Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Applied Thermal Engineering 106 (2016) 1094–1102

Contents lists available at ScienceDirect

Applied Thermal Engineering


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

Research Paper

Influence of aspect ratio of tunnel on smoke temperature distribution


under ceiling in near field of fire source
Jie Ji a,b, Yubo Bi a, Kondapalli Venkatasubbaiah c, Kaiyuan Li d,⇑
a
State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei, Anhui 230026, China
b
Institute of Advanced Technology, University of Science and Technology of China, Hefei, Anhui 230088, China
c
Department of Mechanical Engineering, Indian Institute of Technology Hyderabad, Hyderabad 502205, India
d
Department of Civil and Structural Engineering, Aalto University, 02150 Espoo, Finland

h i g h l i g h t s

 Experiments were conducted using a model tunnel with adjustable width.


 The effect of tunnel width on mass loss rate is studied.
 Correlations for predicting the ceiling temperature distributions are proposed.
 The proposed correlations are compared with literature and show a good agreement.

a r t i c l e i n f o a b s t r a c t

Article history: Former studies usually used model tunnels with fixed width while in reality the tunnel width can be
Received 15 December 2015 different between different tunnels. In the current study, a set of small scale experiments were conducted
Revised 13 June 2016 using a model tunnel with adjustable width. The influences of tunnel width on mass loss rate, maximum
Accepted 14 June 2016
smoke temperature and temperature distributions under the ceiling were investigated. Results show that
Available online 16 June 2016
under the current experimental conditions, the tunnel width has little influence on the mass loss rate and
maximum smoke temperature under the ceiling. However, tunnel width significantly affects the smoke
Keywords:
temperature distribution. The smoke temperature under the ceiling decays more slowly at both
Tunnel fire
Tunnel width
longitudinal and transverse directions as tunnel width decreases. Empirical models are developed for both
Mass loss rate longitudinal and transverse temperature distributions which take into account the effects of tunnel width
Smoke temperature and dimensionless heat release rate. The models include exponential type functions. The index of the
Pool fire exponential function is proportional to the dimensionless heat release rate to the power of n. The value
of n is 0.4 for longitudinal distribution and 0.5 for transverse distribution.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction For tunnels with certain height, the distance between fire source
and sidewalls decreases with tunnel width. As a result, the sidewalls
In recent years, significant amount of researches were carried out will affect the air entrainment and radiation feedback, which will
in model or real tunnels with fixed height and width and thus fixed influence the fire development. The influence of sidewall on air
aspect ratio, where the aspect ratio is height/width for a tunnel with entrainment and radiation feedback is especially strong and notice-
rectangular cross section. These researches covered many important able when the fire source is placed against the sidewall [3,4,12–14].
factors in tunnel fires, such as mass loss rate [1,2], maximum smoke Gao et al. [3,4,12] investigated the sidewall effect on the ceiling jet
temperature [3–7] and longitudinal temperature distribution temperature and flame behavior for fires with different distances
[8–11]. As these factors are important parameters providing guid- to sidewall in a model tunnel with an aspect ratio of 0.5. It is found
ance for fire detection, evacuation and even protection of tunnel that the maximum ceiling temperature increased significantly when
structure. However, in reality, the tunnel width can be very different the fire was placed against sidewall. Moreover, the flame length
between tunnels, leading to different effects on fire development. from fire against the sidewall of tunnel was higher than the one in
open space due to the restriction of air entrainment. Ji et al.
[13,14] studied the influence of sidewall effect on flame characteris-
⇑ Corresponding author.
tics and burning rate using a series of experiments with heptane pool
E-mail address: kaiyuan.li@aalto.fi (K. Li).

http://dx.doi.org/10.1016/j.applthermaleng.2016.06.086
1359-4311/Ó 2016 Elsevier Ltd. All rights reserved.
J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102 1095

Nomenclature

cp specific heat of ambient air (kJ kg1 K1) DT max maximum smoke temperature rise under the ceiling (K)

d characteristic tunnel hydraulic diameter (m) DT max;dl maximum temperature rise at a longitudinal distance of
dl longitudinal distance from the fire source (m) dl from the fire source (K)
dref distance of reference position from fire source (m) DT max;dref reference smoke temperature rise (K)
dt transverse distance from the fire source (m) DT max;dt maximum temperature rise at a transverse distance of
g gravity acceleration (m s2) dt from the fire source (K)
H tunnel height (m) w tunnel width (m)
H hydraulic diameter of the tunnel (m) wf width of the methanol pool (m)
Hef effective ceiling height (m)
L tunnel length (m) Greek symbols
m_ 00 mass loss rate per unit area (g m2 s1) g heat release enhancement coefficient
n power of Q_  qa density of ambient air (kg m3)
Q_ heat release rate (kW) u aspect ratio of the tunnel
Q_ 0 dimensionless heat release rate in Eq. (8)
Q_  dimensionless heat release rate in Eq. (9)
Ta ambient temperature (K)

fires in two model tunnels with aspect ratios of 0.43 and 0.5. They Q_ can be calculated from the mass loss rate at the steady state with
found that the burning rate increased as the effective ceiling height the heat of combustion of methanol of 20 MJ/kg [17]. Two digital
decreased due to the enhanced radiation heat feedback from the cameras with a spatial resolution of 1920  1080 and a frame rate
ceiling and ceiling jet flame, especially for fires against the sidewall. of 25 fps were used to record the experimental process from two
Nevertheless, these studies ignored the influence of tunnel width sides of the apparatus, as shown in Fig. 1. The ceiling temperatures
and the fire was only affected by only one sidewall. For relatively were measured using K-type thermocouples with a diameter of
narrow tunnels, the fire might be restricted by both sidewalls of 1 mm and a response time less than 1 s. As shown in Fig. 1 and 30
which the influence could be different from the above studies. thermocouples were installed 15 mm beneath the ceiling with the
However, little study has been carried out on the effect of aspect transverse distances from the sidewall of 0.01 m, 0.1 m, 0.2 m,
ratio. Lee and Ryou [15] indicated that no demonstrated correlations 0.3 m, 0.4 m and 0.5 m, respectively, and the interval at the longitu-
that could correlate the temperature distribution of the fire-induced dinal direction was 0.2 m. Typical cases were repeated two times
flow along the ceiling of tunnels (or corridors) with various and the outcomes showed good repeatability. A summary of the
cross-section aspect ratios. Based on numerical study, Li et al. [16] experimental conditions is presented in Table 1.
developed a simple correlation for temperature distribution along
tunnels with different aspect ratios. Their results need to be
validated experimentally. 3. Results and discussion
It can be seen from the literature review that fire characteristics
in tunnels are influenced by ceiling, sidewalls and aspect ratio of 3.1. Mass loss rate
tunnels. The change of tunnel width may affect the air entrainment
and heat feedback to the fire source. Hence, systematic study on The average mass loss rates per unit area (m _ 00 ) at the steady state
the effect of tunnel width is worthwhile. In the current study, a are shown in Fig. 2. It can be seen that m _ 00 differs very little with the
set of small-scale experiments were carried out to study the effect tunnel width. The values of m _ 00 in cases with side lengths from 0.075
of tunnel width on mass loss rate, maximum temperature and to 0.25 m are approximately 20 g/(m2 s) at all tunnel widths,
ceiling jet temperature distributions in both longitudinal and indicating that the mass loss rate per unit area of methanol pool fire
transverse directions. is little affected by the tunnel width under the current experimental
conditions. This can be explained with the help of the study
2. Experimental setup conducted by Carvel et al. [18]. Based on the literature data, they
investigated the impact of tunnel width on the heat release
Fig. 1 shows the small-scale model tunnel. The tunnel is 2.0 m enhancement coefficient (denoted as g), which is the ratio of heat
long (L), 0.5 m high (H) with an adjustable width (w) where release rates of fires in a tunnel and in the open. They concluded that
w = 0.4, 0.6, 0.8, 1.0 m. Then the aspect ratio of tunnel height to for methanol pool fires, g changes insignificantly when the width of
width ranges from 0.5 to 1.25. The ceiling, floor and one sidewall the fire over tunnel width wf/w is smaller than 0.7. In our work the
are made of 4 mm thick steel plate with 30 mm thick fire- maximum value of wf/w was 0.625. Using the conclusion from Car-
resistant board as the inner lining. The other movable sidewall is vel et al. to the present study, g is therefore a constant at a given wf/
made of 8 mm thick fire-resistant glass for observation. Methanol w. That is to say, at a given pool size, the mass loss rates of the
pool fires were used as fire sources. The fuel pans were placed at methanol fires in a tunnel will maintain constant without being
the center of the tunnel. There were 5 square pans with side lengths affected by the tunnel width. So the current tunnel widths are large
of 0.075, 0.10, 0.15, 0.20, 0.25 m while their heights are identical of enough for the pool fire and the radiation heat feedbacks from the
20 mm. All pans were made of 2 mm thick steel plates. The fuel sidewalls are ignorable. The experimental results are also compared
depth in each experiment was set as 10 mm before ignition. The to the experimental results in Refs. [17,19], as shown in Fig. 2. In
effective ceiling height, Hef, was defined as the distance between open space, as the methanol pool fires with pool diameter higher
pan bottom and ceiling which was 0.45 m in each experiment. The than 20 cm, the mass loss rates per unit area was experimentally
ambient temperature recorded by a mercury thermometer ranged determined as a constant at 17 g/(m2 s) [17]. This value is slightly
from 291 K to 295 K. The fuel mass versus time was recorded using lower than the current study as caused by the heat feedback from
a CPA34001S balance with an accuracy of 0.1 g. The heat release rate the ceiling. It can also be seen in Fig. 2 that the data of m _ 00 for
1096 J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102

(a) Solid diagram Longitudinal 7KHUPRFRXSOHV


Thermocouples
rse
nsve
Tra

Camera
&DPHUD1 H+ = 0.5

Pan
wZ = 0.4, 0.6, 0.8, 1.0
Camera 2
&DPHUD L /= 2.0
)LUHUHVLVWDQWJODVV
Fire-resistant glass Electronic
(OHFWURQLF
EDODQFH
balance Pan length = 0.075, 0.1, 0.15, 0.2, 0.25

(b) Top view


dt
Longitudinal centerline
dl
Transverse centerline (unit: m)

Fig. 1. Schematic of experimental apparatus.

Table 1 700
Experimental conditions.
Experimental data
Pan size (m  m) Tunnel width (m) 600 w = 1.0 m
w = 0.8 m
0.075  0.075 1.0, 0.8, 0.6, 0.4
0.10  0.10 1.0, 0.8, 0.6, 0.4
500 w = 0.6 m
0.15  0.15 1.0, 0.8, 0.6, 0.4 w = 0.4 m
0.20  0.20 1.0, 0.8, 0.6, 0.4 400 Eq. (3)
0.25  0.25 1.0, 0.8, 0.6, 0.4
Δ Tmax

300

200 Literature data


25
Gao et al. [3]
100 Tunnel A [20]
Mass loss rate per unit area (g/m s)

Tunnel B [20]
2

20 0
0 5 10 15 20 25 30 35 40
2/3 5/3
Q /H
ef
15
Experimental data Fig. 3. Experimental maximum smoke temperature rise and comparison with data
w = 1.0 m w = 0.8 m in literature.

w = 0.6 m w = 0.4 m
10
Literature data
thermocouples positioned right above the fire source. Fig. 3 shows
Karlsson and Quintiere [17]
the maximum temperature rises under the ceiling in all tests with
Liu [19]
various tunnel widths and pan sizes. It can be observed that the
5 maximum smoke temperature rise varies little with the tunnel
0.05 0.10 0.15 0.20 0.25
width at a given pan size and tunnel height, indicating that the
Side length of pan (m) tunnel width has a weak impact on the maximum smoke
Fig. 2. Mass loss rates per unit area of pans with different tunnel widths.
temperature under the ceiling. This is because the mass loss rate
changes little with the tunnel width. As the combustion efficiency
of methanol is close to unity [17], the total energy released by
methanol pool fires in a no wind tunnel from Liu [19] is comparable methanol pool changes little with the tunnel width. As a result,
to the current study, which implies the current results are the maximum smoke temperature changes insignificantly as the
reasonable. tunnel width varies.
Li et al. [20] carried out two sets of small-scale experiments
using two model tunnels. They developed an expression to predict
3.2. Maximum smoke temperature under the ceiling the maximum smoke temperature. When the flame did not reach
the ceiling, the maximum temperature rise can be expressed as
In the experiments, the flames generated by the pool fires with
pan side lengths of 0.075 m and 0.10 m cannot reach the ceiling Q_ 3
2

while those from pan side lengths of 0.15 m, 0.20 m and 0.25 m DT max ¼ 17:5 5
ð1Þ
can reach the ceiling intermittently. Li et al. [20] found that the H3ef
location of maximum ceiling gas temperature is right above
the center of fire source under natural ventilation. Therefore, where DT max is the maximum smoke temperature rise under the
the maximum temperature can be determined using the ceiling, Q_ is the heat release rate, Hef is the effective ceiling height.
J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102 1097

Ji et al. [8] experimentally studied the impact of transverse fire away from the tunnel sidewall rather than at the longitudinal
location on maximum smoke temperature under the tunnel ceiling. centerline of tunnel. Although Ingason and Li [11] compared their
They also developed an equation to predict the maximum smoke results to other researches with different aspect ratios, yet the
temperature with fires located at the longitudinal centerline: tunnel width was not involved in Eq. (5). Since no correlation
was proposed by Li et al. [16] to predict the temperature rise at
Q_ 3
2

DT max ¼ 17:9 5 ð2Þ the reference position DT max;dref , the applicability of Eq. (6) is very
H3ef limited.
In the current study, the longitudinal smoke temperature rises
The above studies indicate that the maximum smoke tempera-
2
are measured using thermocouples along the longitudinal center-
Q_ 3
ture is proportional to 5 . Similarly, the correlation between DT max line. Taking the 0.15 m pan as an example, Fig. 4(a) shows the
H3 DT max;d
ef
2
dimensionless maximum smoke temperature rise DT max
l
against
Q_ 3
and 5 can be obtained by linearly fitting the experimental results the dimensionless distance from the fire source dl
in order to obtain
H3 H
ef
DT max;d
shown in Fig. 3. the temperature decay rate. It can be seen that DT max
l
increases
As shown, the experimental results are well fitted as Eq. (3) with decreasing the tunnel width at a given The smoke temper- dl
.
H
with a correlation coefficient of 0.98: ature decreases owing to the heat losses to the tunnel ceiling and
Q_ 3 sidewalls as well as the fresh air when it travels under the ceiling.
2

DT max ¼ 18:03 5
ð3Þ For the same traveling distance, the area of smoke contacting the
H3ef ceiling becomes smaller as the tunnel width decreases. As a result,
the heat loss to the ceiling decreases with tunnel width. The heat
Then the experimental data are also compared to the data from
loss to the cool air is similar. Consequently, the longitudinal smoke
Gao et al. [3] and Li et al. [20] in Fig. 3. Gao et al. [3] conducted
temperature decays slower as the tunnel width decreases.
experiments in a reduced scale model tunnel using methanol as DT max;d dl
fuel. The maximum temperature rises with fires located at the Fig. 4(b) shows the DT max
l
against H
at a given tunnel width of
longitudinal centerline of the tunnel were presented in Fig. 3. In 0.8 m. It can be seen that as the pan size increases, the temperature
Fig. 3, the data of both Tunnel A and B are adopted from DT max;d
decay rate decreases, which indicates that DT max
l
is also affected by
Ref. [20], which shows that Eq. (3) correlates well with experimen-
the heat release rate.
tal data of this study as well as those in the literature.

3.3. Smoke temperature distribution


(a) 1.1
3.3.1. Longitudinal smoke temperature distribution under the ceiling
1.0 Longitudinal distribution
Many studies have been performed on the longitudinal smoke
temperature distribution under the tunnel ceiling [8,11,16]. When 0.9 w = 1.0 m
the fire was located at the centerline of tunnel, Ji et al. [8] proposed w = 0.8 m
0.8 w = 0.6 m
the following correlation to predict the longitudinal smoke
ΔTmax , d / ΔTmax

w = 0.4 m
temperature distribution. 0.7

DT max;dl 0.6
l

¼ 0:65edl =H þ 0:35 ð4Þ


DT max 0.5
where DT max;dl is the temperature rise at a longitudinal distance of dl
0.4
from the fire source, H is the tunnel height. Using wood crib as a fire
source, Ingason and Li [11] developed an empirical model to predict 0.3
the longitudinal smoke temperature distribution under a ventilated 0.2
tunnel ceiling, which is -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
   
DT max;dl dl dl dl /H
¼ 0:57e 0:13 H þ 0:43e 0:021 H ð5Þ
DT max
(b) 1.1
In the study of Ingason and Li [11], DT max was nearly a constant
1.0 Longitudinal distribution
of 900–1000 K due to the flames hit the ceiling continuously. Li
0.075 m pan
et al. [16] developed a simple correlation and numerically studied 0.9
0.10 m pan
the longitudinal smoke temperature distribution under the ceiling
0.8 0.15 m pan
by taking the aspect ratio into consideration:
ΔTmax, d / ΔTmax

( 0.20 m pan
 0.7
DT max;dl eð0:00733ðdl dref Þ=d u P 1 ¼ H 0.25 m pan
¼ ; d ð6Þ
DT max;dref 
eð0:00406ðdl dref Þ=d u < 1 2u þ 20:3=u 0.6
l

0.5
where DT max;dref is the reference smoke temperature rise, u is the
aspect ratio of the tunnel cross section, u ¼ H=w, dref is the 0.4

reference position at which the distance from the fire source along 0.3
 is the
the ceiling is equal to the tunnel width, that is, dref ¼ w, d
0.2
characteristic tunnel hydraulic diameter. -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
It can be seen that the above correlations follow the exponential d l /H
decay trend. Note that Eqs. (4) and (5) are obtained from tunnels
with specific aspect ratios. And the measuring points of longitudi- Fig. 4. Dimensionless longitudinal temperature distributions for (a) 0.15 m pan
nal temperatures in the study of Ji et al. [8] were located at 1/4w length with various tunnel widths and (b) 5 pans under the tunnel width of 0.8 m.
1098 J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102

Based on the Eq. (4) proposed by Ji et al. [8], the relationship of Determination of B
DT max;d dl Experimental results under each tunnel width are fitted using
DT max
l
and H
can be expressed as
Eq. (11) to come up with the values of A, B and C as listed in Table 2.
DT max;dl dl
The index coefficient, B  Q_  , is a function of tunnel width w and
0:4

¼ Ae H =B þ C ð7Þ
DT max dimensionless heat release rate Q_  . As Q_  is known, B is a function
where A, B and C are coefficients. Lee and Ryou [21] studied the of w. According to Eq. (6), B is a function of u and u1
critical velocity in ventilated tunnels by using the dimensionless w H
B ¼ C1 þ C2 ð12Þ
heat release rate parameter (Q_ 0 ) considering aspect ratio of the H w
tunnel, where Q_ 0 is defined as Both sides of Eq. (12) are multiplied by w
, Eq. (12) can be
H

Q_ 0 ¼ Q_ =ðqa cp T a g 1=2 H5=2 u1=2 Þ ð8Þ rewritten as


w w2
where qa is the density of ambient air, cp is the specific heat of B ¼ C1 þ C2 ð13Þ
H H
ambient air, T a is the ambient temperature, g is the gravity
 2
acceleration, H is the hydraulic diameter of the tunnel and u is Eq. (13) indicates that B w
H
is linearly related to w
H
. The values of B
the aspect ratio of the tunnel. are from Table 2 and the results are well fitted as follows with the
From Fig. 4b, the longitudinal smoke temperature distribution related coefficient of 0.93.
under the ceiling is related to heat release rate. Referring to
w w2
Eq. (8), a new dimensionless heat release rate denoted as Q_  is B ¼ 0:361 þ 1:08 ð14Þ
H H
therefore introduced.
Therefore,
Q_  ¼ Q_ =ðqa cp T a g 1=2 H5=2 u1=2 Þ ð9Þ
w H
B ¼ 0:361 þ 1:08 ð15Þ
The phenomenon that the longitudinal temperature decay rate H w
decreases with increasing heat release rate was also observed by Determination of A and C
Fan et al. [22]. They found that the coefficient B is inversely Substituting Eq. (15) into Eq. (11) gives
proportional to n power of dimensionless heat release rate, namely,  
B ¼ f ðQ_  Þ. Shifting the term Q_  from the right to the left, a new
n n DT max;dl dl
ð0:361w H ÞQ  0:4
¼ Ae H = H
þ1:08w
þC ð16Þ
DT max
correction BQ_  is therefore introduced into Eq. (7) and it can be
n

rewritten as DT max;d
Eq. (16) indicates that the dimensionless temperature rise DT max
l
is
DT max;dl d _ n
d
 Hl =½ð0:361w H ÞQ 0:4 
þ1:08w
¼ AeH=ðBQ Þ þ C ð10Þ linearly related to the exponential term e , as plot- H

DT max ted in Fig. 6. Using the form of Eq. (16) to fit the data, the values of
However, it is hard to determine n because there is more than coefficients A and C can be obtained, as listed in Table 3. Fig. 6
one parameter changing in Eq. (10). In Oka and Kurioka’s study shows that at a given tunnel width, the scatters with various heat
[23], a series of consecutive values of n were used for fitting and release rate are concentrated, showing that coefficients A and C
the one with the highest correlation coefficient was chosen. Similar are strongly dependent on tunnel width while little affected by
method is used in the current study. The correlation coefficients the heat release rate.
with different n values are shown in Fig. 5. It is observed that the The relationship of A and w can be obtained by fitting the data
correlation coefficients are relatively high when n ranges between in Table 3. As A is a dimensionless variable, we use a dimensionless
0.4 and 0.5. Thus 0.4 is adopted as the final value of n. Then Eq. (10) tunnel width w H
for fitting. The results are well fitted linearly with a
can be converted to correlation coefficient of 0.99 as
w
DT max;dl dl _ 0:4 A ¼ 0:0893 þ 0:593 ð17Þ
¼ Ae H =ðBQ Þ þ C ð11Þ H
DT max
The dimensionless smoke temperature rise should be equal to 1
In the following parts, the values of A, B and C will be
right above the fire source where dl equals to 0 and the exponential
determined using the experimental data.
term is equal to 1, which indicates that the sum of A and C should
be equal to 1. As a result
1.0
w
C ¼ 0:0893 þ 0:407 ð18Þ
H
0.9
The longitudinal temperature distribution is finally determined
as
0.8 DT max;dl   dl  
R-square

w w H 0:4
¼ 0:0893 þ 0:593 e H = ð0:361Hþ1:08wÞQ
DT max H
 w 
0.7
w = 1.0 m þ 0:0893 þ 0:407 ð19Þ
H
w = 0.8 m
w = 0.6 m Table 2
0.6
w = 0.4 m Values of coefficients for longitudinal temperature distribution.

w A B C R2
0.5
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.0 0.774 1.30 0.227 0.989
0.8 0.733 1.18 0.263 0.993
n
0.6 0.692 1.26 0.304 0.993
0.4 0.675 1.77 0.325 0.987
Fig. 5. R-square for different n.
J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102 1099

1.0
1.0

0.8
0.8

ΔTmax, d / Δ Tmax
ΔTmax, d / ΔT max

0.6 0.6
0.075 m pan 0.075 m pan

l
0.10 m pan
l

0.10 m pan
0.4 0.4
0.15 m pan 0.15 m pan
0.20 m pan 0.20 m pan
0.2 0.2 0.25 m pan
0.25 m pan
Linear fit Linear fit
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
*-0.4 *-0.4
exp [-dl /HQ /(0.361w/H+1.08 H/w)] exp[- d l /HQ /(0.361w /H +1.08 H /w )]
(a) w = 1.0 m (b) w = 0.8 m

1.0 1.0

0.8 0.8

ΔTmax , d / Δ T max
ΔT max, d / ΔT max

0.6 0.6
0.075 m pan 0.075 m pan
l
l

0.10 m pan 0.10 m pan


0.4 0.4
0.15 m pan 0.15 m pan
0.20 m pan 0.20 m pan
0.2 0.25 m pan 0.2 0.25 m pan
Linear fit Linear fit
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
*-0.4 *-0.4
exp [- dl /HQ /(0.361w /H +1.08H /w )] exp[- dl /HQ /(0.361w /H +1.08H /w )]
(c) w = 0.6m (d) w = 0.4 m
Fig. 6. Dimensionless temperature rise versus exponential term.

Table 3
Values of A and C for longitudinal temperature distribution.
As the current w/H ranges from 0.8 to 2.0, Eq. (19) should be used
w A C R2 with special care out of this range. Fig. 7 shows the comparison of
1.0 0.771 0.232 0.99 measured longitudinal temperatures and the calculated values from
0.8 0.738 0.254 0.99 Eq. (19). The results show that calculated longitudinal temperatures
0.6 0.696 0.296 0.99 are in good agreement with the experimental results.
0.4 0.666 0.338 0.99

3.3.2. Transverse smoke temperature distribution under the ceiling


In tunnel fires, the hot gases travel at the transverse and
longitudinal directions. However, less study has been conducted
on the transverse smoke temperature distribution compared to
600 the longitudinal one. Fan et al. [24] conducted small-scale
Calculated longitudinal temperature rise (K)

0.075 m pan experiments to investigate the transverse smoke temperature


500 0.10 m pan distribution and developed an empirical equation, which is
0.15 m pan
0.20 m pan DT max;dt dt =w
400 ¼ 0:893e0:267 þ 0:107 ð20Þ
0.25 m pan DT max

300 where DT max;dt is the maximum temperature rise at a transverse dis-


tance of dt from the fire source, the expression of DT max is the same
200
as in Eq. (2). However, the tunnel width in their study is a constant,
which indicates that Eq. (20) may not predict the transverse smoke
temperature distribution accurately for different tunnel widths.
100
In the current study, the transverse temperature rises are
measured by the thermocouples along the transverse centerline.
0 Fig. 8(a) shows the transverse temperature distribution of 0.15 m
0 100 200 300 400 500 600 DT max;dt
pan. A dimensionless temperature rise DT max
and a dimensionless
Measured longitudinal temperature rise (K)
distance dt =w are used in Fig. 8(a) to show the temperature decay.
Fig. 7. Comparison of the measured and calculated longitudinal temperature rises. It can be seen that the transverse temperature decays slower as the
1100 J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102

(a) 1.1 1.00


Transverse distribution
1.0 w = 1.0 m
0.9
w = 0.8 m
w = 0.6 m 0.95
0.8 w = 0.4 m
ΔT max, d / ΔT max

0.7

R-square
0.90
t

0.6

0.5 w = 1.0 m
w = 0.8 m
0.4 0.85 w = 0.6 m
0.3 w = 0.4 m
0.2
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.80
d t /w 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
n
(b) 1.1
Fig. 9. Correlation coefficients for different n.
1.0 Transverse distribution
0.075 m pan
0.9 0.10 m pan Table 4
0.8 0.15 m pan Values of coefficients for transverse temperature distribution.
ΔT max, d / ΔT max

0.20 m pan
0.7 w A B C R2
0.25 m pan
1.0 0.762 0.652 0.247 0.98
t

0.6 0.8 0.735 0.806 0.270 0.98


0.6 0.687 1.276 0.322 0.98
0.5
0.4 0.596 2.319 0.414 0.97
0.4

0.3

0.2 Table 5
0.0 0.1 0.2 0.3 0.4 0.5 0.6 Values of A and C for transverse temperature distribution..
d t /w
w A C R2

Fig. 8. Dimensionless transverse temperature distributions for (a) 0.15 m pan 1.0 0.755 0.262 0.98
length with various tunnel widths and (b) 5 pans under the tunnel width of 0.8 m. 0.8 0.742 0.260 0.99
0.6 0.688 0.321 0.99
0.4 0.595 0.415 0.97

tunnel width decreases. As the tunnel width decreases the smoke


traveling distance decreases so the total heat loss along transverse
Determination of A, B and C
direction decreases, which makes more heat accumulating thus the
The same method in Section 3.3.1 is used to obtain A, B and C. B
transverse temperature decays slower.
DT max;d
is a function of w as in the longitudinal temperature distribution
Fig. 8(b) shows DT max
l
against dwt at a given tunnel width of 0.8 m. analysis.
DT max;dt
is plotted against dt
and the values of A, B and
DT max wQ_ 0:5
dt
It can be seen that when reaches 0.5, the dimensionless temper-
w C are listed in Table 4. The relationship between B and w can be
ature rise is higher for pan with larger pan size. Namely, the higher obtained by fitting the data in Table 4. Considering that B is a
the heat release rate is, the slower the transverse temperature dimensionless variable, we use the dimensionless tunnel width w H
decays with a same tunnel width. This indicates that transverse for fitting. The results are well fitted as follow with a correlation
temperature distribution is related to the heat release rate. Com- coefficient of 0.99:
paring Figs. 4 and 8 show that the transverse temperature distribu-
tion is more sensitive to the tunnel width than longitudinal B ¼ 1:67ðw=HÞ1:46 ð23Þ
temperature distribution. According to Eq. (20) and adopting the
Substituting Eq. (23) into Eq. (22) yields
similar analysis on longitudinal temperature distribution, the form
of transverse temperature distribution is expressed as DT max;dt dt 1:46 _ 0:5
¼ Ae w =½1:67ðw=HÞ Q  þ C ð24Þ
DT max;dt dt  DT max
¼ Ae w =ðBf ðQ ÞÞ þ C ð21Þ
DT max DT
Eq. (24) indicates that the dimensionless temperature rise, DTmax;d t
is
The same method as the longitudinal temperature rise is used  max
 w = 1:67ðw=HÞ1:46 Q_ 
dt 0:5

to determine f ðQ  Þ. The correlation coefficients are shown in linearly related to the exponential term e . The
Fig. 9. It can be observed that 0.5 performs the best among all fitting result shows that A and C are functions of w, which is similar
the n values. So 0.5 is taken as the value of n. Then Eq. (21) can to those in longitudinal temperature distribution. The values of A
change to and C for each tunnel width are listed in Table 5. Then we obtain
A and C by fitting the data in Table 5. Similarly, the dimensionless
DT max;dt dt 0:5
tunnel width w
¼ Ae w =ðBQ Þ þ C ð22Þ H
is used. The results are well fitted as follow with
DT max the correlation coefficient of 0.94:
J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102 1101

600 C ¼ 0:361ðw=HÞ0:564 ð25Þ


Calculated transverse temperature rise (K)

0.075 m pan
500 0.10 m pan When dt equal to 0, which means the position right above the fire
0.15 m pan DT max;dt
source, the dimensionless temperature rise DT max
should equal to
0.20 m pan
400 1. So the sum of A and C should equal to 1 accordingly, which means
0.25 m pan

300
A ¼ 1  0:361ðw=HÞ0:564 ð26Þ
The transverse temperature distribution is then obtained as:
200
DT max;dt 
¼ ð1  0:361ðw=HÞ0:564 ÞeQ 0:5
DT max
100
dt
 =ð1:67ðw=HÞ1:46 Þ þ 0:361ðw=HÞ0:564 ð27Þ
0
w
0 100 200 300 400 500 600 Fig. 10 shows the comparison of measured transverse tempera-
Measured transverse temperature rise (K) tures and the calculated values from Eq. (27). The results show that
calculated transverse temperatures are in good agreement with the
Fig. 10. Comparison of the measured and calculated transverse temperature rises.
experimental results.

1.2 3.3.3. Comparison with previous studies


Experimental data from [11] The correlations to predict the longitudinal smoke temperature
Calculated by Eq. (19) distribution were proposed by Refs. [8,11,16], as shown in Eqs. (4)–
1.0 Eq. (5) from [11] (6), respectively. However, these correlations are not suitable to
compare with the current study for the following reasons. The
0.8 measuring points of longitudinal temperatures in the study of Ji
ΔTmax, d / ΔTmax

et al. [8] were located at 1/4 w away from the tunnel sidewall
0.6 rather than at the longitudinal centerline (1/2w) of tunnel. In the
l

tests of Ingason and Li [11], fires were so large that the induced
0.4 flames hit the ceiling continuously, leading to the maximum tem-
perature DT max maintaining constant. While in our work, continu-
0.2
ous flame heights were lower than the tunnel ceiling. There is no
surprising that the calculated longitudinal temperatures using Eq.
(19) from this work will lower than the experimental data in Ref.
0.0
[11], as shown in Fig. 11. Since no correlation was proposed by Li
0 5 10 15 20 25 30
et al. [16] to predict the temperature rise at the reference position
d l /H DT max;dref , it is difficult for us to extract their numerical data to do
comparison. From the above discussions, we can conclude that
Fig. 11. Comparison of the calculated dimensionless longitudinal temperature rises
with the results from Ingason and Li [11].
Eq. (19) for predicting the longitudinal temperature distribution
needs to be validated in the future.

250 250 250


Exp. from [24] Exp. from [24] Exp. from [24]
200 Cal. by Eq. (27)
200 Cal. by Eq. (27) 200 Cal. by Eq. (27)
t
ΔTmax, d

t
ΔTmax, d
t
ΔTmax, d

150 150 150


100 Q =3.38 kW 100 Q =4.93 kW 100 Q=6.95 kW
50 50 50
0 0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
d t /w dt /w dt /w
250 250 250
Exp. from [24] Exp. from [24] Exp. from [24]
200 Cal. by Eq. (27)
200 Cal. by Eq. (27)
200 Cal. by Eq. (27)
ΔT max, dt
t
ΔTmax, d

t
ΔTmax, d

150 150 150


100 Q =9.44 kW 100 Q =12.56 kW 100 Q=16.45 kW
50 50 50
0 0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
dt / w d /w dt / w
t
250 250 250
Exp. from [24] Exp. from [24] Exp. from [24]
200 200 200
ΔT max, dt
t
ΔTmax, d

Cal. by Eq. (27) Cal. by Eq. (27) Cal. by Eq. (27)


t
ΔTmax, d

150 150 150


100 Q =21.21 kW 100 Q =23.80 kW 100 Q =29.57 kW
50 50 50
0 0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
dt / w d t /w dt /w

Fig. 12. Comparison of the calculated transverse temperature rises with the experimental data from Fan et al. [24].
1102 J. Ji et al. / Applied Thermal Engineering 106 (2016) 1094–1102

Fan et al. [24] conducted 1/6 scale experiments to investigate Funds for the Central Universities under Grant No.
the transverse smoke temperature distribution with the tunnel WK2320000033.
dimension of 6 m long, 2 m width and 0.88 m height. Nine
methanol pools with height of 15 cm were used as fire sources References
and the heat release rate were between 3.38 kW and 29.57 kW.
Fig. 12 shows the comparison of the calculated transverse temper- [1] N. Saito, Experimental study on fire behavior in a wind tunnel with a reduced
scale model, Second International Conference on Safety in Road and Rail
ature rises using Eq. (27) from this work with their experimental Tunnels, Granada, Spain (1995) 303–310.
data. It can be observed that the proposed formula performs well [2] J.S. Roh, S.S. Yang, H.S. Ryou, An experimental study on the effect of ventilation
and gives good predictions to the transverse temperature rises velocity on burning rate in tunnel fires—heptane pool fire case, Build Environ.
43 (7) (2008) 1225–1231.
under the tunnel ceiling. [3] Z.H. Gao, J. Ji, C.G. Fan, J.H. Sun, J.P. Zhu, Influence of sidewall restriction on the
maximum ceiling gas temperature of buoyancy-driven thermal flow, Energy
Build. 84 (2014) 13–20.
4. Conclusions [4] Z.H. Gao, J. Ji, C.G. Fan, J.H. Sun, Experimental analysis of the influence of
accumulated upper hot layer on the maximum ceiling gas temperature by
modified virtual source origin concept, Int. J. Heat Mass Trans. 84 (2015) 262–
This paper presents a systematic experimental investigation on 270.
the influence of tunnel width on mass loss rate, maximum smoke [5] Y.Z. Li, H. Ingason, Position of maximum ceiling temperature in a tunnel fire,
temperature, longitudinal and transverse smoke temperature dis- Fire Technol. 50 (4) (2012) 889–905.
[6] L. Li, X. Cheng, Y. Cui, W. Dong, Z. Mei, Effect of blockage ratio on the maximum
tributions under the ceiling of methanol pool fires. The conclusions temperature under the ceiling in tunnel fires, J. Fire Sci. 31 (3) (2012) 245–257.
are as follows. [7] A. Lönnermark, H. Ingason, Gas temperatures in heavy goods vehicle fires in
tunnels, Fire Safety J. 40 (6) (2005) 506–527.
[8] J. Ji, C.G. Fan, W. Zhong, X.B. Shen, J.H. Sun, Experimental investigation on
(1) Tunnel width has little influence on the mass loss rate of influence of different transverse fire locations on maximum smoke
methanol pool fire within the current experimental range. temperature under the tunnel ceiling, Int. J. Heat Mass Trans. 55 (17) (2012)
This is because the tunnel width is large enough so the heat 4817–4826.
[9] L. Li, S. Li, X. Wang, H. Zhang, Fire-induced flow temperature along tunnels
feedback from the sidewall is ignorable. The experimental with longitudinal ventilation, Tunn. Undergr. Sp. Technol. 32 (2012) 44–51.
mass loss rates were compared with previous data and the [10] A. Kashef, Z. Yuan, B. Lei, Ceiling temperature distribution and smoke diffusion
results imply that the current experimental outcomes are in tunnel fires with natural ventilation, Fire Safety J. 62 (2013) 249–255.
[11] H. Ingason, Y.Z. Li, Model scale tunnel fire tests with longitudinal ventilation,
reliable.
Fire Safety J. 45 (6) (2010) 371–384.
(2) Tunnel width has little influence on the maximum smoke [12] Z. Gao, J. Ji, H. Wan, K. Li, J. Sun, An investigation of the detailed flame shape
temperature within the current tunnel width range. An and flame length under the ceiling of a channel, Proc. Combust. Inst. 35 (3)
(2015) 2657–2664.
empirical formula was developed and compared with the
[13] J. Ji, C.G. Fan, Y.Z. Li, H. Ingason, J.H. Sun, Experimental study of non-
experimental data in the current and previous studies. The monotonous sidewall effect on flame characteristics and burning rate of n-
results show that the formula agrees well with the experi- heptane pool fires, Fuel 145 (2015) 228–233.
mental data. [14] J. Ji, Y. Fu, K. Li, J. Sun, C. Fan, W. Shi, Experimental study on behavior of
sidewall fires at varying height in a corridor-like structure, Proc. Combust. Inst.
(3) Tunnel width has significant influence on temperature dis- 35 (3) (2015) 2639–2646.
tribution in both longitudinal and transverse directions. [15] S.R. Lee, H.S. Ryou, A numerical study on smoke movement in longitudinal
The decay of smoke temperature is caused by the heat loss ventilation tunnel fires for different aspect ratio, Build Environ. 41 (6) (2006)
719–725.
from smoke to ceiling, sidewall and cool air. For the same [16] L. Li, X. Cheng, X. Wang, H. Zhang, Temperature distribution of fire-induced
traveling distance at the longitudinal direction, the area of flow along tunnels under natural ventilation, J. Fire Sci. http://dx.doi.org/10.
smoke contacting the ceiling and cool air decreases with 1177/0734904111428896.
[17] B. Karlsson, J.G. Quintiere, Enclosure Fire Dynamics, CRC Press, 2000.
tunnel width. So does the total heat loss. As a result, the lon- [18] R.O. Carvel, A.N. Beard, P.W. Jowitt, D.D. Drysdale, The influence of tunnel
gitudinal smoke temperature decays slower as the tunnel geometry and ventilation on the heat release rate of a fire, Fire Technol. 40 (1)
width decreases. Similarly, the transverse smoke tempera- (2004) 5–26.
[19] S. Liu, Studies on heat feedback and burning rate of pool fires under a
ture also decays slower as the tunnel width decreases.
horizontal air flow Dissertation, University of Science and Technology of China,
Empirical models are developed to predict longitudinal and 2011.
transverse smoke temperature distributions. The compar- [20] Y.Z. Li, B. Lei, H. Ingason, The maximum temperature of buoyancy-driven
smoke flow beneath the ceiling in tunnel fires, Fire Safety J. 46 (4) (2011) 204–
isons of experimentally measured and model predicted tem-
210.
peratures show that the developed models can predict well [21] S.R. Lee, H.S. Ryou, An experimental study of the effect of the aspect ratio on
the temperature distributions. the critical velocity in longitudinal ventilation tunnel fires, J. Fire Sci. 23 (2)
(2005) 119–138.
[22] C.G. Fan, Y.Z. Li, H. Ingason, A. Lönnermark, Effect of tunnel cross section on gas
temperatures and heat fluxes in case of large heat release rate, App. Therm.
Eng. 93 (2016) 405–415.
Acknowledgements [23] Y. Oka, H. Kurioka, Effect of shape and size of a fire source on fire properties in
vicinity of a fire source in a tunnel, Fire Sci. Technol. 25 (1) (2006) 15–29.
This work was supported by National Natural Science [24] C.G. Fan, J. Ji, Z.H. Gao, J.H. Sun, Experimental study on transverse smoke
temperature distribution in road tunnel fires, Tunn. Undergr. Sp. Technol. 37
Foundation of China (NSFC) under Grant No. 51376173, National (2013) 89–95.
Top-Notch Young Talents Program and the Fundamental Research

You might also like