fire in windy and
windless conditions, respectively. Fig. 17 shows the variation of CO
concentration over time. A higher CO concentration developed in
windy conditions than in windless conditions. It would be expected
that, because fresh ventilation air was constantly available, combustion
would improve and CO concentration would therefore decrease; how-
ever, the ventilation air velocity tilted the flame such that part of the
pan containing the fuel received little or no heat feedback from the
flame and smoke, which inhibited combustion. The curves show that
the longitudinal air velocity obviously played a dominant role in CO
concentration. The effect of location was also measured in these tests.
The lag time of up to 225 s before the increase of CO concentration at
the point 120 m from the fire source was longer than at the 60 m lo-
cation, since the smoke spread velocity depending on buoyancy was less
in windless condition.
The variation of CO concentration vs. time 60 m from the fire
source, and at different heights in T3, is shown in Fig. 18. Large dif-
ferences in CO concentration were found at different heights, for ex-
ample, at 5.5 m height it is as twice that at 4.5 m. This phenomenon
reasonably demonstrates the characteristics of smoke stratification.
Then, as combustion proceeds, the CO becomes mixed with the smoke
and cold air and its concentration tends to be uniform at all heights.
The CO concentration distributions in the vehicle fire tests T4 and
T5 (Fig. 19) indicate that the CO concentration for the bus fire was a
little higher than for the car fire, especially in the initial stage of
combustion. This was mainly due to the different combustible materials
in the vehicles. The variation trend of CO concentration was basically
the same when the combustible materials had been burned out. The
maximum CO concentration from the car fire was 160 ppm, and was
480 ppm for the bus fire.
To summarize, variation in CO concentration depended most on the
distance from the fire source. The concentration decreased rapidly with
increasing distance from the fire source, so CO intoxication would be
more probable near the fire source.
4. Conclusions
A full-scale model of an immersed-tube tunnel 150 m long was built
with electrical and mechanical systems and a fire test measuring system
in Zhangzhou Fujian, China, taking the Hong Kong – Zhuhai – Macao
Bridge immersed tube tunnel as prototype. Multiple pool fire and ve-
hicle fire tests were carried out in the experimental tunnel. Smoke
stratification and smoke movement were clearly observed. The location
of the smoke layer interface, the smoke spread velocity, and CO con-
centration distribution in the smoke were measured to predict the
smoke flow patterns in an actual tunnel fire scenario. The following
conclusions were drawn:
(1) During the fires, obvious smoke stratification was observed. A dis-
tinct interface, which was measured as the smoke layer height,
separated the high-temperature smoke from the underlying cold air.
During the smoke spread, the hydraulic jump fluid flow phenom-
enon was also observed.
(2) The smoke layer height obeyed a quartic polynomial trend with
increase in longitudinal distance from the fire source (goodness of
fit > 95%). Based on this, the personal safe visibility range was
predicted to be about 20 m in windy conditions and 80–90 m in
windless conditions. Therefore, the influence of longitudinal air
velocity on the height of the smoke layer was found to be far greater
than other factors, including fuel type and HRR.
(3) The personnel evacuation time was affected by the smoke spread
504
https://www.tarjomano.com/order
Tunnelling and Underground Space Technology 81 (2018) 494–505
velocity. Around the fire source or in windless condition, the
buoyancy of the smoke was the critical factor driving smoke flow;
the influence of other factors was no more than ± 20%. However,
other conditions depended on the ventilation in the tunnel. In
windless conditions, the safe walking speed in an evacuation is at
least 1.5 m/s. In windy conditions, the walking speed should be at
least the longitudinal velocity of the ventilation air stream.
(4) The CO concentration in the smoke, which was barely affected by
factors other than distance from the fire source, was observed to
sharply decrease by about 50% at a point 60 m from the fire source.
(5) By comparing the key parameters in pool fires and vehicle fires, it is
reasonable that burning vehicles were found to be replaced by pool
fire in the research.
Acknowledgments
The authors gratefully acknowledge the support extended by
Chongqing Nature Science Foundation (Award No. cstc2016jcyjA1548),
Chongqing Education Commission Science and Technology Research
Project (Award No. KJ1600505) and the Fundamental Research Funds
for the Central Universities of China and Opening funds of the State Key
Laboratory Cultivation Base for Bridge and Tunnel Engineering in
Mountain Areas under Grant no. CQSLBF-Y16-17 for funding this re-
search. The authors would like to thank the editor and the reviewers for
their helpful comments and constructive suggestions which have sig-
nificantly improved the quality of this paper.
References
Vuilleumier, F., Weatherill, A., Crausaz, B., 2002. Safety aspect of railway and road
tunnel: example of the Lotschberg railway tunnel and Mont-Blanc road tunnel. Tunn.
Undergr. Space Technol. 17, 153–158.
Yang, Y., 2014, March 13. Shanxi road accident death toll rises to 31. Retrieved
October15, 2016, from the Word Wide Web: http://news.xinhuanet.com/english/
china/2014-03/13/c_133184322.htm.
Babrauskas, V., Gann, R.G., Levin, B.C.A., 1998. methodology for obtaining and using
toxic potency data for fire hazard analysis. Fire Safety 31 (4), 345–358.
Babrauskas, V., Gann, R.G., Levin, B.C., 1992. Toxic potency measurement for fire hazard
analysis. Fire Technol. 28 (2), 163–167.
Yang, D., Huo, R., Zhang, X.L., Zhu, S., Zhao, X.Y., 2012. Comparative study on carbon
monoxide stratification and thermal stratification in a horizontal channel fire. Build.
Environ. 49, 1–8.
Hu, L.H., Huo, R., Wang, H.B., Li, Y.Z., Yang, R.X., 2007. Experimental studies on fire-
induced buoyant smoke temperature distribution along tunnel ceiling. Build.
Environ. 42 (11), 3905–3915.
Lee, S.R., Ryou, H.S., 2006. A numerical study on smoke movement in longitudinal
ventilation tunnel fires for different aspect ratio. Build. Environ. 41 (6), 719–725.
Colella, F., Rein, G., Borchiellini, R., Carvel, R., Torero, J.L., Verda, V., 2009. Calculation
and design of tunnel ventilation systems using a two-scale modeling approach. Build.
Environ. 44, 2357–2367.
Li, Y.Z., Lei, B., Ingason, H., 2011. The maximum temperature of buoyancy-driven smoke
flow beneath the ceiling in tunnel fires. Fire Saf. J. 46, 204–210.
Kurioka, H., Oka, Y., Satoh, H., Sugawa, O., 2003. Fire properties in near field of square
fire source with longitudinal ventilation in tunnels. Fire Saf. J. 38, 319–340.
Liu, F., Yu, L.X., Weng, M.C., Lu, X.L., 2016. Study on longitudinal temperature dis-
tribution of fire-induced ceiling flow in tunnels with different sectional coefficients.
Tunn. Undergr. Space Technol. 54, 49–60.
Tang, F., Mei, F.Z., Wang, Q., He, Z., Fan, C.G., Tao, C.F., 2017. Maximum temperature
beneath the ceiling in tunnel fires with combination of ceiling mechanical smoke
extraction and longitudinal ventilation. Tunn. Undergr. Space Technol. 68, 231–237.
Kashef, A., Yuan, Z.Y., Lei, B., 2013. Ceiling temperature distribution and smoke diffusion
in tunnel fires with natural ventilation. Fire Saf. J. 62, 249–255.
Gong, L., Jiang, L., Li, S.Y., Shen, N., Zhang, Y.C., Sun, J.H., 2016. Theoretical and ex-
perimental study on longitudinal smoke temperature distribution in tunnel fires. Int.
J. Therm. Sci. 102, 319–328.
Ji, J., Bi, Y.B., Venkatasubbaiah, K., Li, K.Y., 2016. Influence of aspect ratio of tunnel on
smoke temperature distribution under ceiling in near field of fire source. Appl.
Therm. Eng. 106, 1094–1102.
Liu, C., Zhong, M.H., Shi, C.L., Zhang, P.H., Tian, X.L., 2017. Temperature profile of fire
induced smoke in node area of a full-scale mine shaft tunnel under natural ventila-
tion. Appl. Therm. Eng. 110, 382–389.
Meng, N., Wang, Q., Liu, Z.X., Yang, H., 2017. Smoke flow temperature beneath tunnel
ceiling for train fire at subway station: reduced-scale experiments and correlations.
Appl. Therm. Eng. 115, 995–1003.
Hu, L.H., Huo, R., Chow, W.K., 2008. Studies on buoyancy-driven back-layering in tunnel
fires. Exp. Therm. Fluid Sci. 32, 1468–1483.
Li, Y.Z., Lei, B., Ingason, H., 2010. Study of critical velocity and back-layering length in
 Downloaded from https://iranpaper.ir
P. Xu et al.
longitudinally ventilated tunnel fires. Fire Saf. J. 45, 361–370.
Gannouni, S., Maad, R.B., 2016. Numerical analysis of smoke dispersion against the wind
in a tunnel fire. J. Wind. Eng. Ind. Aerod. 158, 61–68.
Fan, C.G., Yang, J., 2017. Experimental study on thermal smoke backlayering length with
an impinging flame under the tunnel ceiling. Exp. Therm. Fluid Sci. 82, 262–268.
Tsai, K.C., Chen, H.H., Lee, S.K., 2010. Critical ventilation velocity for multi-source tunnel
fires. J. Wind. Eng. Ind. Aerod. 98, 650–660.
Tang, W., Hu, L.H., Chen, L.F., 2013. Effect of blockage-fire distance on buoyancy driven
back-layering length and critical velocity in a tunnel: an experimental investigation
and global correlations. Appl. Therm. Eng. 60, 7–14.
Chow, W.K., Gao, Y., Zhao, J.H., Dang, J.F., Chow, C.L., Miao, L., 2015. Smoke movement
in tilted tunnel fires with longitudinal ventilation. Fire Saf. J. 75, 14–22.
Tian, Xiangliang, Zhong, Maohua, Shi, Congling, Zhang, Peihong, Liu, Chang, 2017. Full-
scale tunnel fire experimental study of fire-induced smoke temperature profiles with
methanol-gasoline blends. Appl. Therm. Eng. 116, 233–243.
Hu, L.H., Tang, W., Chen, L.F., Yi, L., 2013. A non-dimensional global correlation of
maximum gas temperature beneath ceiling with different blockage–fire distance in a
longitudinal ventilated tunnel. Appl. Therm. Eng. 56, 77–82.
Ji, J., Fan, C.G., Zhong, W., Shen, X.B., Sun, J.H., 2012. Experimental investigation on
influence of different transverse fire locations on maximum smoke temperature under
the tunnel ceiling. Int. J. Heat Mass Transf. 55, 4817–4826.
Liu, C., Zhong, M., Shi, C., Zhang, P., Tian, X., 2017. Temperature profile of fire-induced
smoke in node area of a full-scale mine shaft tunnel under natural ventilation. Appl.
Therm. Eng. 110, 382–389.
Tang, F., Li, L., Chen, W., Tao, C., Zhan, Z., 2017. Studies on ceiling maximum thermal
smoke temperature and longitudinal decay in a tunnel fire with different transverse
gas burner locations. Appl. Therm. Eng. 110, 1674–1681.
Hu, L.H., Chen, L.F., Wu, L., Li, Y.F., Zhang, J.Y., Meng, N., 2013. An experimental in-
vestigation and correlation on buoyant gas temperature below ceiling in a slopping
tunnel fire. Appl. Therm. Eng. 51, 246–254.
505
https://www.tarjomano.com/order
Tunnelling and Underground Space Technology 81 (2018) 494–505
Ji, J., Wan, H., Li, K., Han, J., Sun, J., 2015. A numerical study on upstream maximum
temperature in inclined urban road tunnel fires. Int. J. Heat Mass Transf. 88,
516–526.
Buvik, Harald, 2006. The Runehamar test tunnel-R&D of tunnel technology and the need
of full scales tests. In: Proceedings of the International Seminar on Road Tunnel
Operations Management and Safety. Chongqing University Press, pp. 101–103.
Ingason, Haukur, Lonnermark, Anders, 2005. Heat release rates from heavy goods vehicle
trailer fire in tunnels. Fire Saf. J. 40, 646–668.
Lonnermark, Anders, Ingason, Haukur, 2005. Gas temperatures in heavy goods vehicle
fires in tunnels. Fire Saf. J. 40, 506–668.
Ingason, Haukur, Li, Ying Zhen, Lonnermark, Anders, 2015. Runehamar tunnel fire tests.
Fire Saf. J. 71, 134–149.
Ingason, H., Lönnermark, A., Li, Y.Z., 2012a. Runehamar tunnel fire tests, SP Technical
Research Institute of Sweden, Borås, weden, 2011 (SP Report 2011:55). H. Ingason,
Fire Dynamics in Tunnels. In: Beard, A.N., Carvel, R.O. (Eds.), The Handbook of
Tunnel Fire Safety, second ed. ICE Publishing, London, pp. 273–308.
Ingason, H., Lönnermark, A., Li, Y.Z., 2012b. Model of ventilation flows during large
tunnel fires. Tunn. Undergr. Space Technol. 30, 64–73.
Hu, L.H., 2006. Studies on Thermal Physics of Smoke Movement in Tunnel Fires. State
Key Laboratory of Fire Science University of Science and Technology of China.
Hu, L.H., Huo, R., Li, Y.Z., Wang, H.B., Chow, W.K., 2005. Full-scale burning tests on
studying smoke temperature and velocity along a corridor. Tunnel. Undergr. Space
Technol. 20 (3), 223–229.
Xu, P., 2014. Research on Fire Smoke Movement Characteristic in Submarine Immersed
Tunnel. Chongqing Jiaotong University of China.
Xu, P., 2016. Experimental research on fire heat release rate in immersed tube tunnel.
Chin. J. Undergr. Space Eng. 12 (6), 1516–1523.
Alarie, Y., 2002. Toxicity of fire smoke. Crit. Rev. Toxicol. 32 (4), 259–289.
Hietaniemi, J., Kallonen, R., Mikkola, E., 1999. Burning characteristics of selected sub-
stances: production of heat smoke and chemical species. Fire Mater. 23, 171–185.