International Journal of Thermal Sciences 125 (2018) 210–217

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Effects of ambient pressure on transport characteristics of thermal-driven T

smoke flow in a tunnel
Jie Ji∗, Fangyi Guo, Zihe Gao∗∗, Jiping Zhu
State Key Laboratory of Fire Science, University of Science and Technology of China, JinZhai Road 96, Hefei, Anhui 230026, China

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

Keywords: Thermal-driven smoke flow is the greatest threat to the trapped people in tunnels, and the basic understanding of
Thermal-driven smoke flow the smoke transport characteristics is necessary for the engineering applications of smoke control and exhausting
Transport characteristics system. However, most of the previous studies only focused on the smoke transport at normal pressure. As a
Ambient pressure matter of fact, the ambient pressure decreases with the increasing of the altitude, and the influence of different
Mass flow rate
ambient pressures on the transport characteristics of thermal-driven smoke has been rarely considered before.
This paper presents an investigation on the effect of ambient pressure on the smoke transport characteristics in a
tunnel. A series of fire simulations were conducted in a full scale road tunnel with ambient pressure ranged from
100 kPa to 50 kPa and the heat release rate (HRR) varied from 3 MW to 15 MW. The smoke mass transport along
the tunnel for different conditions is analyzed, and results show that for a certain HRR, the smoke mass flow rate
decreases with reduced ambient pressure due to the decreased air density and the weakening of air entrainment
strength. Based on the dimensional analysis, a quantitative model to predict the average smoke mass flow rate in
the tunnel is proposed considering both ambient pressure and HRR. Moreover, the credibility of the predictive
model is validated by comparing with a wide range of experimental results from both full scale and reduced scale
experiments at different pressures.

1. Introduction
Owing to the long-narrow structure of tunnels, the thermal-driven
smoke flow is easier to accumulate below the ceiling and spread to a
long distance along the tunnel, and consequently has caused many se-
vere casualties [1–3]. Statistics shows that smoke is the greatest threat
to the trapped people in fires, and about 85% of the victims are killed
by the hot and toxic smoke [4,5]. Therefore, it is necessary to clarify the
smoke transport characteristics to further guides the engineering ap-
plications of smoke control and exhausting system, and thus the
thermal characteristics and mass transport of the smoke in tunnel fires
have attracted much more attention recently [6–11].
Mass transport is an important aspect of the transport characteristics
of thermal-driven smoke flow, and the related knowledge of smoke
mass transport is vital for predicting smoke behavior, determining
smoke control strategies and designing smoke extraction system.
Extensive researches on mass transport of free smoke plume in open
space have been reported and some quantitative models are proposed to
estimate the smoke production rate. Mccaffrey [12] indicated that the
fire plume in open space can be divided into three regions: the
continuous flame region, the intermittent region, and the buoyant
plume region. Zukoski et al. [13] carried out experiments to measure
the smoke mass flow rate in the buoyant plume region and proposed a
classical model to predict the smoke mass flow rate based on the ideal
plume theory. The equation is given by
1/3
ρ2g
m˙ p = 0.21 ⎛⎜ 0 ⎞⎟
1/3
Q˙ Z 5/3
⎝ cP T0 ⎠ (1)
Further, Heskestad [14] introduced a “virtual origin” and employed
the convective heat release rate (HRR) to quantify the smoke mass flow
rate.
With respect to smoke mass transport in tunnel fires, there are a
certain differences due to the long-narrow structure of the tunnel.
Previous studies [7,15,16] have pointed out that the smoke spread in
tunnel fires can be divided into four stages: free rising plume stage,
radial spread stage, transition stage and one-dimensional spread stage.
The one-dimensional spread stage is the longest one of these four stages
and the smoke mass flow reaches the maximum value in this stage.
Compared with the first three stages, the mass flow rate of the fully
developed one-dimensional stage is much larger and it is more referable

∗
Corresponding author.
∗∗
Corresponding author.
E-mail addresses: jijie232@ustc.edu.cn (J. Ji), gaozihe@ustc.edu.cn (Z. Gao).

https://doi.org/10.1016/j.ijthermalsci.2017.11.027
Received 4 June 2017; Received in revised form 14 September 2017; Accepted 24 November 2017
1290-0729/ © 2017 Elsevier Masson SAS. All rights reserved.
J. Ji et al.
Nomenclature
cp specific heat capacity of air (kJ kg−1 K−1)
D∗ fire characteristic diameter (m)
g gravitational acceleration (m s−2)
H tunnel height (m)
ṁ p mass flow rate of plume in open space (kg s−1)
ṁ average mass flow rate of smoke in one-dimensional
spread stage (kg s−1)
ṁ ∗ dimensionless mass flow rate in one-dimensional spread
stage (−)
P0 ambient pressure (kPa)
for the designing of smoke exhausting system. Ji et al. [15] investigated
the stage partition of smoke spread and the air entrainment in a small
scale tunnel by measuring the temperature and velocity within the
smoke layer, and found that the air entrainment rate of smoke in one-
dimensional spread stage is significantly smaller than the first three
stages which ranges in the near filed of the fire source. Jiang et al. [16]
carried out a series of experiments in a model tunnel with exhaust
system, and found that the air entrainment of one-dimensional spread
stage increased with the smoke extraction rate and HRR. In addition,
Haselden and Hinkley [17,18] gave a comprehensive investigation on
the smoke spread velocity, and the temperature, the opacity and the
depth of the smoke layer in a disused railway tunnel. Yang et al. [19]
measured the velocity and the depth of buoyancy-driven smoke in a
reduced scale corridor and obtained the volume rate of smoke flow at
different HRRs. However, most of the current research on smoke mass
transport in tunnel fires were carried out at normal pressure by default.
211
International Journal of Thermal Sciences 125 (2018) 210–217
Q̇ heat release rate (kW)
*
Q̇P dimensionless heat release rate (−)
T0 ambient temperature (K)
Ts   smoke temperature (K)
v smoke flow velocity (m s−1)
Z height from the fire source (m)
Greek
ρ0 air density (kg m−3)
δx grid size (m)
β air entrainment coefficient (−)
In recent years, the tunnel constructions at high altitude areas (low
ambient pressure) are increasing. For instance, the Changlashan road
tunnel on Tibetan Plateau in China is located at the altitude of 4500 m
where the ambient pressure is about 58 kPa, almost half of the normal
pressure (101 kPa) [20]. The lower ambient pressure implies the
smaller air density, and this change of air density will directly affects
the smoke mass transport in tunnel fires. Tang et al. [21] carried out fire
experiments in a compartment at two ambient pressures (64 kPa and
100 kPa), and found that the air entrainment coefficient of the buoyant
spill plume at ambient pressure of 64 kPa is about 0.8 times that of
100 kPa, indicating that the strength of air entrainment process gets
weaker at lower ambient pressure. Wang et al. [22] conducted a series
of fire tests in a full scale simulated aircraft cargo compartment at four
ambient pressures (100 kPa, 90 kPa, 80 kPa and 70 kPa) and also found
that the air entrainment coefficient of smoke became smaller at lower
pressure.
Fig. 1. Arrangements of the measuring points.
J. Ji et al.
With respect to the transport characteristics of thermal-driven
smoke flow in a tunnel at high altitude areas (low ambient pressure),
Yan et al. [23] conducted six fire tests in a full scale road tunnel at the
altitude of 4100 m, where the ambient pressure is about 63 kPa, and
investigated the HRR, temperature distribution, smoke propagation and
back-layering. They concluded that the longitudinal smoke temperature
decayed more slowly at lower ambient pressure. Tang et al. [22] stu-
died the longitudinal smoke temperature at two ambient pressures
(100 kPa and 64 kPa) in a tunnel by numerical simulation and found
that the temperature decayed faster at lower pressure. Ji et al. [20]
numerically investigated the effects of ambient pressure on smoke
movement and temperature distribution in a road tunnel. They studied
the variation of the starting position of one-dimensional spread stage of
smoke flow with ambient pressures and proposed a uniform model of
longitudinal smoke temperature for different pressures environment. It
is confirmed that the ambient pressure is indeed an important factor
that can affect the air entrainment and temperature distribution, and
ultimately affects the mass transport of the smoke, and the smoke
control method of tunnel fires at lower pressures could be different
from those at normal pressure. However, in comparison to the extensive
work on tunnel fires at normal pressure, the effect of different lower
pressures (at high altitude areas) on smoke mass transport in fire tun-
nels has rarely been reported in previous studies, and whether the re-
search results at normal pressure are applicable to the lower pressure
environments is not yet known.
Therefore, this study mainly focuses on revealing the effect of am-
bient pressure on the smoke mass transport in tunnel fires and attempts
to clarify the change law of the smoke mass flow rate with varying
ambient pressures. We hope this study can increase the basic under-
standing of smoke mass transport characteristics and provide a re-
ference for engineering applications of smoke control and exhausting in
tunnel fires.
2. Numerical modeling
2.1. Fire dynamics simulator
The maturing of computational fluid dynamics (CFD) and the im-
provement of computing power have brought about the development of
CFD applied to fire and smoke research. When some fire experiments
are difficult to carry out due to the limitation of experimental condi-
tions, CFD can be employed as a convenient and effective research tool.
Fire Dynamics Simulator (FDS), a CFD tool specifically designed for
large-eddy simulations (LES) of low-speed and thermal-driven flows,
has been widely used in fire research, and its validity has been ex-
tensively verified. The new version of FDS (Version 6) offers many
improvements in the hydrodynamics, turbulence models and the scalar
transport scheme compared to FDS version 5. The improved hydro-
dynamics and turbulence models provide more dynamic range to the
flow field for coarse resolution and converge to the correct solution at
fine resolution, and these improved scalar transport schemes can pre-
vent over-shoots and under-shoots in species concentrations and tem-
perature of the smoke flow [24]. All the improvements contribute to the
accurate solution of the field of thermal-driven smoke flow. Therefore,
FDS (Version 6.4.0) with LES model is employed in this study.
More relevant information about FDS is available in the User's Guide
and technical reference guide [24,25].
2.2. Fire scenario analysis
Based on the investigation of actual road tunnels [26], a full scale
tunnel with sizes of 150 m long, 10 m wide and 5 m high is adopted in
this study, as shown in Fig. 1. The fire source with the property of N-
heptane is placed at the longitudinal tunnel centerline. The material
properties of ceiling and sidewalls are specified as “CONCRETE”, and
the tunnel openings are set as “OPEN”. The ambient temperature is set
212
International Journal of Thermal Sciences 125 (2018) 210–217
to 20 °C. In consideration of the most common fire scenarios of car fire
(3–5 MW) and truck fire (10–20 MW) in road tunnels [27], the fire
HRRs are designed as 3 MW, 5 MW, 10 MW and 15 MW. As mentioned
above, for tunnels at high altitude area, the ambient pressure can be
much lower than the normal pressure, for example, the ambient pres-
sures of Milashan road tunnel and Changlashan road tunnel decreases
to only 56 and 58 kPa, respectively. Thus, six ambient pressures (50, 60,
70, 80, 90 and 100 kPa) are selected as simulation conditions. The
measuring points of velocity, temperature and the measuring plane of
mass flow rate are arranged as Fig. 1.
2.3. Sensitivity study on the grid system
To ensure the reliability of the simulation, one of the key work is to
select the appropriate grid size. The ratio of the fire characteristic
diameter to grid size, D∗/ δx has been widely adopted as a criterion of
the grid resolution [20,28,29], and D∗ is given by
2
Q˙ 5
D∗ = ⎜⎛ ⎞
⎟
ρ c T
⎝ 0 p 0 g ⎠ (2)
It is recommended by the FDS user's guide that the value of D∗/ δx
should be in the range of 4–16 [24]. Taking HRR of 3 MW for example,
the characteristic fire diameter D∗ for 100 kPa and 50 kPa are 1.45 m
and 1.91 m, respectively, and the recommended grid size range is about
0.1–0.5 m. Thus, six grid sizes (0.500 m, 0.333 m, 0.250 m, 0.200 m,
0.167 m and 0.125 m) are chosen from the recommended range for
comparison. Fig. 2 displays the vertical temperature distribution 15 m
away from the fire source in the quasi-steady state. With the decrease of
the grid size, the temperature profiles trend to be uniform. It can be
observed from Fig. 2 that there is only a slight difference between the
temperature profiles of 0.167 m and 0.125 m, in other words, there is
no significant improvement but the computing time increases when the
mesh size is smaller than 0.167 m. Therefore, 0.167 m (6 grids/m) is
selected as the grid size (see Fig. 2).
Previous study has shown that LES model is valid in simulating the
smoke temperature distribution in tunnel fires [27]. To further verify
the accuracy of the LES model, the smoke flow velocity under the
tunnel ceiling is compared with previous research. He [30] has pro-
posed an empirical formula to predict the smoke flow velocity under
the corridor ceiling, and the formula is in good agreement with Kim's
experimental results [31].
gQT˙ s ⎞1/3
v = 0.8 ⎜⎛ ⎟
2
⎝ 2cp ρ0 T0 W ⎠ (3)
where W is the corridor width, Q̇ is the HRR, Ts is the smoke tem-
perature at the certain longitudinal location, T0 is the ambient tem-
perature. Fig. 3 presents the comparison of smoke flow velocity at 10 m
from the fire source with the value calculated by Eq. (3). The smoke
flow velocity is obtained by averaging the measured values of nine
velocity measuring points illustrated as Fig. 1. As we can see, the results
agree well with the values predicted by Eq. (3), which then confirms the
validity of the LES model and grid size we choose in this study.
3. Results and discussion
3.1. Mass flow rates of the smoke along the tunnel
When a tunnel fire occurs, the process of smoke spread can be di-
vided into four stages in terms of the smoke flow pattern [7,16]. The
free plume rises up until it impinges the tunnel ceiling, and then the
ceiling jet flow spreads radially until intersects the sidewalls. The next
stage is the transition from radially spread to longitudinal one-dimen-
sional spread. Finally, a fully developed one-dimensional smoke flow
forms. Fig. 4 displays the smoke mass flow rate along the tunnel at
J. Ji et al. International Journal of Thermal Sciences 125 (2018) 210–217

80 25

70 Grid size
0.125m 20
0.167m
60

Mass flow rate (kg/s)

0.200m
Temperature (°C)

0.250m 15
50 0.333m
0.500m Ambient pressure
40 10
50kPa
60kPa
30 70kPa
5 80kPa
20 90kPa
100kPa
0
10 0 10 20 30 40 50 60 70 80 90 100
1 2 3 4 5 Distance from fire (m)
Height(m) (a) HRR=5 MW
(a) 100 kPa 30

110
25
100 Grid size
0.125m
Mass flow rate (kg/s)
90 20
0.167m
80 0.200m
15
Temperature (°C)

70 0.250m
Ambient pressure
0.333m 50kPa
60 0.500m 10 60kPa
50 70kPa
80kPa
40 5 90kPa
30 100kPa
0
20 0 10 20 30 40 50 60 70 80 90 100
10 Distance from fire (m)
1 2 3 4 5
(b) HRR=10 MW
Height(m)
Fig. 4. Smoke mass flow rate along the tunnel.
(b) 50 kPa
Fig. 2. Vertical temperature distribution 15 m from the fire source. reasons accounting for this phenomenon. On the one hand, the air
density at lower pressure environment is smaller, which implies less
4.0 mass of fresh air is entrained to the smoke layer for the same entrain-
ment velocity during the smoke spread. On the other hand, the air
Velocity predicted by He's formula (m/s)

3.5 entrainment coefficient of the smoke is smaller at lower pressure en-

vironment [21,22], that is to say, the air entrainment velocity of the
smoke is smaller at lower ambient pressure, and thus the weaker en-
3.0
trainment strength also brings about smaller mass flow rate. Based on
the studies of Wang et al. [32] and Jiang et al. [16], the air entrainment
2.5 coefficient of tunnel smoke flow spreading from x 0 to x is expressed as

HRR Δm˙
2.0 β=
3MW ρ0 us0 w (x − x 0) (4)
5MW
1.5 10MW where Δm˙ is the increment of smoke mass flow rate from x 0 to x , x 0 is
15MW the distance to fire source, us0 is the smoke velocity at x 0 , w is the width
1.0 of the tunnel. Taking HRR of 3 MW as an example, Fig. 5 shows the air
1.0 1.5 2.0 2.5 3.0 3.5 4.0 entrainment coefficient of the smoke flow from x 0 = 20 m to x = 40 m
Simulated results of velocity(m/s) under different ambient pressures. It can be observed that the air en-
Fig. 3. Comparison of the simulated smoke velocity with predicted values by He's for-
trainment coefficient decreases with the ambient pressure, meaning
mula. relatively weaker air entrainment at lower ambient pressure. As both
the air density and air entrainment coefficient are smaller at the lower
ambient pressure, the smoke mass flow rate reduces with the decrease
different ambient pressures. The smoke mass flow rate increases with
of ambient pressure. In addition, the smoke mass flow rate of all stages
the distance from the fire source due to the air entrainment of the
increases with HRR due to the increased smoke production and the
smoke. It can be observed that the smoke mass flow rate decreases with
enhanced air entrainment.
decreasing ambient pressure in the spread process. There are two
Furthermore, it can also be observed from Fig. 4 that the smoke

213
J. Ji et al.
0.0004
0.0003
Air entrainment coeficient
0.0002
0.0001
0.0000
50 60 70 80 90
Ambient pressure (kPa)
Fig. 5. Air entrainment coefficient under different ambient pressures.
mass flow rate increases rapidly along the tunnel in the near field of the
fire source. In fact, the driving buoyancy is large with higher tem-
perature in this near field region and severe smoke fluctuations occur,
as shown in Fig. 6 and Fig. 7, which leads to stronger air entrainment.
This rapid mass flow rate rise approximately corresponds to the radial
spread stage and the transition stage of the four spread stages. With the
spread of smoke, more and more fresh air is entrained into the smoke
layer, which decreases the smoke temperature as well as the entrain-
ment strength, and then the smoke mass flow rate becomes relatively
stable, corresponding to the one-dimensional smoke spread stage.
However, an obvious increase of smoke mass flow appears close to the
end of the tunnel. This is mainly due to the influence of supplying air
flow from external environment. It can be observed from Fig. 6 that the
stable stratification of temperature is destroyed by the supplying air
flow at the tunnel end. Moreover, large-scale vortexes, which result in
strong entrainment, forms as shown in Fig. 7. The phenomenon of
smoke reaching to ground level near the open tunnel end was also
observed by Heselden and Hinkley in full scale tunnel experiments
[17], due to the mixing produced by the air supply flow. Thus, it can be
concluded that the stable smoke stratification in the one-dimensional
spread stage will be destroyed by air supplement from external en-
vironment rather than extend to the end of tunnel and the perturbed
smoke flow near the tunnel opening no longer belongs to the one-di-
mensional spread stage. As shown in Fig. 4, this unstable and mass
increasing region ranges about 25 m from the tunnel end.
International Journal of Thermal Sciences 125 (2018) 210–217
3.2. Average smoke mass flow rate in one-dimensional spread stage
According to the results of section 3.1, the one-dimensional spread
stage is the longest one during the process of smoke spread and the
smoke mass flow rate also reaches its maximum value in this stage,
excluding the mass increasing region near the tunnel end. Thus, it is
more valuable to study the smoke mass flow rate of the one-dimen-
sional spread stage to further provide references for the designing and
operating of tunnel smoke exhausting and air supply system. Kunsch
[33] indicated that the air entrainment coefficient of the smoke in one-
dimensional spread stage was about 0.00015, which proves that the
smoke mass flow in one-dimensional spread stage tends to be steady. As
shown in Fig. 4, the increment of the air entrainment along the tunnel
in one-dimensional spread stage is indeed quite small and the smoke
100 mass flow rate almost keeps a stable value. Based on the above facts, the
average of smoke mass flow rate in the steady one-dimensional spread
stage is employed as an estimation to clarify the relationship between
smoke mass flow rate and the ambient pressure together with HRR.
Based on the results of section 3.1, when the distance from the fire
source exceeds 20 m, the smoke spread under all conditions will reach
the one-dimensional spread stage, while the one-dimensional spread is
destroyed at about 20–30 m close to the tunnel end due to air supple-
ment from external environment. Thus, the average of smoke mass flow
rate from 20 m to 70 m is chosen as the estimation of the smoke mass
flow rate in one-dimensional spread stage, as shown in Fig. 8. It can be
seen that the smoke mass flow rate decreases with the ambient pressure
and increases with the HRR, and the reasons have been elucidated in
section 3.1.
Based on the results of earlier studies [13,14] and the analysis above
in this paper, it can be concluded that the governing parameters for
smoke mass flow rate in fire tunnels include the HRR of fire source Q̇ ,
ambient pressure P0, ambient temperature T0, air density ρ0, specific
heat capacity of air cp, gravitational acceleration g and the tunnel
geometry. Note that the air density and the ambient pressure are not
independent of each other, target parameter P0 is reserved here. Ac-
cording to previous study a survey of 17 urban road tunnels in China,
the aspect ratios of unidirectional tunnels are usually of the same order
[3,34]. Therefore, the tunnel height H was used as the characteristic
length. Consequently, the average smoke mass flow rate in one-di-
mensional spread stage can be expressed as
m˙ = f (Q˙ , P0, T0, cp, g , H ) (5)
Based on the dimensional analysis method, the parameters H, P0, g,
and cp, are chosen as independent variables, and then Eq. (5) can be
transformed into a dimensionless function
Fig. 6. Temperature filed with HRR of 10 MW at 50 kPa.
214
J. Ji et al. International Journal of Thermal Sciences 125 (2018) 210–217

Fig. 7. Velocity vector field with HRR of 10 MW at 50 kPa.

30 m˙ ∗ 0.30
m˙ ∗ = = 0.133(Q˙ P )
HRR P0 H 3/2/ g (10)
3MW
25 5MW
10MW 3.3. Verifications of the smoke mass flow rate model
Mass flow rate (kg/s)

15MW
20 Based on the dimensional analysis, the predictive correlation of the
smoke mass flow rate in the steady one-dimension spread stage was
derived in Section 3.2. To further check the credibility of this correla-
15 tion, it is compared with both previous full scale and reduced scale
experimental results at different pressures.
The full scale experiments were conducted in a disused railway
10 tunnel with the dimensions of 620 m (long) × 7.8 m (wide) × 5.2 m
(high) in Glasgow [17,18], and the ambient pressure is about 100 kPa.
The fire source was designed as one, two or four trays of kerosene,
5
50 60 70 80 90 100 approximately corresponding to the HRR of 2 MW, 4 MW and 8 MW,
Ambient pressure (kPa) respectively. The smoke velocity, smoke layer depth and the tempera-
ture rise of smoke at 45.7 m from the fire source were measured (see
Fig. 8. Average smoke mass flow rate in one-dimensional spread stage vs. ambient
Table 1), and then the smoke mass flow rate through the transverse
pressure.
section of the tunnel can be calculated. In addition, the full scale ex-
periment at lower ambient pressure is also utilized for comparison. Six
m˙ Q˙ T0 ⎞ fire tests were conducted by Yan et al. [23] in a 5180 m long, 10 m wide
= f ⎜⎛ ,
1/2H 5/2 gH / c ⎟
P0 H 3/2/ g P g (6) and 6.85 m high road tunnel at the altitude of 4100 m (62.63 kPa) in
⎝ 0 p⎠
Yunnan Province of China. The HRR of the fire source ranges from
By combining the two dimensionless terms on the right-hand side of 0.5 MW to 1.6 MW. Unfortunately, only one test (Test 3) can be used
Eq. (6), the function can be expressed as here for comparison (see Table 1), because merely the comprehensive
information of Test 3, including the temperature, velocity and layer
m˙ Q˙ depth of the smoke, is reported in the reference.
m˙ ∗ = = f ⎜⎛ ⎞
3/2 / g ⎟
P0 H 3/2/ g P c T
⎝ 0 p 0 H ⎠ (7)

Then, a dimensionless HRR to characterize the magnitude of the fire

of different ambient pressures can be defined as

∗ Q˙
Q˙ P =
P0 cp T0 H 3/2/ g (8)

Finally, the dimensionless smoke mass flow rate in one-dimensional

spread stage can be expressed as a function of the dimensionless HRR.

m˙ ∗
m˙ ∗ = = f (Q˙ P )
P0 H 3/2/ g (9)

Fig. 9 displays the plot of dimensionless smoke mass flow rate

against dimensionless HRR. It can be observed that in double loga-
rithmic coordinates, there is a good linear relation between di-
mensionless smoke mass flow rate and dimensionless HRR, and the
Fig. 9. Dimensionless smoke mass flow rate vs. dimensionless HRR.
fitted correlation is given by

215
J. Ji et al. International Journal of Thermal Sciences 125 (2018) 210–217

Table 1
Details of the full scale experiments.

References Ambient No. of HRR Smoke velocity Smoke layer Temperature rise of the Calculated smoke mass Predicted smoke mass
pressure (kPa) tests (MW) (m/s) depth (m) smoke (°C) flow rate （kg/s） flow rate （kg/s）

Heselden and 100 1 2.0 1.25 1.5 70 14.9 15.1

Hinkley [15,16] 2 8.0 1.25 3.2 175 26.2 22.9
3 4.0 1.33 2.3 125 21.9 18.6

Yan et al. [21] 63 3 0.9 0.48 3.1 40 11.3 10.1

Furthermore, owing to the fact that there is no more comprehensive 60

full scale experimental measurements with available smoke mass flow
rate, the correlation is also compared with the results from reduced 50
scale tunnel fire experiments. Considering that the correlation in this
study is derived from the full scale simulations, thus, the mass flow

Measured values (kg/s)

40
rates of the reduced scale experiments are extrapolated to the full scale
fire scenarios by the approach of Froude modeling for comparison
[35,36]. Yang et al. [19] and Jiang et al. [16] investigated the tem- 30
perature, velocity and depth of the smoke layer in reduced-scale (1/6
and 1/20, respectively) tunnels, and the difference is a top exhaust vent 20
Numerical simulations
being arranged in the downstream in Jiang's experiments. The experi-
Heselden and Hinkley [17,18]
mental details are listed as shown in Table 2. According to the widely Yan et al. [23]
10
used similarity criterion of fire research (Froude modeling) [35,36], the Yang et al. [19]
reduced scale experiments can be extrapolated to the full scale tunnel Jiang et al. [16]
fire scenarios. By keeping the Froude number constant, the relation- 0
0 10 20 30 40 50 60
ships of HRR, temperature and velocity between full scale tunnel fire
Predicted by Eq. (10) (kg/s)
and reduced scale experiment are given by Q˙ f / Q˙ r = (L f / Lr )5/2
Tf / Tr = L f / Lr , and vf / vr = (L f / Lr )1/2 , where L denotes the length size, Fig. 10. Comparison of the correlation with previous experimental results.
the subscript ‘f’ and ‘r’ represent the full scale fire scenario and reduced
scale experiment, respectively. Using this method, the full scale fire a wide range of tunnel fire scenarios. However, owing to the fact that
scenarios (include full scale tunnel sizes, HRRs and mass flow rates) of there is no more available experimental results of different ambient
the reduced scale experiments are obtained and the smoke mass flow pressures, the predictive correlation is compared with limited experi-
rates of these full scale fire scenarios are calculated by the predictive mental results of lower ambient pressures in this study, and its ap-
correlation of Eq. (10) (see Table 2). plicability for different lower ambient pressures will be further verified
Finally, the calculated values by Eq. (10) are compared with the in our future work.
measured mass flow rates of the full sale experiments and the extra- It should be noted that the quantitative model of smoke mass flow
polated full scale mass flow rates based on Froude modeling, as well as rate in this study can be applied to the most common tunnel fire sce-
the simulations of this work. As shown in Fig. 10, the predicted values narios. However, for some special tunnel fires, such as fires in inclined
of the smoke mass flow rate are comparable with the full scale results or ventilated tunnels, fires with varying positions (fires against tunnel
by Heselden and Hinkley at 100 kPa and that by Yan et al. at 63 kPa. sidewall for example), or the super heavy goods vehicles fires and
The correlation are also in accordance with the extrapolated full scale tanker fires in tunnels, the applicability of the proposed model needs to
mass flow rates from Ref. [19], while the results from Ref. [16] is a little be further validated.
lower than the predicted values probably because some of the smoke is
exhausted by the top exhaust vent in the downstream. Overall, the 4. Conclusions
comparative results can be considered acceptable. It can be confirmed
that the predictive correlation put forward in this study is applicable to In this study, a series of fire simulations were conducted to reveal

Table 2
Details of the reduced-scale experiments and corresponding full scale fire scenarios.

References Scale Tunnel size Full scale HRR Full scale Smoke velocity Smoke layer Smoke volume Full scale mass Predicted by Eq.(10)
Tunnel size (kW) HRR (MW) (m/s) depth (m) flow rate (m3/s) flow rate (kg/s) （kg/s）

Jiang et al. [14] 1/20 2.8 5.0 0.25 0.07 - 17.4 20.7
22.0 m 440.0 m 5.6 10.0 0.29 0.07 20.2 25.5
(long) (long) 8.4 15.0 0.38 0.06 22.7 28.8
0.60 m 12.0 m 11.2 20.0 0.48 0.06 28.7 31.4
(wide) (wide) 14.0 25.0 0.50 0.06 29.9 33.5
0.27 m 5.4 m 15.1 27.0 0.52 0.06 31.1 34.3
(high) (high) 16.8 30.0 0.57 0.06 34.1 35.4
22.4 40.0 0.66 0.06 39.4 38.6

Yang et al. [17] 1/6 66.0 m 396.0 m 58 5.1 - - 0.39 31.9 30.6
(long) (long) 83 7.3 0.44 36.0 34.1
1.5 m 9.0 m 129 11.4 0.48 39.2 38.9
(wide) (wide) 149 13.1 0.54 44.1 40.7
1.3 m 7.8 m 191 16.8 0.58 47.9 43.8
(high) (high) 218 19.2 0.60 49.1 45.6

216
J. Ji et al.
the effect of ambient pressure on smoke mass transport in tunnel fires.
Major conclusions are as follows:
(1) For a certain HRR, the smoke mass flow rate decreases with reduced
ambient pressure due to the weakening of air entrainment and the
decreased air density, and the smoke mass flow rate increases with
HRR because of the increased smoke production and the enhanced
air entrainment.
(2) Based on the dimensional analysis and simulated results, a quan-
titative model of the average smoke mass flow rate in one-dimen-
sional spread stage is derived considering both ambient pressure
and HRR. The dimensionless smoke mass flow rate ṁ ∗ is propor-
* Tunn Undergr Space Tech 60 (2016) 1–9.
tional to 0.3 power of the dimensionless heat release rate Q̇P .
(3) The predictive model of smoke mass flow rate is compared well
with experimental data from both full scale and reduced scale ex-
periments at different pressures, and the credibility of the model is
validated.
This study increases the basic understanding of the smoke transport
characteristics in tunnel fires, and helps us better understand the effect
of ambient pressure on transport characteristics of thermal-driven
smoke flow. It also provide an engineering reference for smoke control
and exhausting in tunnel fires for different ambient pressures, espe-
cially for the high altitude areas.
Acknowledgement
This work was supported by National Natural Science Foundation of
China (NSFC) under Grant No. 51376173 and No. 51706219, and the
Fundamental Research Funds for the Central Universities under Grant
No. WK2320000033. Jie Ji was supported by the National Program for
Support of Top-notch Young Professionals.
References
[1] A. Haack, Current safety issues in traffic tunnels, Tunn Undergr Space Tech 17 (2)
(2002) 117–127.
[2] R. Viskanta, Overview of some radiative transfer issues in simulation of unwanted
fires, Int J Therm Sci 47 (12) (2008) 1563–1570.
[3] Z.H. Gao, J. Ji, C.G. Fan, J.H. Sun, Experimental analysis of the influence of ac-
cumulated upper hot layer on the maximum ceiling gas temperature by a modified
virtual source origin concept, Int J Heat Mass Transf 84 (2015) 262–270.
[4] J. Ji, H.X. Wan, Y.F. Li, K.Y. Li, J.H. Sun, Influence of relative location of two
openings on fire and smoke behaviors in stairwell with a compartment, Int J Therm
Sci 89 (2015) 23–33.
[5] Y. Alarie, Toxicity of fire smoke, Crit Rev Toxicol 32 (4) (2002) 259–289.
[6] L. Gong, L. Jiang, S.Y. Li, N. Shen, Y.C. Zhang, J.H. Sun, Theoretical and experi-
mental study on longitudinal smoke temperature distribution in tunnel fires, Int J
Therm Sci 102 (2016) 319–328.
[7] J. Ji, C.G. Fan, W. Zhong, X.B. Shen, J.H. Sun, Experimental investigation on in-
fluence of different transverse fire locations on maximum smoke temperature under
the tunnel ceiling, Int J Heat Mass Transf 55 (17–18) (2012) 4817–4826.
[8] C.G. Fan, J. Ji, Z.H. Gao, J.Y. Han, J.H. Sun, Experimental study of air entrainment
mode with natural ventilation using shafts in road tunnel fires, Int J Heat Mass
Transf 56 (1–2) (2013) 750–757.
[9] M. Zhong, C. Shi, L. He, J. Shi, T. Fu, Full-scale experimental research on fire fume
refluence of sloped long and large curved tunnel, Sci China Technol Sc 54 (1)
(2011) 89–94.
[10] M. Zhong, C. Shi, L. He, J. Shi, C. Liu, X. Tian, Smoke development in full-scale
217
International Journal of Thermal Sciences 125 (2018) 210–217
sloped long and large curved tunnel fires under natural ventilation, Appl Therm Eng
108 (2016) 857–865.
[11] Z.H. Gao, Z.X. Liu, J. Ji, C.G. Fan, L.J. Li, J.H. Sun, Experimental study of tunnel
sidewall effect on flame characteristics and air entrainment factor of methanol pool
fires, Appl Therm Eng 102 (2016) 1314–1319.
[12] B.J. McCaffrey, Purely buoyant diffusion flames: some experimental results, NBSIR
79–1910 National Bureau of Standards, Washington DC, 1979.
[13] E.E. Zukoski, T. Kubota, B. Cetegen, Entrainment in fire plumes, Fire Saf J 3 (3)
(1981) 107–121.
[14] G. Heskestad, Engineering relations for fire plumes, Fire Saf J 7 (1) (1984) 25–32.
[15] J. Ji, R. Huo, Y. Zhang, W. Zhong, L.H. Hu, Exprimental study on the entrainment
mass flow rate acoss the smoke layer interface during horizontal spred in a long
channel, J Univ Sci Technol China 39 (7) (2009) 738–742.
[16] X. Jiang, M. Liu, J. Wang, K. Li, Study on air entrainment coefficient of one-di-
mensional horizontal movement stage of tunnel fire smoke in top central exhaust,
[17] A.J.M. Haselden, P.L. Hinkley, Smoke travel in shopping malls experiments in Co-
operation with Glasgow fire Brigade–Part 1, Fire Research Note No. 832 Fire
Research Station, 1970.
[18] A.J.M. Haselden, Smoke travel in shopping malls experiments in Co-operation with
Glasgow fire Brigade–Part 2, Fire Research Note No. 854 Fire Research Station,
1970.
[19] D. Yang, R. Huo, X.L. Zhang, X.Y. Zhao, On the front velocity of buoyancy-driven
transient ceiling jet in a horizontal corridor: comparison of correlations with
measurements, Appl Therm Eng 31 (14–15) (2011) 2992–2999.
[20] J. Ji, F. Guo, Z. Gao, J. Zhu, J. Sun, Numerical investigation on the effect of ambient
pressure on smoke movement and temperature distribution in tunnel fires, Appl
Therm Eng 118 (2017) 663–669.
[21] F. Tang, L.H. Hu, Q. Wang, K.H. Lu, L.Z. Yang, An experimental investigation on
temperature profile of buoyant spill plume from under-ventilated compartment
fires in a reduced pressure atmosphere at high altitude, Int J Heat Mass Transf 55
(21–22) (2012) 5642–5649.
[22] J. Wang, Y. Pan, S. Lu, K. Lu, W. Chen, CO concentration decay profile and ceiling
jet entrainment in aircraft cargo compartment fires at reduced pressures, Appl
Therm Eng 110 (2017) 772–778.
[23] Z.G. Yan, Q.H. Guo, H.H. Zhu, Full-scale experiments on fire characteristics of road
tunnel at high altitude, Tunn Undergr Space Tech 66 (2017) 134–146.
[24] K. Mcgrattan, S. Hostikka, R. Mcdermott, J. Floyd, C. Weinschenk, K. Overholt, Fire
dynamics simulator User's guide, National Institute of Standards and Technology,
Gaithersburg, Maryland, 2016.
[25] K. Mcgrattan, S. Hostikka, R. Mcdermott, J. Floyd, C. Weinschenk, K. Overholt, Fire
dynamics simulator technical reference guide, National Institute of Standards and
Technology, Gaithersburg, Maryland, 2016.
[26] W. Zhong, C.G. Fan, J. Ji, J.P. Yang, Influence of longitudinal wind on natural
ventilation with vertical shaft in a road tunnel fire, Int J Heat Mass Transf 57 (2)
(2013) 671–678.
[27] J. Ji, Z.H. Gao, C.G. Fan, J.H. Sun, Large Eddy Simulation of stack effect on natural
smoke exhausting effect in urban road tunnel fires, Int J Heat Mass Transf 66 (2013)
531–542.
[28] J. Ji, H. Wan, K. Li, J. Han, J. Sun, A numerical study on upstream maximum
temperature in inclined urban road tunnel fires, Int J Heat Mass Transf 88 (2015)
516–526.
[29] C.G. Fan, J. Ji, W. Wang, J.H. Sun, Effects of vertical shaft arrangement on natural
ventilation performance during tunnel fires, Int J Heat Mass Transf 73 (2014)
158–169.
[30] Y.P. He, Smoke temperature and velocity decays along corridors, Fire Saf J 33 (1)
(1999) 71–74.
[31] M.B. Kim, Y.S. Han, M.O. Yoon, Laser-assisted visualization and measurement of
corridor smoke spread, Fire Saf J 31 (3) (1998) 239–251.
[32] H.B. Wang, J. Ji, W. Zhong, R. Huo, Y. Zhang, L.H. Hu, Experimental study on air
entrainment coefficient during one-dimenional horizontal movement of fire-in-
duced smoke in long channels, Eng Mech 26 (11) (2009) 247–251.
[33] J.P. Kunsch, Critical velocity and range of a fire-gas plume in a ventilated tunnel,
Atmos Environ 33 (1) (1999) 13–24.
[34] Y.Z. Li, B. Lei, H. Ingason, Study of critical velocity and backlayering length in
longitudinally ventilated tunnel fires, Fire Saf J 45 (6–8) (2010) 361–370.
[35] J.G. Quintiere, Scaling applications in fire research, Fire Saf J 15 (1) (1989) 3–29.
[36] G. Heskestad, Modeling of enclosure fires, Proceedings of the fourteenth symposium
(International) on combustion, The Pennsylvania State University, USA, August
1972, pp. 1021–1030.