You are on page 1of 10

Fuel 184 (2016) 725–734

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Flame characteristics and burning rate of small pool fires under


downslope and upslope oblique winds
P. Zhu a, X.S. Wang a,b,⇑, Y.P. He c, C.F. Tao a, X.M. Ni a
a
State Key Lab. of Fire Science, University of Science & Technology of China, Hefei 230026, PR China
b
Anhui Province Center of Collaborative Innovation for City Public Security, Hefei 230027, PR China
c
School of Computing, Engineering and Mathematics, University of Western Sydney, Locked Bag 1797, Australia

h i g h l i g h t s

 Burning rates under upslope wind reduce more obviously than that under downslope wind.
 Inverse variations of flame shape occur between upslope and downslope wind.
 Flame lengths first increase, then decrease with increasing wind speed.
 Flame tilt angle increases until to an asymptotic value with increasing wind speed.
 The modified correlations are presented to describe the flame tilt characteristics.

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

Article history: This paper presents an experimental investigation on burning rate, flame tilt angle and flame length of
Received 30 May 2016 small ethanol pool fires under different oblique winds. Square pool fires with dimension of 4 cm and
Received in revised form 11 July 2016 6 cm with wind attack angles from 0° to 30° and wind speeds from 0 to 3.28 m s1 for downslope airflow
Accepted 15 July 2016
and from 0 to 1.90 m s1 for upslope airflow are tested. Flame tilt angle and flame length are determined
Available online 25 July 2016
based on luminous flame intermittency from the flame images. The results show that the burning rate
under downslope wind is relatively larger than that under upslope wind at the same absolute wind speed
Keywords:
and slope angle. The flame tilt angle and flame length generally increase with the increase of wind slope
Combustion characteristic
Pool fire
angle for downslope airflow, but decrease with the increase of upslope airflow angle. At a given wind
Flame tilt angle slope angle, the flame tilt angle generally increases with the increase of wind speed until to an asymptotic
Flame length value when the wind speed exceeds 0.35 m s1; the flame length first increases and then decreases for
Oblique wind downslope wind, but initially decreases then increases before decrease again for upslope wind. In addi-
tion, some modified correlations are developed to analyze the flame tilt angle and flame length data
under oblique wind conditions.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction circumstance, the combustion of the fuel can be affected not only
by the wind speed, but also by the wind direction. The fire hazard
Pool fires are the most frequently occurred industry accidents would be extended by the deflected flame under oblique wind
[1,2], which may be either the initiators, or the consequences in conditions.
most industry accidents involving fires and explosions [3,4]. In fact, As an important fundamental topic in fire and combustion
pool fires may occur under many topographical and meteorological research, liquid pool fire behaviour has been studied for decades
conditions, such as oblique wind conditions. For instance, the oil [6], and most of them focused on flame characteristics under qui-
tanker fires [5] occurred in oil tank battery located at hilly land escent or horizontal wind conditions, such as burning rate [7–
and the fires occurred in a sloped road tunnel are all related to both 11], flame thermal radiation [12–14], flame tilt angle [15–25],
of the downslope and upslope wind conditions. In this kind of flame length [15–18,20,24]. Flame shape is a dominant factor for
determination of flame radiant emission hazards of a real fire
scenario. Under quiescent condition, flame height would be the
⇑ Corresponding author at: State Key Lab. of Fire Science, University of Science & most important dimension in characterizing the flame geometry.
Technology of China, Hefei 230026, PR China.
However, under wind condition, the flame will deflect with an
E-mail addresses: wxs@ustc.edu.cn (X.S. Wang), Y.HE@uws.edu.au (Y.P. He).

http://dx.doi.org/10.1016/j.fuel.2016.07.059
0016-2361/Ó 2016 Elsevier Ltd. All rights reserved.
726 P. Zhu et al. / Fuel 184 (2016) 725–734

Nomenclature

L pool size (cm) Ta ambient temperature (K)


H the depth of fuel pan (cm) t time (s) pffiffiffiffiffiffi
Lf average flame length (m) Fr c combustion Froude number (Frc ¼ m _ 00 =qa gD)
0 0
uw airflow velocity (m s1) d ; e ; b; k; c; d, m, n, l, p, K, B constant parameter in the equations
uv ; uh the vertical and horizontal component of uw
(uv ¼ uw sin hw ; uh ¼ uw cos hw ) Greek symbols
u dimensionless airflow speed (u ¼ uw =uf ) hf flame tilt angle (°)
uf flame characteristic velocity (uf ¼ g m _ 00 D=qa ) hw airflow attack angle (°)
g gravitational acceleration (m s2) pffiffiffiffiffiffiffiffiffi q fluid density (kg m3)
D fire source diameter (m) (for square pool, D ¼ 2 A=p) a; g constant parameter in the equations
2 2
A fuel surface area (m ) (L ) v combustion efficiency
m fuel mass (g) / a dimensionless pan shape factor
m_ 00 burning rate (or mass burning rate per area)
(kg m2 s1)   Subscripts
Fr w Wind Froude number u2w =gD
f flame
Fr v ; Frh the vertical and horizontal component of Fr w w wind
Q_  dimensionless pffiffiffiffiffiffi heat release rate v vertical
(Q_  ¼ Q_ =ðqa cp T a gDD2 Þ)
h horizontal
Q_ heat release rate (kW) (Q_ ¼ vAm _ 00 DHc )
a ambient air
DHc heat of combustion (kJ/kg) c combustion
cp specific heat (kJ kg1 K1)

angle [18] which would increase the radiation intensity to the length Lf is defined as the distance from the centre of the burner
downwind area. Therefore, flame deflection angle (or tilt angle) surface to the visible flame tip with flame intermittency of 0.5.
and flame length would be the main fire geometric parameters of The flame tilt angle hf is defined as the angle between the normal
interest under wind conditions. direction of the fuel surface and the axis of the flame, where the
Most of the previous studies [7–24] on flame burning beha- positive value means downward sloping while the negative value
viours were conducted under horizontal airflows. However, to a means upward sloping.
practical fire scenario, the airflow field is complicate and the obli- A number of correlations to predict flame tilt angle subjected to
que airflow generally occurs, but few studies being conducted horizontal wind were used during the earlier times. The first one is
focus on the effect of oblique airflow on flame geometric character- the correlation between flame tilt angle hf and the dimensionless
istics. The studies by Tao et al. [25] and Zhu et al. [26] considered wind speed u [15–18],
the effects of oblique airflow, but only pool fire burning rate under ( 0
0
downslope wind conditions was determined. Studying on flame d ðu Þe u P 1
cosðhf Þ ¼ ð1Þ
orientation under oblique windy condition is fundamental to 1 u < 1
understand the behaviour of tank fires in hilly land or inclined tun-
nel fires, where the deflecting flame would generate large thermal uw
hazards to the nearby oil tanks or other objects and persons. There- u ¼ ð2Þ
uf
fore, the flame geometric parameters including flame tilt angle and
0
flame length would be crucial to evaluate the fire behaviours under where d and e0 are constants, u is the dimensionless wind speed,
oblique wind conditions. being the ratio of the wind speed uw and a flame characteristic
The objective of the current study is to experimentally investi- _ 00 D=qa Þ1=3 ), m
velocity uf (uf ¼ ðg m _ 00 is fuel burning intensity (or mass
gate the effects of oblique airflow on both the burning and flame burning rate per unit area), D is fire source diameter, g is the grav-
geometric characteristics of pool fires. The experiments were con- itational acceleration, qa is the ambient air density. Another corre-
ducted in a small scale wind tunnel with adjustable tilt angle. lation expresses flame tilt angle as a function of the Froude number,
Burning intensity (i.e., burning rate), flame tilt angle and flame Frw ; being given as [19–20],
length were measured and analyzed.
tanðhf Þ / Fr pw ð3Þ
2. Influence of wind on flame tilt characteristics where Fr w is defined as u2w =gD,
which is a characteristic of the ratio
of inertial force on the flames due to the wind to the buoyant force
Fig. 1 schematically presents the typical flame structure under generated by the fire. The value of the power p depends on fire
upslope and downslope wind conditions, where uw is oblique wind geometry. Oka et al. [21] also proposed a flame tilt angle model
speed, uf is flame rise characteristic velocity due to thermal buoy- based on the balance between the momentum wind flow and flame
ancy. The wind attack angle hw is defined as the angle between the buoyancy flow as,
wind direction and the horizontal plane. The vertical and the hor-
tan hf / Frm _ n
izontal components of the wind speed are uv (uv ¼ uw sin hw ) and w Q ð4Þ
uh (uh ¼ uw cos hw ), respectively. In this work, hw is determined by pffiffiffiffiffiffi
the slope angle of the wind tunnel. The differentiation between where Q_  ¼ Q_ =ðqa cp T a gDD2 Þ, is the dimensionless heat release
the downslope and upslope oblique airflow directions is achieved rate, m and n are fitting values, Q_ (vAm _ 00 DHC ) is the heat release rate
by uw , where the positive value represents downslope direction of fire, v is the combustion efficiency, which is approximated as 1 of
and the negative value represents upslope direction. The flame ethanol fire, A is the fuel surface area, DHC is the heat of combustion,
P. Zhu et al. / Fuel 184 (2016) 725–734 727

Fig. 1. Definition of flame length and title angle under different oblique wind: (a) upslope and (b) downslope.

T a is ambient temperature cp is the specific heat of ambient air. The angle and flame length was omitted in these two studies, which
values obtained by Oka et al. are 3/4 and 2/3, respectively. would be very important for defining the thermal radiation hazard
According to the studies on wood cribs fires [15] and LNG fires in the downstream of fires.
[16,17], flame length was usually correlated as,
pffiffiffiffiffiffi g b 3. Experimental apparatus and methodology
_ 00 =qa
Lf =D ¼ a  ðm gDÞ  u ð5Þ

where Lf is flame length, a, g and b are coefficients whose values 3.1. Experimental facilities
depend on the pool size and fuel type. Another correlation pre-
sented by Thomas is [15], Fig. 2 shows the schematic diagram of the experimental appara-
tus, whose width in Z-axis direction is 60 cm. The airflows are pro-
Lf =D ¼ kðFr c Þc  ðFrw Þd ð6Þ duced by a two-way axial wind fan, where the clockwise rotation
pffiffiffiffiffiffi generates downslope wind and anticlockwise rotation generates
where Fr c (=m_ 00 =qa gD) is the dimensionless burning rate, k, c and d the upslope wind. The wind flow is conditioned by a rectifying
are constant coefficients. Oka et al. [21] also correlated flame length mesh at a distance of 95 cm to the pool center to minimize the
of propane gas fires as a function of the dimensionless heat release fan disturbance. The wind speeds are measured before the ignition
rate Q_  and the Froude number Frw ; but they did not give specific of the fuel. The average speed of the downslope airflow ranged
constant coefficient of their correlation. from 0 to 3.28 m s1, and from 0 to 1.90 m s1 for the upslope air-
However, the above correlations for flame tilt angle and flame flow. The fuel used in the experiments is ethanol (C2H5OH). The
length are limited to horizontal wind only, and did not include fuel mass loss history is recorded by a load cell. Two typical
the influence of wind on the burning intensity. The burning rate conduction-controlled square pans with edge length of 4 cm and
of liquid fuel is strongly influenced by flow conditions of the sur- 6 cm are considered in this study. The depth of the stainless steel
rounding air [10,11,22–24]. The studies by Tao et al. [25] and pan is 15 mm and the wall thickness is 1 mm. The selection of
Zhu et al. [26] on the effects of downslope oblique air flow on burn- the pan size is based on the consideration of minimizing both
ing rates of square ethanol pool fires showed that the mass burning the blockage area by the pan and the interference to the flame
rate increases with the increase of the oblique angle but is inver- by the tunnel ceiling. The maximum area ratio of the pool top sur-
sely proportional to the pool size. But the analysis of the flame tilt face to tunnel cross section is 1.43%, and the maximum flame

Fig. 2. Schematic diagrams of the experimental apparatus.


728 P. Zhu et al. / Fuel 184 (2016) 725–734

Table 1
Summary of experimental scenarios.

Test cases Initial fuel mass (g) L (cm) Wind direction, hw Wind speed, uw (m s1)
1 17 4 0°, 10°, 20°, 30° 0/0.12/0.35/0.63/0.94/1.15/1.39/1.63/1.90
2 17 4 0°, 10°, 20°, 30° 0/0.32/0.87/1.15/1.44/1.74/2.02/2.72/3.28
4 35 6 0°, 10°, 20°, 30° 0/0.12/0.35/0.63/0.94/1.15/1.39/1.63/1.90
3 35 6 0°, 10°, 20°, 30° 0/0.32/0.87/1.15/1.44/1.74/2.02/2.72/3.28

length to ceiling height ratio is 25%. A method [22] based on flame atmospheric pressure. Moreover, all the other wind tunnels sur-
intermittency image analysis was used to quantitatively measure faces were simulated as wall boundary with no-slip condition.
the flame length and flame tilt angle from flame image series The turbulence quantities at the inlet are specified in terms of tur-
recorded by a digital camera in the steady burning state. bulence intensity (6%, according to our previous papers [25,26] and
hydraulic diameter (equivalent diameter of tunnel cross section).
3.2. Experimental conditions and procedure The other parameters were defaulted by ANSYS Fluent. The theo-
retical models and governing equations are not described here,
All of the test cases are specified in Table 1. In each test, the which can be seen elsewhere [30,31].
door and the windows of the laboratory room where the tunnel Three cases for horizontal wind, upslope and downslope wind
is housed are closed to avoid the influence of ambient wind. The with wind direction of 20° and wind speed of 2 m s1 are consid-
fuel pan is filled with ethanol to the rim. The fan and the data ered. Fig. 3 shows the slice of velocity contour at Z = 0, which is
acquisition system are activated at the same time. Then after the vertical centre plane in the flow direction. It can be seen that
approximately 30 s settling period, the fuel was ignited. Each test for different wind conditions, the recirculation and stagnant region
case was carried out two or three times and each repeated test would always be formed in the leeward side of the fuel pan system,
was conducted after the wind tunnel and fuel pan returned to but the effect of wind on the fuel pan surface are different with the
the initial ambient conditions. The quantitative results presented variation of wind direction. From Fig. 3a, under horizontal airflow,
in this paper are the ensemble average of all the repeated tests. the variation of velocity streamline is small near the fuel pan sur-
face due to the blockage effect. However, under oblique airflow
from Fig. 3b and c, the variations of velocity streamline are remark-
4. Results and discussion
able and the wind speed always tend to be horizontal at the fuel
surface, and the trend would be weakened with the increase of dis-
4.1. Visualization of flow field by numerical simulation
tance from the fuel pan surface.
Therefore the effect of fuel pan blockage effect would weaken
For present configurations of the experimental apparatus, the
the initial oblique flow characteristic near the fuel pan surface as
fuel pan is held horizontally, while the wind tunnel is tilted to pro-
the upstream, and this behaviour means that the effects of airflow
vide downslope or upslope wind. The blockage effect of the fuel
on flame base and flame tip for oblique winds in present study
pan in wind field is existent, which would lead to the different flow
have some difference, but the blockage effects are reality in practi-
field near the fuel pan for slope and horizontal wind, and then this
cal scenario, such as the oil tank, or the flammable objet in an open
might impact the behaviours of the flames. Therefore the flow field
space, etc.
structure near the fuel pan would be help to interpret the following
fire results fundamentally and physically. Here a simple numerical
4.2. Visualization of flame under different airflows
investigation for the case without fire is conducted to illustrate
flow field under different airflows.
Fig. 4 presents the typical flame images of the 6 cm pool fire
The commercial CFD code ANSYS Fluent 14.0 is used to simulate
under different oblique airflows. The flame tilt behaviour can be
the airflow field in the oblique tunnel. The simulation was con-
interpreted combined with the flow field as shown in Fig. 3. The
ducted with a 3D steady Reynolds-averaged Navier–Stokes (RANS)
absolute flame tilt angle value jhf j increases with the increase of
computation based on the Finite Volume Method (FVM) for solving
the flow equations. The turbulent component of the airflow was juw j for both of the slope winds, but decreases for upslope wind
modelled using the standard k-e turbulence model [27,28]. All of and increases for downslope wind with the increase of hw , which
 
the transport equations were discretized using a second-order is bounded, i.e., hf  6 ð90  hw Þ for upslope wind and
upwind scheme. The SIMPLE algorithm was used for pressure- hf 6 ð90 þ hw Þ for downslope wind.
velocity coupling. The computational domain is set as similarly At a given slope angle, the flame length appears to increase first
to the experimental tunnel with 2.5 m  0.6 m  0.4 m, where and then decrease with the increase of uw . The flame tilt angle hf
the fuel pan with size of 6 cm  6 cm is assumed as one solid object increases with the increase of wind speed uw and attack angle hw .
in the tunnel. Structured Hexahedra grids with 4,799,292 elements Flame downwash occurs for uw > 1.44 m s1 and hw > 20°. Flame
were used for horizontal airflow case. Due to the complexity of the length has the same variation trend with increase of uw , but has
geometry under oblique airflow cases, unstructured tetrahedral an inverse trend with increase of hw comparing to the cases under
grids were used to discretize the surface and volumes of the com- downslope winds. The phenomenon of flame detachment and
putational domains, which has 3,139,564 elements for downslope flame wrapping also occur at relatively high wind speed, which
airflow case, and 3,248,520 elements for upslope airflow case. The has been described in previous studies [25,26]. In most cases, the
used grid size is relatively fine (about interval size 0.005 m of aver- flame length decreases when flame detachment occurs. Quantita-
age value) for the current computational domains according to the tive results are presented in the following subsections.
similarly simulations of flow field by existing papers [27–29].
Three different boundary conditions were used in this work: veloc- 4.3. Burning rate
ity inlet, pressure outlet and walls. The airflow inlet which is the
end of the tunnel with fan was set as velocity inlet, and the other Fig. 5 presents the typical time evolution of fuel mass for the
end is set as pressure outlet, which is considered equal to 6 cm pool fire with hw = 20° under upward and downward oblique
P. Zhu et al. / Fuel 184 (2016) 725–734 729

Fig. 3. The velocity field near the fuel pan for different cases: (a) horizontal wind; (b) downslope wind and (c) upslope wind.

airflows. It can be seen that the variations of fuel mass with time
are linear for the range of 20–5 g, which means that the pool fires
are burning in a steady state during the corresponding periods,
although the depth of the fuel is decreasing with increasing burn-
ing time. It can also be seen that the flame burning period under
upslope wind is relatively longer than that under downslope wind.
The burning rate at the steady state is presented in Fig. 6 against
wind speeds at various wind attack angles. The results indicate
that at a given attack angle the burning rate increases nonlinearly
with the increase of wind speed. This outcome is similar to that of
some previous studies [25,26]. In addition, the burning rate under
downslope wind is relatively larger than that under upslope wind
at the same absolute wind speed and slope angles. The difference is
attributable to the flame wrapping in the downslope flow, which
would enhance the pool wall heating effect and the fuel evapora-
tion. But at upslope wind, the fuel surface is shielded by the bottom
of the pan from the direct wind attack and the flame wrapping
phenomenon would be weak. However, it should be noted that
at hw ¼ 0 and the same absolute value of uw , the measured burning
intensities under the forward and backward flows are slightly dif-
ferent. This difference may be due to the only one rectifying mesh
location near the wind fan, where the effect of the fan disturbance
would be different when the flow direction in the tunnel is
changed.

Fig. 4. Typical flame images of the 6 cm pool fire under different airflow conditions: 4.4. Flame tilt angle
(a) upslope wind and (b) downslope wind. The pan side is indicated with red dash
lines. (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.) Fig. 7 presents the flame tilt angle, hf , under various wind attack
angles and speeds. For downslope airflow conditions, the flame tilt

Fig. 5. Typical time evolution of mass loss for the 6 cm pool fire under different wind speeds (attack angle is 20°).
730 P. Zhu et al. / Fuel 184 (2016) 725–734

Fig. 6. The flame burning rate under different airflow speeds and attack angles.

Fig. 7. The variation of flame tilt angle under different wind attack angles and speeds.

angle hf increases with the increase of wind speed and tends to buoyant flow. However, the difference between the horizontal
asymptotically approach j90 þ hw j. However, the increase rate wind flow and the oblique wind flow is that there are two momen-
reduces dramatically when uw > 0.32 m s1. For upslope wind, jhf j tum components in the latter as illustrated in Fig. 1, i.e., the hori-
increases with increasing of juw j and asymptotically approaches zontal component uh and vertical component uv . The flow field
j90  hw j. With the increase of wind slop angle, jhf j increases structure can be seen from Fig. 3 as reference. It can be seen that
under downslope wind, but decreases under upslope wind. uh always tends to cause hf increases, whereas uv tends to help
As discussed in previous studies by many researchers [19–21], flame to retain its vertical orientation if it is the same as uf or oppo-
the flame tilt behaviour under wind conditions is mainly domi- site to g. When uv and g are in same direction, i.e., under down-
nated by the balance between momentum wind flow and flame slope wind, uv will reduce or counter balance the buoyancy
force, and the change in hf will be enhanced.
Fig. 8 presents the collapse of the experimental data by Eq. (1)
for e0 = 0.49 and d0 = 0.7 by Thomas model [15] and e0 = 0.5
and d0 = 1 by AGA model [16]. The data points in Fig. 8 do not agree
with the linear lines of the existing model. Therefore, the previous
correlation of Eq. (1) is found to be not appropriate for current
experiment results. The correlation of Eq. (3) does not consider
the mass burning rate which would be considerably enhanced by
the wind for liquid pool fires. So the correlation form of Eq. (4) is
employed to describe the relationship between the flame tilt angle,
wind velocity, orientation and burning rates. The experimental
results show that the flame tilt angle varies between 0° and 120°.
To avoid the singularity of the tangent function at 90° and consider
the negative sign of hf under upslope wind, hf is shifted by 90° in
the model, where the tangent function is modified as
tanðjhf j  90 Þ. Corresponding to the two velocity components,
Fig. 8. Plots of flame tilt angle by correlation of Eq. (1). the Froude number should also has two components as,
P. Zhu et al. / Fuel 184 (2016) 725–734 731

Fr h ¼ u2h =gD ð7Þ instabilities of the flame tilt angle and burning intensity at low
wind speed. Specifically, when the attacking angle is zero, Fr v
Fr v ¼ u2v =gD ð8Þ would be zero, and then Eq. (11) for upslope and downslope has
the same form, but the power exponent of Fr h was 1/5 for down-
where uv ¼ uw sin hw and uh ¼ uw cos hw , which has been introduced slope wind and 2/5 for upslope wind. Therefore, in order to illus-
in Section 2. When considering the different slope direction of air- trate the applicable and consistent of Eq. (11) at zero attacking
flow, as shown in Fig. 1, uh , or Frh always tends to increase jhf j, angle, the correlations of tanðjhf j  90 Þ vs Fr 1=5  Q_ ð2=15Þ ,
h
but uv , or Frv will decrease jhf j when it has a same direction as uf , tanðjhf j  90 Þ vs Fr 2=5  Q_ ð2=15Þ and Eq. (11) separately for both
h
and increase jhf j when it has a same direction as g. Namely, Frv kind of slop winds in one figure were made, as shown Fig. 10(a)–
has a negative effect on jhf j for upslope wind and has a positive (c). It can be seen from Fig. 10 that the three correlations all could
effect on jhf j for downslope wind. It is, therefore, postulated that express the data relatively well except for the low wind speed
Eq. (4) can be modified into, below 0.35 m s1 for both slope winds, but the correlation of Eq.
  (11) was relatively better. Therefore, Eq. (11) at zero attacking
tanðjhf j  90 Þ / Fr m1 m1
h  Fr v  Q_ ðn1Þ ðfor upslope windÞ ð9Þ angles were appropriate.

 
tanðjhf j  90 Þ / Fr m2 m2
h þ Fr v  Q_ ðn2Þ ðfor downslope windÞ 4.5. Flame length
ð10Þ
Fig. 11 presents the experimental results of flame length against
The values of the power coefficients, m1, n2, m2 and n2 are wind speed and attack angle. It can be seen that the variation of
obtained through regression. The experimentally measured data flame length depends on pan size, wind speed and direction. Gen-
 
Q_ ð2=15Þ for upslope
2=5
are plotted in Fig. 9(a) against Fr h  Fr 2=5v erally, the flame length increases with the increase of pool size. For
  a given pan size D under upslope flow conditions, flame length Lf
wind and in Fig. 9(b) against Fr 1=5 þ Fr 1=5
v Q_ ð2=15Þ for downslope
h initially decreases with increase of juw j then increases before
wind. It can be seen that approximate linear variations are applica-
  decrease again at about juw j = 1.2 m s1. Under downslope winds,
ble in certain range of Fr 2=5
h
 Fr 2=5
v Q_ ð2=15Þ or there appears that only one turning point can be seen to each case.
  These results are different from the previous studies [7,8,11,16,19],
1=5 1=5 _ ð2=15Þ
Fr h þ Fr v Q . Then linear regression functions of the fol- where Lf has a linearly correlation with the wind speed. The vari-
lowing form are suggested: ation of flame length with attack angle hw may or may not be

8    
>
< 0:334 Frh2=5  Fr 2=5 Q_ ð2=15Þ  0:708 upslope; 0:2 < Fr 2=5  Fr2=5 Q_ ð2=15Þ < 1:6
v h v
tanðjhf j  90 Þ ¼     ð11Þ
>
: 0:332 Frh1=5 þ Fr 1=5 Q_ ð2=15Þ  0:655 downslope; 1 < Fr h1=5 þ Fr1=5 Q_ ð2=15Þ < 2:6
v v

The values of coefficient R2 of the above two expressions are monotonous depending on the values of juw j and D. The variations
0.781 and 0.882, respectively. of flame length due to the change of hw are mainly caused by the
The correlation of Eq. (11) can express the variation of flame tilt different fuel evaporation rates and flame flow motions.
angle hf for both oblique winds well within the specified ranges of Fig. 12 presents the plots of the experimental data by Eq. (5) for
the relevant non-dimensional group. The lower limits of the appli- a = 55, g = 0.67 and b = 0.21 of Thomas model-1 [15], and a = 62,
cable range of the correlations are mainly associated with the small g = 0.254 and b = 0.044 of Moorhouse model [17] and by Eq. (6)
wind speed, such as 0.12 m s1 and 0.35 m s1 for upslope wind for k = 70, c = 0.86, d = 0.11 of Eq. (6) [15]. The data points in
and 0.32 m s1 for downslope wind, which may be caused by the Fig. 12 collectively exhibit a non-linear fashion and all existing

Fig. 9. Correlations of jhf j with Fr h ; Fr v and Q_  under different oblique airflows.


732 P. Zhu et al. / Fuel 184 (2016) 725–734

Fig. 10. The results of correlations under downslope and upslope wind condition for hw = 0°: (a) plot of tanðjhf j  90 Þ vs Fr1=5
h
 Q_ ð2=15Þ ; (b) plot of tanðjhf j  90 Þ vs
Fr h  Q_ ð2=15Þ and (c) plot of Eq. (11).
2=5

models overestimate the data of flame length in most cases. There- the same height of pan rim may have slightly difference, which
fore, the previous correlations of Eqs. (5) and (6) are found to be is due to the different surface area exposed to the airflows. So to
not applicable for the current study of flame length. demonstrate the difference for this effects, a dimensionless pan
However, the form of Eq. (6) is suggested to correlate the flame shape factor / is considered, which is defined as / ¼ H=D, where
length under oblique airflows including the influence of wind on H is the pan rim height (1.5 cm), D is the equivalent diameter of
the burning intensity. But considering previous models were square pan. So the correlation between Lf /D and the parameter
obtained under horizontal airflow, some modification should be Fr h ; Fr v ; Fr c and / are attempted. After a regression analysis, the
conducted for Fr w under oblique airflows, which has the same proposed correlation takes on the following form,
modified form as discussed in Section 4.4, i.e., being divided into 8   
pffiffiffiffiffiffi >
< Fr m1
 Fr m1 n1
K  Fr l1 p1
/ þ B Upslope
_ 00 =qa gD) is still
Fr h and Fr v . The combustion Froude number Fr c (m Lf h v 1 c 1

used. For shallow fuel pan, the effects of oblique airflows on the D >
¼   
m2 n2
 ð12Þ
: Fr h þ Frv
m2 l2 p2
K 2  Fr c / þ B1 Downslope
fuel vapour diffusion for the different diameter D of pool with

Fig. 11. Flame lengths for different airflow speeds and attack angles of all cases.
P. Zhu et al. / Fuel 184 (2016) 725–734 733

Fig. 12. Plots of flame length data by previous correlations.

θw θw
Downslope wind
20 20 0° 20
0° Upslope wind
10°
uw>1.44m/s
10°
(Lf /D)(Frh0.6+Frv0.6)
(Lf /D)(Frh1.1- Frv1.1)

20°
(Lf /D)(Frh1.1+Frv1.1)

2
16 R =0.9218 16
15 20° 30°
30° Correlation
Correlation 12 12
10 1.44m/s ≥uw
2
8 R =0.7926 8
5
R2=0.8776 4
(a) (b) 4
0 0
0
0.03 0.06 0.09 0.12 0.06 0.09 0.12 0.15 0.18
Frcφ −0.9 Frcφ −0.9
Fig. 13. Correlations of Lf with Fr h ; Fr v , Fr c and R for (a) upslope wind and (b) downslope wind.

The values of the parameters m, n, l, p, K and B can be deter- For downslope wind, as shown in Fig. 13(b), the regression anal-
mined by regression. For upslope wind, as shown in Fig. 13(a), ysis yields by,
the regression analysis can be yielded by plotting 8 1
  >
ðLf =DÞ  Fr 1:1 1:1
vs Fr c /0:9 as, Lf < Frh þ Fr v
1:1 1:1
ð212:58Fr c /0:9  19:61Þ uw 6 1:4 m s1
h  Fr v
¼  1
D >: Fr0:6 þ Fr 0:6
Lf  1:1 1
h v ð130:40Fr c /0:9  7:01Þ uw > 1:4 m s1
¼ Fr h  Fr1:1
v ð179:64Fr c /0:9  7:57Þ ð13Þ
D ð14Þ
734 P. Zhu et al. / Fuel 184 (2016) 725–734

the correlation coefficient is 0.7926 for uw 6 1.44 m s1, and 0.9218 References
for uw > 1.44 m s1. This difference may be caused by the changes of
the flame flow field due to occurrence of flame detachment. There- [1] CCPS. Guidelines for chemical process quantitative risk analysis. 2nd ed. New
York: Centre for Chemical Process Safety, AIChE; 2005.
fore, the flame length can be predicted by the correlation including [2] Lees FP. Lee’s loss prevention in the process industries: hazard identification,
the parameter of Frh ; Frv ; Fr c and /, which represents the effect of assessment, and control. 3rd ed. Oxford, UK: Elsevier; 2005.
oblique wind momentum, flame buoyancy and fuel pan geometric [3] Abbasi T, Abbasi SA. Accidental risk of superheated liquids and a framework for
predicting the superheat limit. J Loss Prev Process Ind 2007;20:165–81.
construction.
[4] Abbasi T, Abbasi SA. The boiling liquid expanding vapour explosion (BLEVE):
From Fig. 13 and Eqs. (13) and (14), it can be concluded that mechanism, consequence assessment, management. J Hazard Mater
compared with previous correlation of Eqs. (5) and (6), the vertical 2007;14:489–519.
[5] Lois E, Swlthenbank J. Fire hazards in oil tank arrays in a wind. Symp (Int)
component of the oblique wind is a crucial parameter in determin-
Combust 1979;17:1087–98. http://dx.doi.org/10.1016/S0082-0784(79)80104-
ing the flame geometric characteristics, which is the main differ- 6.
ence comparing to the cases under horizontal wind conditions. In [6] Joulain P. Behavior of pool fires: state of the art and new insights. Proc
addition, the flame detachment would lead to the increase of flame Combust Inst 1998;27:2691–706. http://dx.doi.org/10.1016/S0082-0784(98)
80125-2.
resistance and reduce the effect of wind on flame stretch. [7] Babrauskas V. Estimating large pool fire burning rates. Fire Technol
1983;19:251–61. http://dx.doi.org/10.1007/BF02380810.
[8] Chatris JM, Quintela J, Folch J, Planas E, Arnaldos J, Casal J. Experimental study
5. Conclusions of burning rate in hydrocarbon pool fires. Combust Flame 2001;126:1373–83.
http://dx.doi.org/10.1016/S0010-2180(01)00262-0.
The effects of different oblique winds on flame characteristics of [9] Woods J, Fleck BA, Kostiuk LW. Effects of transverse air flow on burning rates of
rectangular methanol pool fires. Combust Flame 2006;146:379–90. http://dx.
small square ethanol pool fires have been studied experimentally
doi.org/10.1016/j.combustflame.2006.02.007.
with a small-scale wind tunnel. The following conclusions can be [10] Hu LH, Liu S, Xu Y, Li D. A wind tunnel experimental study on burning rate
obtained: enhancement behavior of gasoline pool fires by cross air flow. Combust Flame
2011;158:586–91. http://dx.doi.org/10.1016/j.combustflame.2010.10.013.
[11] Ji J, Fan CG, Li YZ, Ingason H, Sun JH. Experimental study of non-monotonous
(1) The flame burning rate under downslope wind is relatively sidewall effect on flame characteristics and burning rate of n-heptane pool
larger than that under upslope wind with the same absolute fires. Fuel 2015;145:228–33. http://dx.doi.org/10.1016/j.fuel.2014.12.085.
wind speed and slope angles. At a given flow speed, the [12] Muñoz M, Arnaldos J, Casal J, Planas E. Analysis of the geometric and radiative
characteristics of hydrocarbon pool fires. Combust Flame 2004;139:263–77.
burning rate tends to decrease with the increase of attack http://dx.doi.org/10.1016/j.combustflame.2004.09.001.
angle under upslope wind, but increase with the increase [13] Jill MS, Thomas KB, Allen JR, Alexander LB. Characterization of thermal
of attack angle under downslope wind. radiation spectra in 2 m pool fires. Proc Combust Inst 2009;32:2567–74.
http://dx.doi.org/10.1016/j.proci.2008.06.044.
(2) The flame tilt angle tends to increase with the increase of [14] Raj PK. Large hydrocarbon fuel pool fires: physical characteristics and thermal
wind attack angle under downslope airflow conditions, but emission variations with height. J Hazard Mater 2007;140:280–92. http://dx.
decreases with the increase under upslope wind conditions. doi.org/10.1016/j.jhazmat.2007.07.051.
[15] Thomas PH. The size of flame from natural fires. Proc Combust Inst
At a given wind attack angle, the flame tilt angle generally 1963;9:844–59.
tends to increase with the increase of wind speed for both [16] LNG safety research program. Report IS 3-1. American Gas Association; 1974.
upslope and downslope winds. However, the rate of this [17] Moorhouse J. Scaling criteria for pool fires derived from large-scale
experiments. J Chem Sym 1982;71:165–79.
increase diminishes with increasing of the wind speed.
[18] Drysdale DD. An introduction to fire dynamics. 3rd ed.; 2011.
(3) The variation of flame length under different wind condi- [19] Pipkin OA, Sliepcevich CM. Effect of wind on buoyant diffusion flames. Ind Eng
tions is non-monotonous. For a given pan size under upslope Chem Fund 1964;3:147–54. http://dx.doi.org/10.1021/i160010a011.
flow conditions, flame length Lf initially decreases with [20] Welker JR, Sliepcevich CM. Bending of wind-blown flames from liquid pools.
Fire Technol 1966;2:127–35. http://dx.doi.org/10.1007/BF02588541.
increasing juw j, then increases before decrease again at about [21] Oka Y, Kurioka H, Satoh H, Sugawa O. Modelling of unconfined flame tilt in
juw j = 1.2 m s1. For the downslope flows, there appears to cross-winds. Fire Safety Sci 2000;6:1101–12. http://dx.doi.org/10.3801/IAFSS.
be only one turning point in each variation curve. With the FSS.6-1101.
[22] Hu LH, Liu S, deRis JL, Wu L. A new mathematical quantification of wind-blown
increase of wind attack angle, flame length will decrease flame tilt angle of hydrocarbon pool fires with a new global correlation model.
for upslope wind, but first increase up to hw = 20° and then Fuel 2013;106:730–6. http://dx.doi.org/10.1016/j.fuel.2012.10.075.
slightly decrease for downslope wind. In addition, fame [23] Lam CS, Weckman EJ. Wind-blown pool fire, Part I: experimental
characterization of the thermal field. Fire Safety J 2015;75:1–13. http://dx.
detachment occurs at wind speed about 1.44 m s1. doi.org/10.1016/j.firesaf.2015.04.009.
(4) Some modified semi-empirical correlations are presented for [24] Lam CS, Weckman EJ. Wind-blown pool fire, Part II: comparison of measured
determine flame tilt angle and flame length under oblique flame geometry with semi-empirical correlations. Fire Safety J
2015;78:130–41. http://dx.doi.org/10.1016/j.firesaf.2015.08.004.
winds except for small wind speeds. The modified Fr w [25] Tao CF, He YP, Li Y, Wang XS. Effects of oblique air flow on burning rates of
includes horizontal and vertical Froude number Fr h and Fr v square ethanol pool fires. J Hazard Mater 2013;260:552–62. http://dx.doi.org/
would be helpful for correlating the experiment data under 10.1016/j.jhazmat.2013.06.015.
[26] Zhu P, Wang XS, Tao CF. Experiment study on the burning rates of ethanol
oblique airflows.
square pool fires affected by wall insulation and oblique airflow. Exp Therm
Fluid Sci 2015;61:259–68. http://dx.doi.org/10.1016/
The current study is limited to hydrocarbon fires with small j.expthermflusci.2014.11.006.
[27] Calautit JK, Hughes BR. Measurement and prediction of the indoor airflow in a
pool size. Theoretical and empirical models need to be established
room ventilated with a commercial wind tower. Energy Build
on a more rigorous basis. 2014;84:367–77. http://dx.doi.org/10.1016/j.enbuild.2014.08.015.
[28] Calautit JK, Hughes BR. A passive cooling wind catcher with heat pipe
Acknowledgements technology: CFD, wind tunnel and field-test analysis. Appl Energy
2016;162:460–71. http://dx.doi.org/10.1016/j.apenergy.2015.10.045.
[29] Perta ES, Agizza MA, Sorrentino G, Boccia L, Pindozzi S. Study of aerodynamic
The authors appreciate the support of the Natural Science Foun- performances of different wind tunnel configurations and air inlet velocities,
dation of China (Grant Nos. 51574210, 51323010), the Anhui using computational fluid dynamics (CFD). Comput Electron Agric
2016;125:137–48. http://dx.doi.org/10.1016/j.compag.. 05.0007.
Provincial Natural Science Foundation (Grant No. [30] ANSYS FLUENT Theory Guide, Release 14.0. Southpointe: ANSYS, Inc.; 2011.
1408085MKL95) and the Fundamental Research Funds for the Cen- [31] ANSYS FLUENT User’s Guide, Release 14.0. Southpointe: ANSYS, Inc.; 2011.
tral Universities (Grant No. WK2320000035).

You might also like