Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 135 (2016) 261 274

Pool Fire Mass Burning Rate and Flame Tilt Angle under Crosswind in
Open Space
Ping Jianga,b, Shou-xiang Lua,*
a

State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, China
b
Shen Yang Aerospace University, Daoyi South Avenue 37, Shenyang, Liaoning 110136, China

Abstract
To study the influence of crosswind on pool fires mass burning rates and flame tilt angles in open space, aviation fuel pool fires were
investigated. Diameters of fuel pans were 0.1, 0.2, 0.3, 0.4, 0.5 and 0.6m in quiescent ambient air and 0.3, 0.4, 0.5 and 0.6m in crosswind
up to 4.71m/s. Taking advantages of a two axial flow fan wind wall system, small velocity increments in low crosswind velocity range
were obtained, which produced compact and abundant measured data under different crosswind velocities. It was found that the mass
burning rates variation process of a circular aviation fuel pool fire at different crosswind velocities could be divided into three stages as
the first increase stage, a radiation dominant stage, the second decrease stage, a radiation and convection combined dominant stage, and
the last increase stage, a convection dominant stage. First, based on experimental data of mass burning rates in quiescent air condition, the
two empirical factors in Zabetakis and Burgess Formula were obtained which could be used to predict mass burning rates for different
diameter fuel pan. Second, taking wind Froude Number as independent variable, a lognormal distribution regression equation to predict
the dimensionless mass burning rate was established which could be used as Fr Number is greater than zero and less than 0.75. Third, as
Fr Number was greater than 0.75 and less than 7.55, in consideration of that there was a thermal boundary layer between liquid fuel
surface and main flow of flame gas, neglecting variation of Prandtl Number, a relationship between Nusselt and Reynolds Number was
set up based on experimental data for the last stage and this correlation could be used to predict the forced convection heat transfer
coefficient between flame gas with different crosswind velocities and liquid fuel. Flame image sequences captured in qusi-steady period
were processed with a color image binarization program based on threshold of gray scale value to produce an iso-intermittency contour of
luminous flames and flame tilt angles were determined. Based on flame tilt angles of experimental data, parameters in different types of
regression equations generally used were calculated. According to the adjusted determination coefficient, correlations with the best
goodness of fit in terms of the experimental data were recommended.
2016
Authors.
Published
by Elsevier
Ltd. This
an openaccess
accessunder
articleCC
under
the CC BY-NC-ND
2016The
The
Authors.
Published
by Elsevier
Ltd.is Open
BY-NC-ND
license.license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of School of Engineering of Sun Yat-sen University.
Peer-review under responsibility of the organizing committee of ICPFFPE 2015

Keywords: Aviation fuel pool fire; mass burning rate; flames tilt angle; crosswind velocity; forced convection heat transfer coefficient

Nomenclature
D
D
dv
Fr
h
L

fuel pan diameter (m)

magnitude of wind drags (m)
volume of a gas element (m3)
2
wind Froude number, Fr uw
gD
coefficient of convective heat transfer (kw/m2K)
characteristic length (m)

* Corresponding author. Tel.: +86-551-6360-3141; fax: +86-551-6360-3141.

E-mail address: sxlu@ustc.edu.cn

1877-7058 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of ICPFFPE 2015

doi:10.1016/j.proeng.2016.01.122

262

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

Q / Ua c pTa D2 gD

Q*

dimensionless heat release rate, Q*

heat release rate (kw)

Ri

Richardson numbe, Ri

Re

Reynolds number, Re

TF
TB
TR

flame gas temperature (K)

temperature of liquid fuel surface (K)
reference temperature to evaluate thermal properties of gas (), TR =800

cp
g

air constant pressure specific heat (kJ/ kgK), cp=1.005 kJ/ (kgK)
acceleration of gravity (m/s2)
vaporization enthalpy of the fuel (kJ/kg), H=260 kJ/kg.

g L  Ua  U  T  Ta gD
or
2
U a uw 2
T uw
uw D / Q a

mcc

mass burning rate (g/ m2s)

n
Nu

output frequency of transducer (Hz)

Nusselt number, Nu
hD / O

Pr
uw

Prandtl number, Pr Q / a , Pr=0.713 at reference temperature

crosswind velocity(m/s)

u*

dimensionless wind velocity, u*

uw / gm" D /

1/3
a

Greek symbols

density (kg/m3)

flame tilt angle (e)

thermal conductivity of gas (kw/mK), =0.0718 kw/mK at reference temperature

Kinematic viscosity of the air (m2/s), = 1.55110-4 m2/s at reference temperature
extinction coefficient (1/m)
thermal diffusivity of gas (m2/s), =1.88810-4m2/s

Stefan-Boltzmann constant,
566910-8 Wm2K4

1. Introduction
Mass burning rate, flame radiation properties, such as emissivity and temperature, and flame geometrical characteristics
including height, shape and tilt angle etc. are dominant factors in determination of thermal radiation flux from pool fires [1].
A pool fire is characterized by the establishment of a diffusion flame with inenarrable chemical kinetics on top of a
horizontal fuel where buoyancy forces, baroclinicity, turbulence and molecular diffusion are all transport mechanisms[2].
So up to now, it still cant be described and solved fully from fundamental laws and experimental method is still the main
research means for people to described pool fires. It is even the case for pool fires under crosswind.
1.1. Mass burning rate
Previous burning rate measurements of qusi-steady period were, for the most part, made under calm conditions. Hottel[3]
stated that, for large diameter fires, the conduction term will become negligible, the convection term will be constant, and
the radiation term will be dominant and constant. The burning rate, therefore, should become constant for large diameter
liquid fires. For most fuels the data can be presented in predictive form as

m"

mf" (1  eNE D )

(1)

This form was first proposed by Zabetakis and Burgess[4] . It requires determining two empirical factors: mfcc and to
predict

mcc of a pool fire with diameter of D.

Compared to studies on pool fires in quiescent ambient condition, investigations on characteristics of pool fires under
cross air flow are not as abundant as the former. Some former investigator [4-9] indicated that the mass burning rates

263

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

increase monotonically, or even linearly, with the increase of cross air flow velocity, but some other investigators got a
contrary conclusion [3, 10, 11].
For a certain diameter pool under a certain crosswind velocity, the energy release rate from it, the mass burning rate, the
combustion efficiency, the flame temperature, emissivity, the flame geometry, such as its length and tilt angle, so is the view
factor, etc. are interdependent. The burning rate is determined by the rate of heat feedback to the liquid fuel. For a circular
pool, Hottel[3] expressed this rate of heat feedback as

mcc ~ qcc

4k TF  TB 
 h TF  TB   V F TF4  TB4 1  e NE D 
D

(2)

Equation (2) assumes that all the incident energy is absorbed by the fuel. Hottel[3] indicated that if the mechanism of
heat feedback as expressed in Equation (2) was considered, the decrease in burning rate was readily explained. For a
particular pool diameter, consider k, h, TF, and to be constant, the only remaining factor that could affect the rate of heat
feedback, hence the burning rate was the geometrical view factor. The geometrical view factor represents the radiant energy
returned to the surface of the pool, expressed as a fraction of the total radiant energy released by the flame. As the wind
velocity increases, the flame is tilted to greater angles and the view factor decreases, thus decreasing the rate of heat
feedback. So the effect of decreased radiative feedback was more noticeable for larger fires, because larger fires received a
greater fraction of energy feedback in the form of radiation than did small fires as shown by Hottel's analysis. Nevertheless
Zabetakis and Burgess[4] assumed that the effect of wind would increase the burning rate by increasing the extinction
coefficient . This same work also pointed out that increases in view factor could be important in increasing burning rates.
Whatever the case, the effect of wind was assumed to be one of increasing the burning rate.
Ping[12]pointed out that it was different from pool fires in quiescent ambient air conditions that for medium or larger
ones radiation was the predominant heat transfer mechanism, but the predominant heat transfer mechanism would be
changed under different cross wind velocities, which resulted in variation of heat feedback to liquid fuel. A little later, this
conclusion was verified further by a more detailed research by Longhua Hu[13] on the evolution of heat feedbacks of
conduction, convection and radiation in relatively small square pool fires (10-25 cm) with horizontal cross air flows ranged
in 03.0 m/s. Ethanol and heptanes were used as representative fuels, and a stagnant layer solution theory was then
proposed. Results in this study[13] indicated that fractions of conduction and radiation feedback didnt change obviously
when wind Fr Number exceeded about unity and convection became predominantly.
Generally the increase of surface flow velocity should increase the forced convection heat transfer coefficient if other
conditions remain unchanged. For larger fuel pans and wind velocities, more study is needed to examine the convective heat
transfer regularity and the forced convection heat transfer coefficients need to be quantified under different crosswind
velocities. Adopting some predecessors' points and the experimental data of this study, the effects of crosswind on factors
affecting heat transfer are analyzed. Assumptions that variations of heat conduction and radiation can be ignored when wind
Fr Number exceeds a certain value, the increase of heat feedback is solely caused by forced heat convection of crosswind
are made. According to the increase of mass burning rate and energy balance relationship, forced convection heats are
calculated. Based on theories of forced convection, assume that there is also a thermal boundary layer between hightemperature gas and liquid fuel like what exists between gas flow and a flat plate, then, there should be a certain relationship
among Nusselt number Nu, Reynolds number Re and Prandtl number Pr, that is:

Nu

F( Re,Pr )

(3)

Based on experimental data, a semi-empirical function is founded.

1.2. Flame tilt angle
Flame tilt angle , under wind conditions have been studied by several investigators for decades of years. Walker,
Sliepcevick[14] and Emori and Saito [14] derived correlations as Equation (4) from small-scale experiments. They correlated
the angle of inclination as a function of crosswind velocity, but the results do not compare well with larger-scale data [1].
Where Cf is dimensionless flame drag coefficient defined by Walker[14]. One target is to seek for the dimensionless flame
drag coefficient.

264

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

tan T
cos T

2C f uw2
U  Uf
S D a
Ua

(4)

Muoz [15] suggested that usually two types of correlations are used to predict flame tilt. The first type relates flame tilt
to the dimensionless wind speed u*, as
a ' u* b
 

'

cos T

u* t 1
u

(5)

The main drawback of these correlations is that for low wind velocity, they predict the flame tilt angle to be zero [1].
The second type of correlation expresses the tilt angles as a function of the wind Froude and Reynolds Numbers, as
tan
cos

c ' Fr d ' Ree '

(6)

Although these correlations allow the tilt to be determined at low wind velocities, but none of them led to good results [15].
J.A. Fay[16] also proposed an empirical correlation in terms of Froude Number, as
sin

Fr

Fr
0.19

(7)

But there was considerable scatter in the measured flame tilt angles to build up this correlation, and there were almost no
measured data for low wind velocity range.
Longhua Hu et al[17] suggested that for hydrocarbon pool fire the mass burning rate determines buoyancy strength of
the fire, which in competition to inertial force of the wind, in turn dominates the flame tilt angle. In fact buoyancy force on
fluid element depends on magnitude of (TF-Ta)/TF, and it has no definite correlation with mass burning rate. Experiments
were carried out for ethanol and heptanes square pool fires with dimensions of 10 cm, 15 cm, 20 cm and 25 cm in wind
velocity of 0-2.5 m/s(0Fr2.3), but also only a limited number of experiments were in wind velocity of 0-0.5m/s[17].
Because mass burning rates and flame tilt angles of pool fires are both very sensitive to variation of velocities in low
velocity range[16] and the two interact with each other, there might be some significant facts about the pool fire mass
burning rate and flame tilt angle were not brought to light by former investigators. One possible cause for this problem to
arise might be that the crosswind velocity increments (generally 0.5 m/s or larger [13, 14, 17]) in pool fire experiments were
too big, especially in low velocity range.
2. Experiment
2.1. Experiment facility
Aviation fuel is a kind of common fire hazard liquid on some occasions such as on an aircraft carrier deck or trial flight
airfields, mass burning rates and flame tilt angles of circular aviation fuel pool fires were investigated in crosswind field
supplied by a wind wall system. The height and width of its cross section equal to 2m and 1m. The schematic of the
experimental setup is shown as Fig. 1(a). The wind wall system is located in a large space building with length width
height = 20 m 12 m 20 m in its three dimensions.
The wind wall system consists of two axial flow fans (1.5 kW each, total wind pressure 191 Pa), honeycomb structure for
air flow smoothing, framework structure, outside shell and two power transducers for each of the electromotor of the fans.
The output frequency of the two power transducers can be adjusted from 0Hz to 50Hz, the adjustment accuracy of the
output frequency is 0.1Hz, and the rotation rates of the fans from 0 to 1450rpm (round per minute) can be obtained by
adjusting the output frequency of the transducers. As in Fig. 1(b), smoothed air stream flows out of B-B opening and blows
downward freely. The B-B section is rectangular in shape with its width height = 1.0 m 2 m.
Horizontal air flow velocities at uniformly distributed points on cross section A-A were measured with an anemometer.
Take the average velocity on the measuring points as crosswind velocity uw. The best fit formula of uw derived from
measured data to n is as

265

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

uw

0.102n  0.388

(8)

Equation (8) is applicable for n>3.8 Hz for there was kinetic energy loss in the honeycomb structures.
Before igniting the pool, the fans were started to reach its steady state and the crosswind velocity were not adjusted in an
experimental duration. For all of the experiments the distance between the fuel pan centre and the B-B cross section was
always set to 4 meters. As in Fig. 1(c), a fire resistant board with a circular hole in its centre was sported above the
electronic balance. The accuracy of the electronic balance is 0.1g and the sampling interval was set to 1s. A CCD camera
was set orthogonal to the direction of crosswind to record the instantaneous changes of the pool fire flames.

Fig. 1. Schematic of experimental setup (a), (b) top view and (c) vertical section of fuel pan and electronic balance lay out.

2.2. Experimental data processing method

2.2.1 Mass burning rate
Fig.2 presents a typical time-mass variation curve in an experiment and instantaneous mass burning rates derived from it.
It can be seen that the instantaneous mass burning rate changes very little about one minute after the test begins and the pool
fire reaches its so called quasi steady period. Mass burn rates of the pool fires in the steady period are calculate by averaging
the data in the steady-state and then divided by the areas of the circular fuel pans.
2.2.2 Time-average tilt angle from continuous flame images
The flame image sequences in the quasi steady period were captured in KMplayer. The durations of the capturing time
were all 10s and 300 continuous frames images were obtained for each test in its steady period. Each flame image was
processed by a color image binarization program based on gray scale value on each pixel dot to produce the binarization
image of the luminous flame. Examples of flame binarization images are shown in Fig. 3.
400

2.5

mass burning rate 300

fuel mass left

2.0

200
1.5
100

1.0

0.5
-20

mass left(g)

mass burning rate(g/s)

3.0

20 40 60 80 100 120 140 160 180

time(s)

Fig. 2. Time-mass curve and instantaneous mass burning rates

Fig. 3. Binarization processing of luminous flame

By accumulating the 300 binarization images the iso-intermittency contour were acquired, as shown figure 4. The tilt
angle being investigated in this paper refers to the tilt angle of the main fire plume, which could still retain its characteristic
of axial symmetry and it could be verified from the iso-intermittency contour as shown in Figure.4.
3. A concise derivation about Ri and Fr Number
The flame gas elements begin to accelerate upward rapidly when buoyancy force acts on it, which owes to gas density
difference to surrounding fresh air. The streamline of the incandescent gas element is embodied globally by the luminous
flame trajectory, so the tilt angle of the flame can be taken as the angle between the tangential of gas element streamline and
the vertical line. The crosswind velocity can be taken as the horizontal velocity of the gas in fire plume [21].

266

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

Fig. 4. Flame iso-intermittency contour and tilt angle and flame wind drag

For a gas element of volume dv, net buoyancy force acting on it is FB, and
FB ~  Ua  U  gdv

(8)

Horizontal inertia force of the gas element is FI, and

FI ~

U a uw 2 dv

(9)

Divide Equation (9) by Equation (10) and replace characteristic length L with fuel pan diameter D, it becomes
U a uw 2
FI
~
FB g D  Ua  U 

(10)

The reciprocal of right hand side of Equation (10) is Richardson Number Ri. The relative magnitude of the buoyancy and
inertia forces in flows can be captured by Ri Number. Using ideal gas law under constant pressure condition, Ri-1 will
becomes
Ri 1

U a uw 2
g D  Ua  U 

TF uw 2

TF  Ta gD

(11)

For high temperature flame gas, the ratio TF/(TF-Ta) approaches to a constant and it is not sensitive to gas temperature
variation. Take TF as a constant will not bring much error to TF/(TF-Ta). TF in hydrocarbon pool fire flame in open space
depends on excess air coefficient, ''combustion heat of the fuel, radiation heat fraction and combustion efficiency etc, and it
has no definite correlation with m . The highest measured gas temperature of large aviation fuel pool fires under crosswind
was roughly 1400 K[21], but for medium size pool fires, TF might be a little lower, for the lessening of smoke obscuration
effect [22].If TF changes from 1100K to 1300K, the ratio will change from 0.734 to 0.775, both less than 3% compared to the
value at 1200K, 0.756 as it is.
Combine Equation (4 ) and (11) , then:
tan T
cos T

2C f uw2
U  Uf
S D a
Ua

2C f

Ri1

(12)

Equation (12) indicate that tan/cos could be linearly related to Ri-1, this still need experimental verification and the
dimensionless flame drag coefficient Cf need to be determined based on the experimental data.
In fact, the dimensionless wind Froude Number Fr, representing the ratio between inertia forces to gravity forces can be
expressed as:
Fr

uw 2
gD

(13)

267

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

From Equatio (11) and Equation (13) indicate that wind Fr Number is in proportion to reciprocal of Ri-1, so if the flame
tilt angle could be scaled with Ri-1, it should be scaled with Fr Number, which will be discussed combining with the
experimental data later.
4. Results and Discussion
4.1. Mass burning rates results
4.1.1 Mass burning rates in quiescent ambient air conditions
Mass burning rates in quiescent ambient air conditions are presented in Figure 6. According to the data in Figure 6, the
regression equation can be written as

m"

30.46(1  e1.824 D )

(14)

"

So the estimated maximum burning rate mf of the aviation fuel is 30.46g/ (m2s), and is 1.82427m-1.
4.1.2 Mass burning rates under different crosswind velocities
As shown in figure 7, under the same cross air flow velocity, for different fuel pans, the mass burning rates of pool fires
increase with the increase of fuel pan diameters. With the increase of cross air flow velocities, the mass burning rates of a
pool fire do not vary monotonically, while the mass burning rates of different pool fires vary analogously.
24

m3''

32
30

20

28

mass buring rate(g/m s)

Mass burning rate(g/m2s)

34

mass burining rate

regression function

22

18
16
14
12
10

Model

emissivityFunction1 (User)

Equation

a*(1-exp(-b*x))

Reduced
Chi-Sqr

3.69473

Adj. R-Square

0.86367

mass burining a
rate
b

0.2

0.3

0.4

m''2

22
20
18
16
14
12

11.48265

10

1.82427

1.05839

0.5

m''0

24

30.46023

6
0.1

26

Standard Error

Value

dia.=0.3m
dia.=0.4m
dia.=0.5m
dia.=0.6m

m1''

36

6
-0.5

0.6

0.0

0.5

1.0

Fuel pan diameter(m)

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

crosswind velocity(m/s)

Fig. 5. Mass burning rates in quiescent ambient air.

Fig. 6. Mass burning rates under different crosswind velocities.

4.1.3 Analysis of the influential factors on pool fire burning rate under crosswind
The characteristic mass burning rates are listed in Table 1. For each pool diameter, with the increase of crosswind
"
velocities, the mass burning rates increase abruptly to a maximum value m1 at a very low cross air flow velocity, then
"
decrease promptly to a minimum value m2 . For the last stage, as the crosswind velocities continue to increase, the mass
"
burning rates increase to m3 . So the variation process of mass burning rates with crosswind velocities can be divided into
three stages as the first beginning increase stage, the second decrease stage, and the last increase stage.
Table 1. Characteristic mass burning rates
D(m)

m0(g/m2s)

m1(g/m2s)

u1(m/s)

Fr

m2(g/m2s)

u2(m/s)

Fr

m3(g/m2s)

u3(m/s)

Fr

0.3

12.16

15.83

0.33

0.04

33.8

7.8

1. 1

0.40

59.2

24.2

4.7

7.55

85.2

0.4

14.48

22.58

0.53

0.07

29.1

13.33

1.3

0.43

61.1

27.59

4.7

5.66

85.0

0.5

16.71

29.36

0.53

0.06

26.1

16.37

1.9

0.74

66.7

28.69

4.7

4.53

84.6

0.6

22.16

34.19

0.53

0.05

11.6

19.71

1.5

0.38

53.5

29.95

4.7

3.78

84.4

Equation 2 indicates that burning rate of a pool fire depends on at least 7 factors listed in Table 2, and analysis according
to experiments and predecessors study is listed herein.
The view factor between the flame and the liquid fuel surface equals to the ratio between the incident heats from the
flame to the liquid fuel surface to the total radiation heat released by the flame. The incident heat that reaches the fuel
surface comes from any position within the whole flame volume, so the view factor depends on the geometrical relationship
between the flame and the fuel surface.

268

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

From the CCD images (appendix A) of the pool fires it was found that in the first beginning stage, the length of the
flames has a perceptible enlargement at very low cross air flow velocities compared to the flame length in quiescent ambient
condition. This is because that the low velocity cross air flow enhances the shear strain of fluid particles in pool fire plumes,
that is, more small scale vortexes may be created locally at the peripheral interfaces of the air and fire plume flows because
of the interaction of horizontal air stream and gas flow in fire plumes. This facilitates the air entrainment at the edge of the
fire plumes and improves the mixture of fresh air and fuel vapor in flame base so as to the combustion efficiency.
Considering the above factors and the display and state of the fire plumes under low velocity air flow, it could be deduced
that the maximum temperatures in fire plumes also have a considerable elevations, so does the amount of radiation heat
back to the fuel in the pans.
Table 1 and FigureA2 shows that at low velocity air flows when the maximum mass burning rate appears, the tilt angle
of the pool fires flames are about 26, so it can be deduced that the effect of shape factor lessening of the flames to the pan
on the amount of radiation heat back to the liquid fuel in the pan is not very much compared to the effect of temperature
elevation of the pool fire flames. After this, as the air flow velocities continue to increase, the amount of radiation heat back
to the liquid fuel in a pan lessens significantly because the flame tilt angels increase obviously so the flame view factors to
fuel pans decrease a lot. For example, for the 0.5 m diameter pool fire, the minimum mass burning rate appears when the air
flow velocity is about 1.9 m/s and the flame tilt angle exceeds 67. At the same time, the temperature of fire plumes may
decrease to some degree resulting from the fact that too much fresh air mixes with fuel gas under the present air flow
velocity. So in the second stage, the variations of flame temperature and the view factor make the radiation heat feedback
decrease promptly, and the minimum mass burning rate appears.
The convective heat transfer coefficient would increase with the increase of crosswind velocity near the liquid surface.
This makes the mass burning rate increase rapidly.
Table 2. Analysis of influential factors on pool fires burning rates under crosswind
factor

Variation trend along with increase of crosswind velocity

Influence on heat feedback

Generally k reduces 15% per 100K with the increase of liquid fuel temperature, but the
liquid fuel temperature decreases in 15K about with the increase of crosswind velocity

Could be neglected.

Flame emissivity increases exponentially with D. For large diameter fires, the conduction
term becomes negligible; the larger the diameter is, the less the conduction term increase
with the increase of crosswind velocity.

Conduction variation could be

neglected for large diameter fires
under different crosswind velocity.

TF

Gentle breezes contribute to the mixing of air and fuel gas and the improvement of the
combustion efficiency, which observably increases the flame temperature, so does the
radiation heat feedback for the radiation heat scales with four times power of flame
temperature. Further increase of crosswind velocity makes the excess air coefficient
exceed one, and TF decrease a lot.

Burning rate increases evidently in

low crosswind velocity so does
the flame length, but when TF
decrease, the effect of temperature
change can be neglected.

TB

liquid fuel temperature decreases in 15K with the increase of crosswind velocity

Could be neglected.

Flame length increases in low crosswind velocity range because of the increase of
burning rate so the view factor may increases; But the deflection of the flame has the
opposite effect on view factor. The two opposite effects may counteract each other as the
tilt angle is less than 30e. Further investigation is needed to determine the two opposite
effects on the view factor. As the wind velocity increases further, the flame is tilted to
greater angles and the flame length lessens, so the flame view factor decreases, thus
decreasing the rate of radiation heat feedback as the flame tilt angle is about in 30-65e.
The effect of decreased radiation feedback is more noticeable for larger fires, because
larger fires receive a greater fraction of their energy feedback in the form of radiation
than do small fires. The flame tilt angle is very sensitive to variation of crosswind
velocity in low velocity range, but as the flame tilt angle is greater than 65e(Fr>0.75),
flame tilt angle increases slowly with increase of crosswind velocity, it can be assumed
that the view factor change a little when its tilt angle is greater than 65e.

As flame tilt angle is less than

30e, increase of view angle and
flame temperature lead to the
increase of radiation feedback; As
flame tilt angle is in the range of
30 e ~65 e , decrease of flame
view factor and temperature leads
to decrease of radiation feedback;
As flame tilt angle is greater than
65 e , view factor and flame
temperature both change a little,
so does the radiation feedback.

1  e NE D

The effect of crosswind to flame emissivity is still not clear and might increase with
wind velocity.

Assume that
negligible.

Convective heat transfer intensity increases with increase of the crosswind velocity and
its fraction might increase from 10-25% to 60-80% of the total heat feedback as Fr
Number increase from zero to 7.55 about.

Convective heat increases with the

increase of crosswind velocity.

4.1.4 Dimensionless mass burning with wind Fr Number (Fr<0.75)

the

change

is

269

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

As analysis in table 2, for the first and second stage the wind Fr Number is the key factor, so take each m0 of different
diameters as the normalization factor, dimensionless mass burning rate results versus Fr Number and lognormal distribution
regression equation are shown in figure 8. Dimensionless mass burning rates fit approximately to lognormal distribution
indicates that low crosswind velocity range is also a sensitive interval for mass burning rate so as for flame tilt angle.
dimensionless mass burning rate
LogNormal Fit of Sheet1 B"dimensionless mass burning rate"
1.8

D=0.3m
D=0.4m
D=0.5m
D=0.6m

1.6
1.4

dimensionless mass burning rate

dimensionless mass burning rate

1.8

1.2
1.0
0.8

(a)

Model

LogNormal

Equation

y = y0 + A/(sqrt(2*PI)*w*x)*exp(-(ln(x
/xc))^2/(2*w^2))

Reduced
Chi-Sqr

0.02112

Adj. R-Square

0.64205

1.4

Value
y0
dimensionless xc
mass burning
w
rate
A

1.2

Standard Error

0.93237

0.05066

0.08372

0.01869

0.85386

0.13638

0.07065

0.01261

1.0
0.8
0.6

0.6
-0.1

1.6

0.0

0.1

0.2

0.3

0.4

Fr

0.5

0.6

0.7

0.8

-0.1

0.0

0.1

0.2

0.3

(b)

0.4

0.5

0.6

0.7

0.8

Fr

Fig. 7. Normalized results of mass burning with Fr Number (a) normalized results of different D and (b) lognormal distribution regression equation

4.1.4 Convective heat transfer coefficient(0.75< Fr< 7.55)

To analyze the experimental data from perspective of forced convection theories of external flow (as shown in Figure 9)
as the Fr Number exceeds 0.75, take the liquid surface as a flat plate to inspect the correlation of Nu Number with Re and Pr
Numbers where the thermal properties of flame gas are evaluated at 800 and choose the fuel pan diameter D as
"
characteristic length. Based on study by Longhua Hu [23], assume that contribution of convective heat fraction to m2 is
about 67.5% and after this point, the increase of mass burning rate solely arise from the increase of convective heat. Assume
the temperature difference between the liquid fuel and flame gas is 600, 700, 800 and 900K for fuel pan diameters of 0.3,
0.4, 0.5 and 0.6m. An attempt was made to seek for a correlation fitting for Equation (4). The results are shown in figure 10.
Nu
Allometric1 Fit of Sheet1 J"Nu"

70

60

Nu

50

40

30

Model

Allometric1

Equation

y = a*x^b

Reduced
Chi-Sqr

7.3086

Adj. R-Squar 0.9687

Value

20
Nu

Standard Err

0.0119

0.00417

0.8860

0.03776

10
2000

4000

6000

8000

10000 12000 14000 16000 18000

Re

Fig. 8. External forced convection on the liquid fuel surface under crosswind. Fig. 9. Experimental results of Nu and Re Number and regression equation.

Then the regression equation is as

Nu

0.0119 Re0.886

(15)

Figure.10 indicates that the power function of Equation (15) fit the experimental data well and it can be used to predict
the forced convection heat transfer coefficient under different crosswind velocities.
4.2. Flame tilt angle results and discussion
4.2.1 Flame tilt angle results under different crosswind velocities
Flame tilt angle results under different crosswind velocities are presented in figure 11. It can been seen that there are
three regions for the tilt angle increasing process with the increase of crosswind. The first region is the initial rapid
increasing region and the tilt angles change from zero to about 55e, corresponding to 0-0.301 for Fr Number and 0-0.398
for Ri Number. In this region, flame tilt angles almost increase linearly with the increase of Froude or Richardson Number.
The increasing curves in this region are very steep indicating that flame tilt angles are very sensitive to the variation of
Froude or Richardson Number and the slope of the curve are very important to the accurate prediction of flame tilt angle.
The second region is the transition region and the tilt angles changes from about 55eto 77e, corresponding to 0.3011.685 for Fr Number and 0.398-2.227 for Ri Number. In this region, the curves make an arc turn. Afterwards the tilt angles
begin to increase very slowly with the increase of Froude or Richardson Number and this is the third incremental region.

270

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

(a)

(b)

(c)

100

100
80

80

60

dia.=0.3m
dia.=0.4m
dia.=0.5m
dia.=0.6m

60

40

40

40

20

20

20

dia.=0.3m
dia.=0.4m
dia.=0.5m
dia.=0.6m

60

tilt angle

D=0.3m
D=0.4m
D=0.5m
D=0.6m

tilt angle

Flame tilt angle

80

-1

Crosswind velocity (m/s)

10

1/Ri

Fr

Fig. 10. Experimental results of flame tilt angles (a) under different crosswind velocity and (b), (c) with Fr or Ri-1.

4.2.2 Regression equations and the one with the best goodness
Regression equations are as the followings.
x The first type as shown in Figure 12.
tan
cos

c ' Fr d ' Ree '

(16)

11.11
0. 3 7 Fr1.15948
Re0.38181
0.1317

Table 3. Parameters for Equation 16

Investigators

c'

d'

e'

Ref.

Moorhouse

1.9

0.339

0.05

[8]

Johnson

0.7

0.428

0.109

[9]

Pritchard and Binding

0.666

0.333

0.117

[10]

Rew and Hulbert

3.13

0.431

This study

0.1317

1.15948

[11]
0.38181

Adj. R-Square=0.99149

x The second type as shown in Figure 13

tan T
cos T

2C f

Ri1 10.7526 Ri1

(17)

So the average dimensionless flame drag coefficient Cf is 16.89.

120

tan/cos
Linear Fit of Sheet1 B"tan/cos"

tan/cos
Linear Fit of Sheet1 B"tan/cos"

120

100
100

80

40

Equation

y = a + b*x

Weight

No Weighting

tan/cos

tan/cos

80

60

Residual Sum 1487.45092

of Squares
0.98421
Pearson's r

Intercept
Slope

Standard Error
0

--

10.75261

0.23116

0.96821

20

Value
tan/cos

Intercept
Slope

Standard Erro
0

--

14.14817

0.30416

1/Ri

Fig. 11. tan/cos and Re, Fr Number

x The third type as shown in Figure 14.

y = a + b*x

No Weighting
Weight
Residual Sum 1487.44958
of Squares
0.98421
Pearson's r
Adj. R-Square

Value
tan/cos

Equation

40

0.96821

Adj. R-Square

20

60

10

-1

Fr

Fig. 12. tan/cos and 1/Ri, Fr Number

271

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

9.45 1/ Ri 

sin T

0.9896

1  9.3 1/ Ri 

(18a)

0.9896

12.4 Fr 0.9896
1  12.2 Fr 0.9896

sin T

(18b)

x The fourth type

a '  u* 

cos T

b'

0.75  u * 

0.28

(19)

Table 4 parameters for Equation 19

a

notes

ref

M. Muoz

0.96

-0.26

Large gasoline and diesel pool fires

[5]

AGA

-0.50

LNG pool fires

[6]

Thomas

0.7

-0.49

Wooden crib fires

[7]

Moorhouse

0.86

-0.25

Cylindrical flames

[8]

This study

0.75069

-0.28186

Aviation pool fires

cosine
Allometric1 Fit of Sheet1 L"cosine

1.0
1.0

1.0

sine

0.6

0.4

Model

LangmuirEXT1

Equation

y = (a*b*x^(1-c))/(1 + b*x^(1-c))

Reduced
Chi-Sqr

0.00935

Adj. R-Square

0.90263

0.4

sine

Model

LangmuirEXT1

Equation

y = (a*b*x^(1-c))/(1 + b*x^(1-c))

Reduced
Chi-Sqr

0.00935

Adj. R-Square

0.90263

Standard Error

1.01646

0.03166

9.29783

2.68542

0.01044

0.10542

sine

0.2

Model

Allometric1

Equation

y = a*x^b

Reduced
Chi-Sqr

0.03385
0.6331

Adj. R-Square

Value
Value

0.2

sine

0.6

0.8

Standard Error

1.01646

0.03166

12.19976

3.85633

0.01041

0.10542

cosine

sine
LangmuirEXT1 Fit of Sheet1 C"sine"

0.8

sine
LangmuirEXT1 Fit of Sheet1 B"sine"

0.8

Value

0.6
cosine

Standard Erro

0.75069

0.02608

-0.28186

0.02692

0.4

0.2

0.0

0.0
0

1/Ri

10

-1

Fr

Fig. 13. sin and 1/Ri or Fr Number

0.0
0

10

12

14

u*

Fig. 14. u* and cos.

According to adj.R-Square (adjusted determination coefficient), equations (16) and (17) are the correlations with the best
goodness of fit in terms of the experimental data.
5. Conclusions
Pool fire mass burning rates and flame tilt angles of different sizes circular aviation fuel pool fires under crosswind in
open space were investigated. Taking advantage of a new invented wind wall system, small velocity increments in low air
flow velocity range were obtained, which produced compact and abundant measured data for the quantification. It is found
that the mass burning rates of a circular aviation fuel pool fire do not vary linearly or monotonically with the increase of
cross air flow velocities in an open space and the mass burning rates variation process can be divided into three stages as the
first beginning increase stage, the decrease stage, and the last increase stage. At the beginning, with the increase of cross air
flow velocities, the mass burning rates increase abruptly to a maximum value at a very low cross air flow velocities, then it
begins to decrease promptly to a minimum value. For the last stage, as the cross air flow velocities continues to increase, the
mass burning rates begin to increase with crosswind velocities. The first and second stage can be seen as radiationconvection combined control stage, and the last stage is mainly dominated by heat convection.
First, based on experimental data of mass burning rates in quiescent air condition, the two empirical factors in Zabetakis
mcc
and Burgess Formula, f and , were obtained, which can be used to predict mass burning rates for different diameters
fuel pan. Second, take wind Fr Number as independent variable, a lognormal distribution regression equation to predict the
dimensionless mass burning rate was established which can be used as Fr Number is greater than zero and less than 0.75.

272

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

Third, as Fr Number is greater than 0.75 and less than 7.55, in consideration of that there is a thermal boundary layer
between liquid fuel surface and main flow of flame gas, neglecting variation of Pr Number, a relationship between Nu and
Re Number was set up based on experimental data for the last stage and this correlation can be used to predict the forced
convection heat transfer coefficient with different crosswind velocities of liquid fuel. Based on flame tilt angles of
experimental data, parameters in different type of regression equations were calculated. According to adj.R-Square, two
correlations with the best goodness of fit in terms of the experimental data were recommended. These two models all fit
well compared to measured data.
For the first and the second stages, further studies are needed to quantify the radiation and convection heat separately,
which involve temperature, emissivity and view factor between the flame and the liquid fuel under different crosswind.
Under high velocity cross air flow, the enhancement of evaporation by convection heat transfer becomes momentous.
This indicates that for different volatile combustible liquids, the effect of cross air flow would be quite different. Under a
relatively low velocity air flow, the mass burning rates may be enhanced by convection heat transfer evidently for high
volatile combustible liquids. So it is necessary to investigate pool fire mass burning rates of liquids with different volatilities.

Acknowledgements
It will be added later.

References
[1] C. Beyler., Fire Hazard., 2002. Calculations for Large Open Hydrocarbon Fires, The SFPE Handbook of Fire Protection Engineering, NFPA, Quincy,
MA, Chapter 3-11 .
[2]Pierre Joulain., 1998.The behavior of pool fires: state of the art and new insights, Twenty-Seventh Symposium (International) on Combustion/The
Combustion Institute, pp. 26912706.
[3] V. I. Blinov and G. N. Khudiakov., H. C. Hottel., 1959. Certain Laws Governing Diffusive Burning of Liquids, Fire Research Abstracts and Reviews,
Vol. 1 ,p 41-44.
[4] M. G. Zabetakis and D. S. Burgess., 1961.Research on the Hazards Associated with the Production and Handling of Liquid Hydrogen, Bureau of Mines
R. I. 5707.
>@J.R. Welker and C.M. Sliepcevich., 1966. Bending of Wind-Blown Flames from Liquid Pools, Fire Technology, 2, 2, pp. 127-135 .
>@I.Emori and K. Saito., 1983.Scaling Correlation and Smoke Observations of Oil Tank Fires Under Wind-blown Conditions, Chemical and Physical
Process in Combustion, Fall Technical Meeting, Combustion Institute/Eastern States Section, Providence, RI, Vol. 67, pp. 14 November .
[] M. Muoz, J., Arnaldos, J., Casal, E., 2004. Planas, Analysis of the geometric and radiative characteristics of hydrocarbon pool fires, Combustion
and Flame. 139,p263277
[] LNG Safety Research Program, Report IS 3-1, American, Gas Association (1974).
[] P.H. Thomas, The Size of Flames from Natural Fires, Ninth Symposium (International) on Combustion, Combustion Institute, Pittsburgh, pp. 844
859 (1962).
[] Johnson AD., 1992. A model for predicting thermal radiation hazards from large scale, LNG fires. I Chem E. Symp. Ser. 130, pp 507-524
[] M.J. Pritchard., T.M. Binding., 1992. Symposium on Major Hazards Onshore and Offshore, pp. 491-505.
[1] P.J. Rew., W.G. Hulbert., 1996. Development of Pool-Fire Thermal Radiation Model, HSE Books, Sudbury, Suffolk, p. 99.
[1] J.A. Fay., 2006 .Model of large pool fires, Journal of Hazardous Materials B136, pp219-232
[1] Longhua Hu et al, 2013. A new mathematical quantification of wind-blown flame tilt angle of hydrocarbon pool fires with a new global correlation
model. Fuel 106 p730-736
[1] T.Romen Singh et al., 2011. A New Local Adaptive Thresholding Technique in Binarization. IJCSI International Journal of Computer Science Issues,
Vol. 8, Issue 6, No 2, November.
[1] Lopez, A.R., Gritzo, L.A., and Sherman, M.P., 1998. Risk Assessment Compatible Fire Models (RACFMs). SAND Report 97-1562, Sandia National
Laboratories, Albuquerque, NM.
[1] Raj, P. K., 2010.A Physical Model and Improved Experimental Data Correlation for Wind Induced Flame Drag in Pool Fires, FIRE TECHNOLOGY;,
Technology and Management Systems, p579-609
[1] Kevin B. McGrattan., Howard R. Baum., Anthony Hamins., 2000.Thermal Radiation from Large Pool Fires, NISTIR 6546,
[1] Woods,JAR., Fleck,BA., Kostiuk,LW., 2006. Effects of transverse air flow on burning rates of rectangular methanol pool fires. Combust
Flame,146:37990.
[] Thomas K. Blanchat., Vernon F. Nicolette., W. David Sundberg., Victor G. Figueroa., 2006. Well-Characterized Open Pool Experiment Data and
Analysis for Model Validation and Development, SANDIA REPORT, SAND-7508
[] MILES GREINER., Radiation Heat Transfer and Reaction Chemistry Models for Risk Assessment Compatible Fire Simulations, Journal of FIRE
PROTECTION ENGINEERING, Vol. 16, May 2006.
[2] G.COX., combustion fundamentals of fire, Elsevier Science & Technology Books, 1995.
[2] J. Moorhouse., 1982. Scaling Criteria for Pool Fires Derived from Large-Scale Experiments, I. Chem. Sym., 71, pp. 165-179.
[2] Longhua Hu., Junjun Hu., Shuai Liu ., 2015. Evolution of heat feedback in medium pool fires with cross air flow and scaling of mass burning flux by
a stagnant layer theory solution, Proceedings of the Combustion Institute ,35p25112518
[2] Ping Jiang., Shouxiang Lu., 2013. Effects of cross air flow on mass loss rates of circular aviation fuel pool fires in large open space, Procedia
Engineering, 62 p 309 316

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274
[2] J. R. WELKER and C. M. SLIEPCEVICH., 1966. Burning Rates and Heat Transfer from Wind-blown Flames, Fire Technology, , 2(3):211-218
[2] F. Tang., L.J. Li., K.J. Zhu., 2015. Experimental study and global correlation on burning rates and flame tilt characteristics of acetone pool fires under
cross air flow, International Journal of Heat and Mass Transfer, Volume 87, August, P 369-375

Appendix A. Flame images under different crosswind velocities

u = 0 m/s

u = 0.12m/s u = 0.22 m/s u = 0.33 m/s u = 0.43 m/s u = 0.53 m/s u = 0.63 m/s

u = 0.73 m/s

u = 0.83 m/s

u = 0.99 m/s

u = 0.89 m/s

u = 1.1 m/s

u = 1.5 m/s

u = 1.3 m/s

u = 1.9 m/s

u = 2.5 m/s

u = 4.71 m/s
Fig. A1. Flame images under different crosswind velocities, dia. = 0.5 m.

273

274

Ping Jiang and Shou-xiang Lu / Procedia Engineering 135 (2016) 261 274

Fig.A2 iso-intermittency contour of luminous flame for diameter of 0.3, 0.4, 0.5 and 0.6m pool fires under u1, u2 and u3.

Ping Jiang. Data and place of birth: May 1974, Heilongjiang Province, China. State Key Laboratory of Fire Science,
USTC, China, Hefei, Dr.. Major Field of study: fire dynamics and fire risk assessment.

Te
Test4-16