You are on page 1of 29

Fire Technology

 2019 Springer Science+Business Media, LLC, part of Springer Nature.


Manufactured in The United States
https://doi.org/10.1007/s10694-019-00830-y

Optimal Safe Layout of Fuel Storage Tanks


Exposed to Pool Fire: One Dimensional
Deterministic Modelling Approach

Angan Sengupta*, Department of Chemical Engineering, Birla Institute of


Technology & Science, Pilani, Hyderabad Campus, Hyderabad,
Telengana 500078, India

Received: 5 July 2018/Accepted: 12 February 2019

Abstract. Fire hazard is one of the main risks associated to fuel storage tanks in
petroleum and in the petrochemical industries. Such a hazard includes pool fires in
the storage tanks or in the bunds, fire propagation from the source tank to target
tanks both in absence and in presence of wind, and also the cascading/domino effect
due to confined and unconfined vapour cloud explosion and or BLEVE associated
with the source tank. In the present work, a radiation shield of flowing water has
been introduced at a distance from the source fuel storage tank to prevent the dom-
ino effect originating from this source tank, under fire, to the target fuel storage
tanks in a tank farm. A simple one dimensional model has been developed from the
steady state energy balance to simulate the safe distances (i.e. rim–rim distance)
between fuel storage tanks containing class-I fuel (e.g. gasoline), both in presence and
absence of a water-shield under no-wind and cross-wind conditions. The model pre-
dictions have shown that the maximum safe inter-tank separation distance of 28.42 m
is anticipated at a wind velocity of 6 m/s, compared to 16.34 m in no-wind condition,
beyond which the centroid of the parallelepiped (a solid-flame geometry) falls outside
the base of the tilted flame geometry causing flattening of flame and a very sluggish
increase in the flame tilt angle as the inverse of the Richardson number in the pres-
ence of wind velocity vector increases. Furthermore, the present one dimensional
mathematical model has also been extended to show that introduction of a water-
shield with an appropriate thickness (dopt) is able to prevent the propagation of radia-
tion heat flux under both no-wind and cross-wind conditions to a much lower dis-
tance close to 8.34 m between tanks (measured from the centre of the source tank),
than those predicted from the existing empirical models; viz. Point Source model and
Shokrie-Beyler’s model.

Keywords: Safe inter-tank separation distance (Lsafe), Water-shield, Cross-wind, Flame tilt angle (h),
Flame drag (Ddrag), Domino effect

* Correspondence should be addressed to: Angan Sengupta, E-mail: angan@hyderabad.bits-pi-


lani.ac.in

1
Fire Technology 2019

1. Introduction
Fire and explosion hazards are common incidents associated with the fuel tank
farms in chemical, petroleum and petrochemical industries. The frequency of such
hazards, as has been reported in literature [1, 2], is increasing over years. Necci
et al. [3] have also developed a methodology, based on the past accident data
analysis, to identify the causes that trigger such accident scenarios. The Jaipur
incident in October 2009, where the 12 fuel storage tanks containing 105 kl of die-
sel and gasoline caught fire and the uncontrollable fire, which is left open to extin-
guish by itself over a period of 1 week (as there was no other alternative for the
fire fighters to combat the fire and explosions), resulted in several fatalities and
about 200 injuries besides the damage of 3 9 103 million INR worth property [1,
4]. Such an incident shows the importance of the proper layout of fuel storage
tanks in the petroleum and petrochemical industries, with optimal safe distances,
considering the fire catch situations with/without windy conditions. In literature
[1, 4–6], various standards and empirical models (viz. Point source model, Shok-
rie-Beyler’s model and Mudan’s model) have been proposed to calculate safe
inter-tank distances for preventing the domino effects (i.e. a low frequency cascad-
ing incident) during a fire spread. Researchers, therefore, have used these existing
empirical models to obtain the safe inter-tank distances under various fire and
explosion conditions [1, 6]. These available methods, however, do not take into
account the effect of wind on the flame height and the fire spread. Beer in 1991
has argued that the rate of spread of the fire-fronts are faster with wind speed
(uwind) ranging between 2 m/s and 6 m/s [7], which tends to increase the Lsafe val-
ues between the fuel storage tanks under fire catch conditions, due to the flame
drag (Ddrag) [8] in addition to the flame tilt [4, 8]. Under the cross-wind conditions
the standing fuel is swept in the downward direction of the wind which tilts the
flame over the edge of the pool (resulting in the flame drag [8]) and consequently
the fire-front approaches the target storage tank near to the pool, thus increasing
the radiative heat flux onto the impinged tank. The cross-wind, therefore, results
in an elongated flame base in the direction of the wind (Ddrag> diameter of the
fuel storage tank). Lautkaski [8] in his work has also referred to the possibility of
fire impingement on a nearby storage tank due to the flame drag, under high wind
velocity, which makes it difficult to cool the target tank with water spray; and that
was exactly the accident scenario during the Jaipur incident. Sengupta et al. [4]
had, therefore, modified the existing point source model to incorporate the effect
of cross-wind on the flame height and fire spread. It has been noted that great
variations exist among the values of the safe inter-tank distances, proposed by the
various standards and empirical models. Experiments on small pool fires (pool
diameter not > 0.6 m) are conducted by the researchers [9] to study the effect of
cross-wind velocity (uwind) on the flame tilt angle (h) and on the fuel mass burning
rate ðm_ 00 Þ. The researchers have shown that the flame tilt follows different correla-
tion with uwind values in the three distinct regions [9]. Lam and Weckman [10, 11]
in their recent research on buoyant pool fires have shown that while the flame
length in presence of wind is mainly governed by the fuel vapour density along
with the heat of combustion of the fuel and the fuel mass burning rate; the h
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

value, Ddrag value and the flame length are consistently obtained from the temper-
ature contours of the flame. Recently, researchers have also shown that the fuel
mass burning rate decreases with the pool size under certain cross-wind level con-
ditions; unlike that of no-wind conditions where m_ 00 value increases as the pool
size has been increased [12]. The quantification of the effect of cross-winds on h
value, flame length elongation, heat feedback mechanism and scaling of mass
burning flux for various hydrocarbon fuels in different pool size fires have also
been extensively studied experimentally [13–18]. A more realistic fire incident
involves the presence of multiple fire sources, which are the less investigated sce-
narios. The effect of cross-wind and fuel pan spacing on such multiple pool fires
in terms of its burning rate, flame tilt, flame length and flame merging have been
systematically studied by Fan and Tang [19] through series of experiments. Com-
pared to the empirical models, a better method based on graph theory and Baye-
sian network [20, 21], which is capable to determine the safe layout of tanks in a
tank farm; under fire spread conditions in presence and absence of wind effect has
been recently formulated by Khakzad and his research group. This new method
has also been deployed by the researchers in designing safe layouts and conduct-
ing risk assessment for tanks in a tank farm, taking into account the direction of
the cross-wind [21]. Due to the simplicity of the empirical models over the Baye-
sian network methodology, researchers had continued to use the empirical models
for calculating safe inter-tank distances, during a fire incident; and subsequently
developed heat transfer models to predict the thermal performance of the materi-
als used to construct the target tanks [22] and also to obtain the temperature pro-
files inside the target tanks [23] due to the fire impingement and the radiation
effect from the adjacent flame.
Different empirical models available in literature to calculate the incident radia-
tion heat flux from a fire, provides different safe distances that need to be main-
tained to prevent any domino effect caused by a pool fire. The calculation of
thermal radiation reaching a target, based on view factor times the black body
radiation was first used by Sullivan et al. [24] to model fire radiation from bush-
fires. This methodology for calculating the incident radiation flux from a pool fire
was later used by Zárate et al. [25] to prescribe safe distances for various scenarios
based on the critical value of the incident radiation heat flux. Subsequently, also
CFD calculations [10, 26] of temperature profiles for various fire sizes have been
performed. These results are then coupled with the modified radiation models
(that are based on detailed view factor calculations) to obtain safe separation dis-
tances from different pool fires [26]. However, the gaps in these researches remain
that though the radiation models include detailed view factor calculations and
estimations of flame drag and flame tilt are modified, still the effect of wind on
the calculation of safe separation distance has not been accounted for. In the pre-
sent paper, a simple one dimensional mathematical model has been developed in
absence and presence of wind (Sects. 2.1 and 2.2, respectively), based on the
energy balance at steady state. Initially, this one dimensional model is used to
obtain the precise value of the incident radiation heat flux between the fuel stor-
age tanks (10 m diameter) during a fire incident under no-wind as well as cross-
wind conditions and compare the results with data from the existing empirical
Fire Technology 2019

models and also with the available standards. Subsequently, in Sect. 2.3, novel
attempt have been made to obtain an optimal safe distance between the source
and the target tank (both with a diameter of 10 m) by extending the developed
one dimensional model equations to a scenario where a radiation shield made up
by flowing water layer has been introduced at a given distance from the source
tank. This methodology can, therefore, be used to design safe layout of a tank
farm exposed to fire. A schematic diagram for the system is shown in Fig. 1a. The
simulated results obtained from the present 1-D model under various scenarios are
discussed in Sect. 3, followed by the conclusion in Sect. 4.

2. Model Development
Calculation of the safe distances during a domino effect, caused due to fire inci-
00

dents, is retrieved from the critical value of the heat flux q_ st reaching the target
from the flame. The reported safe distance (i.e. the linear distance of a target
material at which it remains unaffected by the radiation from the adjacent flame)
for class-I fuels is where q_ 00st ¼ 4:732 kW=m2 [4] and for living beings is where q_ 00st ¼
1:4 kW/m2 [6, 25]. This critical value of the heat flux, which governs the safe sepa-
ration distance between target   and the source tank, is equal to the view factor
times the heat flux emitted q_ 00r from the flame [25]. In literature [1, 4–6], various
standards and empirical models (viz. Point source model, Shokrie-Beyler’s model
and Mudan’s model) have been proposed to calculate safe inter-tank distances for
preventing the domino effects during a fire spread. The simplest point source
model is based on the assumption that the total heat of the fire is radiated from a
single point [27] and that the incident radiation onto the adjacent target is homo-
geneous within an arc-radius determined by the view angle from the point fire
source. The most robust and accurate solid-flame geometry model [9, 25, 28],
which considers the visible flame to be a geometrical body and also takes into
account the radiation heat loss from the flame surface thus giving a better pre-
dicted value of the incident radiant heat flux onto the adjacent target [28], as com-
pared to the point source model that leads to an under prediction of the radiant
heat flux. Also it has been shown by Zarate et al. [25] in their work that the gen-
eralised way to approximate a flame-front with a plane surface corresponds to a
parallelepiped. In the present one dimensional (1-D) model the flame-front from
the pool fire is therefore, modelled as a parallelepiped (a solid-flame geometry),
both in absence and presence of wind, to calculate the safe inter-tank distance in a
tank farm exposed to fire.

2.1. Absence of Wind


Development of a 1-D steady state model helps in the determination of the radia-
tion heat flux, the temperature profile and thus provides the safe separation dis-
tance between the fuel storage tanks during a fire incident in a tank farm. A high
capacity fuel storage tank under fire and with no cross-wind velocity is a case of
an uncontrollable pool fire, where the radiation heat flux, the flame height, the
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

(a)
W
Zone – I A Zone – II
T
E
R
-
Source z z Target
y S y
Tank H
Tank
x I x
E
L
D
D D
3.33 m δopt

D/2 + Lsafe

(b)

z
y
x
Hf qr at x qr at x+∆x

∆x

(c)
As
b
Φ

a At

Figure 1. (a) Schematic diagram of class-I fuel storage tanks in a


tank farm under fire both in presence and absence of wind condition
and in presence of a water radiation shield. (b) Schematic
representation of the control volume of the flame front with height Hf.
q_ r00 is the radiant heat flux. (c) Schematic diagram of 3-D geometric
representation for calculating the view factor. b = flame length, a and
c are the equivalent dimensions of the top surface of the target tank
of diameter 10 m and U is the included angle between the flame-front
and the top surface of the target tank that varies between 90o (under
no-wind condition) and (90-h)o value (for tilted asymmetric flames)
under cross-wind condition.
Fire Technology 2019

emission rate and the safe separation distance between the storage tanks is com-
pletely dependent on the storage tank diameter (D) and on the fuel properties [4,
29]. To calculate the safe separation distance between the storage tanks the fol-
lowing one dimensional steady state energy balance equation can be written over
an elemental control volume of constant surface area and of thickness Dx (shown
in Fig. 1b)

d q_ 00r AT AC
¼ hr ðT Ta Þ hair ðT Ta Þ ð1Þ
dx V V

where d q_ 00r (a change in the transport flux) is the change in the amount of radiant
heat emitted per unit time from the constant unit surface of the elemental control
volume of the flame-front, hr is the radiation heat loss coefficient, hair is the natu-
ral convective heat loss coefficient of the surrounding air, AT is the total area
from which heat is predominantly being lost by radiation, AC is the cross-section
area for heat loss through natural convection and V is the volume of the control
volume, which is calculated as, V ¼ Dx  w  Hf ; where w is the dimension of the
control volume in the y-direction and Hf is the flame height in the z-direction,
which have been calculated using the Heskestad relationship [30]. The schematic
of the elemental control volume of constant surface area is shown in the Fig. 1b.
Following assumptions can be made for solving this steady state energy equation
of change:

(i) The radiation heat loss coefficient (hr) is a strong function of the absolute
value of the temperature (T).
(ii) Homogeneous temperature (Tf) at the surface of the flame, which does not
vary along flame height.
(iii) Ambient air temperature (Ta) is much less than the characteristic tempera-
ture, T, near to the source and also natural convection heat loss takes place
from the surrounding air layer along with the radiation heat loss.
(iv) The radiation heat flux is governed by the Stefan–Boltzmann’s law with emis-
sivity of air (ea) equals to 1 [31].
(v) w  Hf and Dx.

Using the above assumptions under steady state condition, the following ordi-
nary differential equation and the transcendental equation are obtained:

dT 2hr ðT Ta Þ hair ðT Ta Þ
4rea T 3 ¼ ð2Þ
dx Hf Hf

hr ¼ 4rea T 3 ð3Þ

The coefficient for the natural convection heat loss (hair) in Eq. (2) has been calcu-
lated as [32],
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

 1=6
2
0:387Raair
0:825 þ 8=27 kair
½1þð0:492= Prair Þ9=16 Š
hair ¼ ð4Þ
Hf

where kair is the thermal conductivity of the surrounding air, Prair is the Prandtl
number for the convective heat transfer and Raair is the Rayleigh number in the z-
direction, along the flame height, which takes into account the buoyancy effect
through the Grashoff’s number. Equations (2)–(4) along with the boundary condi-
tion (Eq. 5) complete the description of the tank farm under the no-wind fire
spread incident.

At; x ¼ 0 : T ¼ Tf ð5Þ

where Tf is the steady state surface temperature of the solid-flame geometry (refer
to Appendix for considering the surface flame temperature), in K, during the pool
fire under no-wind condition and is obtained from the conservation of energy
km_ still DHc 1=4
 00 
inside the flame as er [33].
Equations (2)–(5) is based on the steady state energy balance from the surface
of the flame-front of a fully developed homogeneous pool fire and the model
equations have been solved numerically by the fourth order Runge–Kutta method
in MATLAB, version 9.0.0.341360, using the ode15s solver to obtain the safe sep-
aration distance between the fuel storage tanks. The safe separation distance
(Lsafe) in no-wind condition for the class-I fuel storage tanks equals to a linear
distance in the x–y plane where incident radiation received by the target tank,
q_ 00st = Fst q_ 00r = Fst rea T4 ¼ 4:732 kW/m2 [4, 5]. In this study we have considered
two cases where: (a) the view factor Fst = 1 and (b) Fst has been calculated using
the parallelepiped geometry of the flame-front and also by considering the geome-
try of the target tank; as discussed in Sect. 2.4.

2.2. Presence of Wind


Under the cross-wind condition, the flame tilts in the downward direction of wind
by an angle h with the vertical, which can be obtained by the following empirical
relation [8, 34]:

for u  1

1
cosh ¼ 1 pffiffiffiffiffi ð6Þ
u for u  1

where u* is the dimensionless wind speed and is given by [8, 34],


uwind
u ¼   1=3 ð7Þ
gm_ 00wind D qv
Fire Technology 2019

where uwind is the cross-wind velocity (m/s), g is the acceleration due to gravity
(9.81 m/s2), qv is the fuel vapour density (kg/m3) and m_ 00wind is the mass burning
rate per unit pool area (kg/m2/s) under the cross-wind effect. Babrauskas [33] have
used the results of Blinov and Khudyakov to obtain a relation between the mass
burning rate and the cross-wind velocity, which is as follows:

m_ 00wind uwind
00 ¼ 1 þ 0:15 ð8Þ
m_ still D

The tilt in the flame, due to the effect of cross-wind; results in an oblique Carte-
sian co-ordinate system, x0 and z0 , in the direction of wind flow; such that

x0 1 tan h x
¼ ð9Þ
z0 cos ineh sec h z

where x and z are the orthogonal Cartesian co-ordinates and h is the flame tilt
angle. The solid-flame height (i.e. the height of the tilted parallelepiped flame-front
geometry) in this oblique co-ordinate is given by [35]
0:67
m_ 00wind 0:21
Hf ¼ 55D pffiffiffiffiffiffi ðu Þ ð10Þ
qa gD

The wind effect also elongates the flame base in the downward wind direction, so
that the new flame base diameter can be obtained from the following relation [8]
as suggested by Moorhouse

Ddrag
¼ 1:5Fr0:069 ð11Þ
D

where the Froude number (Fr) under wind condition is calculated as [8, 9]

u2wind
Fr ¼ ð12Þ
gD

In addition to the elongation of the flame base, wind velocity vector also tends to
reduce the flame height [4, 36] and develops an asymmetric flame along flame axis
(flame surface temperature facing the wind being different from that in the down-
ward direction of the wind flow) [7, 9], with convective
 heat losses (either forced
or natural) from the pool fire of increased size Ddrag > D together with the radia-
tive heat loss [7, 9]. The mode of convective heat loss is determined by the order
1=4
of the Ra ratio [37]. Inside this asymmetric flame the 1-D steady state energy
Re0:5 Pr1=3
balance equation, over an elemental control volume of thickness DxI in the oblique
co-ordinate system, can be written as
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

dTI 2hr ðTI Ta Þ hfuel ðTI Ta Þ


4reTI3 ¼ ð13Þ
dx0I Hf ðsechÞ þ xðcosinehÞ Hf ðsechÞ þ xðcosinehÞ

where x is calculated from Eq. (9), hr and hfuel are the radiative and convective
heat loss coefficient [37], respectively and can be obtained as follows:

hr ¼ 4refuel TI3 ð14Þ

8( )2
> 1=6
> 0:387Ra
fuel
< 0:825þ kfuel
>
>
9=16 8=27

hfuel ¼ ð
1þ 0:492= Prfuel Þ ð15Þ
> Hf ðsechÞ for natural convection
> 0:8 0:4
: 0:023Refuel Prfuel kfuel
>
for forced convection
>
Hf ðsechÞ

where kfuel is the thermal conductivity of the fuel, Prfuel is the Prandtl number for
the convective heat transfer in fuel and Rafuel is the Rayleigh number of the fuel
in the z-direction, Refuel is the Reynold’s number of the fuel flow and efuel is the
fuel emissivity, which have a value equal to 1 for all hydrocarbon fuels [27]. Equa-
tions (13)–(15) can be solved with the following boundary condition to obtain the
temperature profile inside the flame under wind condition:

At, x0I ¼ 0 : TI ¼ Tf ð16Þ

where Tf is the average flame temperature in K, along the centreline of the tilted
flame and can be calculated according to the Eqs. (17) and (18), based on energy
conservation inside the flame [33] as first shown by Hottel,

1=4
km_ 00wind DHc
Tf ¼ ð17Þ
er

k ¼ ð0:21 0:0034Ddrag Þ ð18Þ

where DHc is the heat of combustion of the class-I fuel in J/kg and k represents
the fraction of the total heat radiated from the flame [34]. Equations (13)–(18)
have been solved numerically by the fourth order Runge–Kutta method in
MATLAB, version 9.0.0.341360, using the ode15s solver to obtain the interface
temperature (TI) of the pool fire and the surrounding air.
In presence of wind, heat transfer across the surrounding air takes place by
convection in addition to the radiation mode of heat transfer, unlike in no-wind
condition where there is only the radiative heat transfer. However, the heat loss
from the surrounding air in presence of the wind velocity vector continues to be
through radiation and by natural or forced convection. The corresponding 1-D
Fire Technology 2019

steady state energy balance equation for the surrounding air over a control vol-
ume of thickness DxII in the oblique co-ordinate system is
_
 dTII 2hr ðTII Ta Þ hair ðTII Ta Þ
4rea TII3 þ uwind qa cpa

¼
dx0II Hf ðsechÞ þ xðcosinehÞ Hf ðsechÞ þ xðcosinehÞ
ð19Þ

where qa and cpa are the density and specific heat capacity of air, respectively. The
_
radiative heat loss coefficient (hr) and the convective heat loss coefficient (hair ) can
be calculated from Eqs. (14) and (15), using the properties of the surrounding air.
Equations (9) and (13)–(19) along with the boundary condition (Eq. 20) complete
the description of the tank farm under fire in presence of wind.
At, x0II = 0 (interface of the flame and surrounding air, refer to the Appendix
for considering the outer-surface temperature of the flame):


TII ¼ TI Ddrag ð20Þ
2

To obtain the safe separation distance in presence of wind for class-I fuel storage
tanks, which corresponds to a linear distance in x–y plane where the incident radi-
ation received by the target tank, q_ 00st = Fst q_ 00r = Fst rea T4 ¼ 4:732 kW/m2 [4, 5];
which have been presently obtained by solving Eqs. (9) and (13)–(20) numerically
in MATLAB using the fourth order Runge–Kutta algorithm. As mentioned in
Sect. 2.1, in this study we have considered two cases where: (a) the view factor
Fst = 1 and (b) Fst has been calculated using the parallelepiped geometry of the
flame-front and also by considering the geometry of the target tank; as discussed
in Sect. 2.4.

2.3. Introduction of Water Radiation Shield


The heat transfer during a fire incident is predominantly by the radiation mode in
absence of wind; while both radiation and convection heat transfer take place in
presence of cross-winds. The radiation heat flux from a source tank (i.e. the fuel
storage tank under fire) to the target tank (i.e. the fuel storage tank, which is at a
risk of the domino effect) can be greatly reduced by introducing a radiation shield
of appropriate thickness (dopt  Lsafe) and with high reflectivity (or low emissiv-
ity) [31]. The schematic of the tank farm layout under the given situation is shown
in Fig. 1a. The role of the radiation shield, with low emissivity and hence high
resistance to radiation heat transfer is to reduce the transfer of radiative heat flux
from the source tank to the target tank. The 1-D mathematical model to deter-
mine the safe inter-tank distance (Lsafe) in presence of a water radiation shield
with thickness d, has been developed considering the following assumptions under
steady state conditions:
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

(i) High mass flow  rate ofwater inside the water-shield, which has been placed at
a distance of D2 þ 3:33 m from the centre of the source tank, where D is the
pool diameter.
(ii) Heat transfer through the water-shield takes place by radiation and conduc-
tion, since (d/Lsafe 1), while the heat loss from the water-shield is by forced
convection.
(iii) The radiative heat flux is governed by the Stefan–Boltzmann’s law with emis-
sivity of water (ew) equals to 0.45 [31] and the conductive heat transfer can be
governed by the Fourier’s law of heat conduction.
 
(iv) The water-shield placed at the distance of D2 þ 3:33 m from the centre of
source tank is subjected to film boiling, in the presence of high temperature
pool fire, and the heat loss from the boundary of the water-shield close to the
pool fire takes place only through evaporation.
(v) The water radiation shield has been modelled in the orthogonal Cartesian co-
ordinate system both in the absence and presence of wind.

A distance of 3.33 m has been added to the radius of the source tank, to obtain
the position for insertion of the water-shield, as the minimum rim to rim distance
between the class-I fuel storage tanks as suggested by NFPA is 3.33 m [5]. The
steady state 1-D model which governs the temperature variation within the water-
shield has been obtained from the energy balance and is as follows:

dTw d 2 Tw hw ðTw Tin Þ


4rew Tw3 þ kw 2 ¼ ð21Þ
dxw dxw Hf

where Tw is spatial temperature, in K, inside the water-shield, Dxw is the thickness of


the control volume considered within the water-shield, Tin is the inlet temperature, in
K, of water in the water-shield, kw and hw are the thermal conductivity and forced
convective heat loss coefficient of water, respectively. The forced convective heat loss
coefficient (hw) inside the water-shield has been calculated by using the constitutive
relation in Eq. (22) and considering a high velocity falling liquid [38].

0:4 1=3
3 C v2w
hw ¼ 11:4  10 ð22Þ
lw gkw3

where lw is the viscosity of water (Pa-s), g is the acceleration due to gravity


(9.81 m/s2) and C is the water loading in the water-shield (kg-water/m/s) and is
calculated as,
C ¼ qw vw m_ 00ev d
 
ð23Þ

where m_ 00ev provides the evaporation rate of water per unit surface area of the
water-shield (kg-water/m2/s), which can be calculated from the energy balance
(Eq. 24), assuming evaporation at the surface of
 the water-shield by using the
radiation heat flux q_ 00rin at the linear distance of D2 þ 3:33 from the centre of the
 

source tank.
Fire Technology 2019

q_ 00rin
m_ 00ev ¼ ð24Þ
cpw ðTb Tin Þ þ kw

where cpw is the specific heat capacity of water (kJ/kg/K) at constant pressure of
the inlet water, kw is the latent heat of vaporisation of water (kJ/kg) and Tb is the
boiling point of the water in K. The amount of water inlet in the water-shield is
given by qw vw , where vw is the inlet velocity (m/s) of water and qw is the water
density (kg/m3). Thickness of the water-shield (d) at any instance, assuming no
eddy formation, is determined from the following expression [39]:
rffiffiffiffiffiffiffiffiffiffiffiffiffi
vw l w
d¼ 2 ð25Þ
qw g

Equation (21) has been further simplified into a set of two first order ordinary dif-
ferential equations, as given by Eqs. (26) and (27), and can be solved using the
boundary conditions given in Eq. (28), which accounts for the temperature and
the heat flux at the position of the water-shield insertion.

d T1 
¼ T2 ð26Þ
dxw

d T2 hw ðT1 Tin Þ


4rew T13 T2 þ kw ¼ ð27Þ
dxw Hf

where T1 ¼ Tw and T2 ¼ ddxTw1 . These set of ordinary differential equations can be
solved using the following boundary conditions:
At xw ¼ 0: temperature T1 ¼ T 0 and the
 00 
d T1  m_ kw q_ 00rin
¼ T2 ¼ ev ð28Þ
dxw kw

where T 0 is the temperature in K, at the linear distance of D2 þ 3:33 from the


 

centre of the source tank.


Equations (26) and (27) along with the Eqs. (22)–(25) and the boundary condi-
tions, as given by Eq. (28), are solved numerically by the fourth order Runge–
Kutta method in MATLAB, version 9.0.0.341360, using the ode15s solver to
obtain the temperature profile inside the water-shield and also to determine the
optimal safe separation distance between the fuel storage tanks; both in the
absence and presence of cross-wind velocity. The optimal safe separation distance
(Lsafe) for class-I fuel storage tanks, in presence of a water radiation shield corre-
sponds to a linear distance in the x–y plane where the incident radiation heat flux
received by target tank is q_ 00st = Fst q_ 00r = Fst rea T4 ¼ 4:732 kW/m2 .
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

2.4. Calculation of View Factor


As discussed in literature [40] at any point above the ground level where every-
thing viewed by the target is the complete flame, the view factor (Fst) is equal to
1, which provides the worst case situation. However, a minimum safe separation
distance between the source and the target, with 20% safety factor [25],can be
obtained by calculating the Fst value using the appropriate geometric considera-
tion of the flame-front, target tank and also by considering the flame tilt and the
ground slope; such that the radiation received by the target is
q_ 00st = Fst q_ 00r = Fst rea T4 [24, 25]. Also it has been shown recently that the view fac-
tor defined between ‘‘two dimensional pixel to a rectangular surface’’ is four times
greater than the view factor defined between two rectangular surfaces; and this
view factor was used to estimate the incident radiant heat flux onto a target by
moving the position of the emitting surface away from the axis of the pool and
also along the height of the flame [26]. However, this may result in a larger safety
factor as the geometric representation in this radiation model seems to be more
simplified without the consideration of an included angle between the flame-front
and the target as well as the effect of wind on the flame-front is not taken into
account. In this work the heat flux from the flame-front is propagated along the
linear distance between the source and the target fuel tanks using the 1-D model
equations, both in presence and absence of cross-winds. The shape factor at each
inter-tank separation distance is calculated using Eq. (29) that considers a 3-D
surface model for the radiation between two plane surfaces with a common edge
and an included angle U [41]:

sin2U p  A BcosU
U A2 + B2 + B2 tan 1

Fst ¼ ABsinUþ
4pB 2 BsinU
B AcosU
+ A2 tan 1
AsinU
( " # " #
2
1þA2 1þB2

sin U 2 2 B2 ð1þCÞ
þ 1 ln + B ln 
sin2 U 1 + B2 C

4pB 1+C
" #)
A2 ð1þA2 Þcos2U
+ A2 ln
Cð1 + CÞcos2U
pffiffiffiffi
1 1 1 A 1 1 C 1
þ tan þ tan tan 1 pffiffiffiffi
p B pB A pB C
sinUsin2U AcosU B AcosU
þ AD tan 1 + tan 1
2pB D D
2 0 1 0 13
ZB qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cosU B ncosU C B A ncosU C7
þ 1þn2 sin2 U4tan 1 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA + tan 1 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA5dn
6
pB 2 2
0 1þn sin U 1þn2 sin2 U
ð29Þ
Fire Technology 2019

where

A ¼ a/c
B ¼ b/c
ð30Þ
C ¼ A2 + B2 - 2ABcosU
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
D ¼ 1 þ A2 sin2 U

The schematic for this 3-D geometric representation to calculate the view factor is
shown in Fig. 1c. As pointed out in literature [24, 25] introduction of a view fac-
tor requires simplification of the scenario, such as the geometry of the flame-front
and the target. In the present work we have considered the flame-front as a paral-
lelepiped with plane surfaces (which represents a solid-flame geometry) and the
top surface of the target tank is also modelled as a plane surface; such that the
top surface area of the target tank is equal to a cross-sectional area corresponding
to a diameter of 10 m. Under no-wind condition the angle between the radiating
plane surfaces, U = 90 and hence Eqs. (29) and (30) reduces to the standard
equation for view factor calculation between two perpendicular surfaces [41, 42].
However, in the presence of cross-wind U = (90 - h), where the flame tilt angle
calculated from Eq. (6). The radiant heat flux at each point along the inter-tank
distance can be estimated using the governing model equations as described in
Sects. 2.1 and 2.2. The calculated value of Fst from Eqs. (29) and (30) is subse-
quently used to calculate the critical value of the radiant heat flux q_ 00st reaching
the target from the flame (as discussed in the Sects. 2.1 and 2.2); using the relation
q_ 00st = Fst q_ 00r = Fst rea T4 .

3. Results and Discussion


In this section we have reported and discussed the calculated inter-tank safe dis-
tance in absence and presence of cross-winds and also shown the heat flux varia-
tion over the inter-tank separation distance under no-wind and windy conditions,
obtained from the present 1-D model with Fst = 1 and with Fst < 1. We have
also reported the optimal safe distance between the class-I fuel storage tanks (cal-
culated from the present 1-D mathematical model) on introduction of a water
radiation shield.

3.1. Absence of Wind


The temperature profile over the inter-tank separation distance, in absence of
wind velocity, can be calculated using Eqs. (2)–(5). The corresponding incident
radiation heat flux (calculated as q_ 00st = Fst rea T 4 ), at various separation distances
between gasoline storage tanks is shown in Fig. 2. In Fig. 2 we have also com-
pared the incident radiation heat flux calculated from the present model with the
values obtained from the empirical models, viz. the Point Source model and the
Shokrie-Beyler’s model. It is clear that under no-wind condition, the incident radi-
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

ation heat flux decreases with an increase in the separation distance between the
fuel storage tanks. While the point source model gives the lowest value for the
Lsafe, the Shokrie-Beyler’s method and the present 1-D model predictions (assum-
ing parallelepiped geometry for the flame-front from the pool fire and with Fst =
1) for Lsafe values show a good agreement for the pool fire with diameter 10 m
and with no-wind condition (refer Table 1). The incident radiation heat flux from
the present 1-D model at the given tank-height above ground level has been calcu-
lated as q_ 00st = Fst rea T 4 where Fst = 1 or Fst < 1; whereas the Shokrie-Beyler’s
method uses view factor times the average emissive power (which is significantly
less and uncertain than the emissive power that can be ‘‘attained locally’’ [40]) to
estimate the radiant heat flux from a fire source that would be received by the tar-
get. Moreover, it has also been reported in the literature [40] that the major
uncertainty in the Shokrie-Beyler’s method predictions arises from the definition
of the average emissive power and not from the view factor model, which is used
in the Shokrie-Beyler’s method. However, for the value q_ 00st  5 kW/m2 (for which
the Shokrie-Beyler’s method prediction is recommended without any safety factor
[40]), we observe that the present 1-D model predictions of incident radiant heat
flux values, with Fst = 1, show an excellent agreement with that predicted by the
Shokrie-Beyler’s method. We also find that the fraction of incident heat flux radi-
ated (k) calculated by the present model is in excellent agreement with that
obtained from the point source model. The reported flame heights for all the
models are calculated using the Heskestad relationship [30]. A minimum safe sepa-
ration distance between the fuel tanks under no-wind condition has also been cal-
culated from the present 1-D model using the value of Fst = 0.1138 [calculated
from Eqs. (29) and (30)]. It is noted that the introduction of a view factor has
reduced the safe inter-tank distance to 7.01 m from 16.34 m; as the incident radi-
ant heat flux at the source reduces from 353.378 kW/m2 to 40.214 kW/m2 and this
still remains at a higher value than that predicted by the Shokrie-Beyler’s method
(refer Fig. 2), which uses the average emissive power definition [40] to calculate
the incident radiative heat flux.

3.2. Presence of Wind


In presence of a cross-wind velocity vector the flame of the pool fire gets tilted
and the flame tilt angle is governed by the dimensionless wind velocity, as given
by Eqs. (6) and (7). In the present study where the gasoline storage source tank
diameter is 10 m, an observable flame tilt (h = 9.65) occurs at uwind = 0.99 m/s.
Figure 3 shows the variation of the flame tilt with the inverse of the Richardson
number (Ri-1) under cross-wind conditions. We have observed that though h
increases monotonically with an increase in the Ri-1 values, there exist three dis-
tinct regions of variation: (i) region where the h value increases and rapidly from
9.65 to 45.95 (with much steep slope) as the Ri-1 value increases from 0.0123 to
0.0504 (this corresponds to the wind velocity ranging from 0.99 m/s to 2 m/s), (ii)
transition region where the h value changes from 45.95 to 66.09 with Ri-1 value
increasing from 0.0504 to 0.4528 (i.e. the wind velocity ranging from 2 m/s to
6 m/s); at this region the tilt angle versus the Ri-1 curve makes an observable arc-
Fire Technology 2019

Figure 2. Radiation heat flux versus inter-tank separation distance


from the rim of source tank (assuming homogeneous surface
temperature of the flame) in absence of wind, plotted in orthogonal
x–y plane.

Table 1
Safe Separation Distance Under No-Wind Condition Between the Fuel
Storage Tanks with 10 m Diameter And Containing Gasoline

Rim–Rim inter-tank
Models safe distance, Lsafe (m) Flame height, Hf (m) k References

Point Source 13.60 18.24 0.176 [4]


Shokrie-Beyler’s 16.25 18.24 — [4]
Present 1-D model 16.34 with Fst= 1.0 18.24 0.176 —
7.01 with Fst = 0.1138 18.24 0.176 —

turn and (iii) the region where the h value increases very slowly (showing a slug-
gish slope in the curve) with the inverse Richardson number and this corresponds
to uwind> 6 m/s. In this sluggish region the centroid of the tilted parallelepiped
geometry (reported in Table 2) has been observed to move outside the flame base
(determined by the flame drag value, Ddrag, under wind condition [8]), resulting in
the slower flame tilt. Stability of the tilted flame with centroid falling outside the
base of tilted parallelepiped geometry at uwind > 6 m/s (refer Table 2) will ulti-
mately result in complete flattening of the flame (Hf  flame length,
Lf ¼ Hf cos h) [4, 9, 36] and the inter-tank safe separation distance is then equals
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

the Lsafe value corresponding to the uwind = 6 m/s (refer Table 3). Thus, rapid fire
spread due to flame tilt and flame drag (i.e. elongation of the flame base) in
downward direction of wind for uwind = 6 m/s provide an upper limit to the flame
propagation and fire impingement onto the target tank (i.e. at uwind = 6 m/s,
Lsafe> Lf Hf) [7, 8]. Since an increase in wind speed beyond 6 m/s causes com-
plete flattening of the flame, thus lowering the possibility of fire impingement and
rapid fire spread; we therefore, have restricted our further study to the wind speed
ranging from 0.99 m/s to 6 m/s.
The incident radiation heat flux (calculated as q_ 00st = Fst rei Ti4 , where the sub-
script i represents the zone of heat transfer) over the inter-tank separation dis-
tance is shown in Fig. 4 and has been obtained from the temperature profiles
calculated from Eqs. (13)–(20). Each of the curves in Fig. 4a (where Fst = 1, i.e.
the worst possible situation) plotted for various wind velocities, initiate from x-
values that measures the actual deviation (on the orthogonal x–y plane) of the
centre of the elongated flame base from the centre of source tank under cross-
wind effect. This deviation significantly increases from 2.17 m to 20.8 m with
increasing wind speed from 0.99 m/s to 6 m/s that enhance the flame tilt (shown
in Fig. 3). Similar to no-wind condition the cross-wind incident radiation heat flux
decreases as the inter-tank separation distance increases; however, the formation
of an asymmetric flame in presence of cross-winds [9, 36] results in two different
zones of heat transfer (governed by the Eqs. 13 and 19), as shown in Fig. 4a with
two different slopes in the q_ 00st values. Increasing wind velocities results in an
increase of the convective heat loss from the surrounding unstable atmosphere,

Figure 3. Variation of flame tilt angle (h) with inverse Richardson


number (Ri21) in presence of cross-wind.
Fire Technology 2019

Table 2
Flame Tilt and Flame Drag Under the Cross-Wind Conditions for Pool
Fires in the Fuel Storage Tanks with 10 m Diameter and Containing
Gasoline

uwind (m/s) Ri-1 m_ 00wind (kg/m2/s) h (o) Ddrag (m) Centroid of solid-flame geometry (m)

0.99 0.0123 0.0445 9.65 10.92 1.45


1.30 0.0213 0.0447 30.58 11.33 4.76
1.50 0.0283 0.0448 30.69 11.56 5.84
1.80 0.0408 0.0450 42.90 11.85 7.03
2.00 0.0504 0.0452 45.95 12.03 7.67
2.50 0.0787 0.0455 51.49 12.41 8.95
3.00 0.1133 0.0458 55.31 12.72 9.95
4.00 0.2014 0.0465 60.40 13.24 11.54
5.00 0.3146 0.0471 63.71 13.65 12.79
6.00 0.4528 0.0478 66.09 14.00 13.86
7.00 0.6161 0.0484 67.91 14.30 14.80
8.00 0.8043 0.0491 69.35 14.56 15.66
8.50 0.9077 0.0494 69.98 14.69 16.06
9.00 1.0174 0.0498 70.57 14.80 16.48
9.50 1.1333 0.0501 71.06 14.91 16.82
10.00 1.2554 0.0504 71.53 15.02 17.18
12.00 1.8060 0.0517 73.12 15.40 18.56

which is represented in the plot by a steeper slope in the q_ 00st values in this zone of
heat loss in the surrounding air (refer Fig. 4a). Also for uwind = 3.0, 4.0 and
5.0 m/s both natural convection heat loss followed by the forced convection heat
loss in the surrounding unstable atmosphere is noted, which results for a second
steeper slope in the q_ 00st versus distance plot (as shown in Fig. 4) within this sur-
rounding air heat loss region; while for uwind = 6.0 m/s and uwind < 3.0 m/s, heat
loss from the surrounding air is predominantly by forced convection and natural
convection, respectively. The flame height of the pool fire and the safe inter-tank
distance during the pool fire in presence of wind are given in Table 3. It is seen
from Table 3 that the flame height (Hf) decreases with an increase in wind speed,
resulting a decrease in the fraction of radiant heat flux (k) and hence lowering the
value of safe inter-tank distance (Lsafe) as uwind increases from no-wind condition
to 1.5 m/s (comparing Tables 1 and 3), which is then followed by an increase in
the Lsafe value as uwind increases from 1.5 m/s till 6 m/s (Table 3). This effect of
cross-wind on the lowering of flame height and hence the radiant heat loss from
the pool fire is consistent with observations predicted in literature [4, 36]. Decrease
in the radiant heat loss from the pool fire is due to the onset of convective mode
of heat loss from the asymmetric flame with elongated flame base [7, 9], in the
presence of cross-winds (shown by Eq. 13). We also note an inverse relation of the
k value with the mass burning rate (Tables 2 and 3), as the wind velocity increases
from 0.99 m/s to 6 m/s. Such an inverse relation of the radiant heat fraction with
the mass burning rate had already been validated for large heptane pool fires [43].
This initial lowering of the Lsafe value in presence of wind (1.5 m/s ‡ uwind ‡ 0.99
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

Table 3
Flame Height and Safe Inter-tank Distance under the Cross-Wind
Conditions for Pool Fires in the Fuel Storage Tanks with 10 m
Diameter and Containing Gasoline

Flame Rim–Rim inter-tank safe Rim–Rim inter-tank safe n 9 104 n 9 105


uwind height, distance, Lsafe (m) with distance, Lsafe (m) with with with
(m/s) Hf (m) Fst = 1 Fst < 1 k Fst = 1 Fst < 1

0.99 12.76 15.23 9.84 (0.1183) 0.173 - 3.03 - 3.58


1.30 12.09 14.67 13.05 (0.2701) 0.172 - 1.41 - 3.82
1.50 11.76 16.06 14.94 (0.2919) 0.171 - 1.40 - 4.08
1.80 11.35 17.79 17.19 (0.3119) 0.170 - 1.42 - 4.43
2.00 11.13 18.76 18.17 (0.3198) 0.169 - 1.46 - 4.67
2.50 10.68 20.74 20.14 (0.3307) 0.168 - 1.55 - 5.12
3.00 10.33 22.29 21.67 (0.3344) 0.167 - 1.64 - 5.47
4.00 9.83 24.71 PHAST: 25 [1] 24.02 (0.3345) 0.165 - 1.79 - 5.97
5.00 9.48 26.62 25.86 (0.3362) 0.164 - 1.92 - 6.45
6.00 9.22 28.42 27.58 (0.3284) 0.162 - 2.03 - 6.66

Values in the parenthesis are the view factors calculated using Eqs. (29) and (30)

m/s) as compared to the Lsafe value in absence of wind is because of the increase
in the value of n, which is defined as follows:

kwind kstill
n ¼ Fst ð31Þ
h

Thus, n takes into account the effect of the view factor along with the effect of the
cross-wind velocity on the change in the fraction of incident heat flux radiated
from the pool fire to the flame tilt. From Tables 1 and 3 we also note that the
fraction of the radiated heat flux (k) in presence of wind is less than the fraction
of heat flux radiated during no-wind condition. However, an increase in Lsafe
value with the increase in cross-wind velocity at uwind> 1.5 m/s and uwind £ 2.0
m/s (as compared to the no-wind condition, Tables 1 and 3), is predominantly due
to the steep increase in flame tilt over a 0.5 m/s range of wind speed (refer
Table 2); while for uwind> 2.0 m/s the increase in Lsafe value has been because of
the slow increase of the flame tilt over the same wind speed range of 0.5 m/s
(shown as an observable arc-turn in Fig. 3), which results in a rapid decrease in n
value for Fst = 1 (Table 3). Therefore, cross-wind with velocities greater than
1.5 m/s enhances the impingement of flame onto the target tanks under the worst
situation. The Lsafe value, which is obtained from the present 1-D model with
Fst = 1 for gasoline storage tanks under the wind velocity, uwind = 4 m/s (refer
Table 3), shows a very good agreement with the value reported by Abbasi et al.
[1] using the PHAST software.
Comparing Tables 1 and 3, we observe that the view factor values show an
increase as the wind velocity increases from uwind = 0 till uwind = 5 m/s, beyond
which the Fst value then start decreasing with an increase in the wind velocity.
Fire Technology 2019

(a)
350

300

Slope for
250 heat loss
inside tilted,

uwind = 0.99 m/s

uwind = 1.50 m/s


asymmetric

uwind = 2.0 m/s

uwind = 3.0 m/s

uwind = 4.0 m/s

uwind = 6.0 m/s


uwind = 5.0 m/s
q r (kW/m )
2

200 flame
"

150

100 Slope for


heat loss in
surrounding
50 air

0
5 10 15 20 25 30
Inter-Tank Seperation Distance (x, m)

(b)
350

300 Slope for


heat loss
inside tilted,
250 asymmetric
flame
uwind = 0.99 m/s

uwind = 2 m/s

uwind = 3 m/s

uwind = 6 m/s
q"st (kW/m )
2

200
Drastic
heat loss
150 at flame
boundary

100
Slope for
heat loss in
surrounding
50 air

5 10 15 20 25 30
Inter-Tank Seperation Distance (x, m)
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

b Figure 4. (a) Radiation heat flux under worst situation (i.e. Fst = 1)
versus inter-tank separation distance from the centre of elongated
flame-base (assuming no temperature variation along the axis of the
flame) in presence of wind, plotted in orthogonal x–y plane. (b)
Radiation heat flux with Fst< 1 versus inter-tank separation distance
from the centre of elongated flame-base (assuming no temperature
variation along the axis of the flame) in presence of wind, plotted in
orthogonal x–y plane.

The decrease in the view factor with at the wind velocity of 6 m/s is due to the
over-tilting of the flame that abruptly increases the b/c and b/a ratios [refer
Fig. 1c and Eq. (30)] thus approaching to a situation where flattening of the flame
initiates and finally beyond 6 m/s wind speed the flame flattens as the centroid of
the solid-flame geometry falls outside the base of the geometry (refer Table 2).
Similar to the no-wind condition, introduction of Fst decreases the safe inter-tank
distance also in the presence of cross-winds (values reported in Table 3). The safe
distance calculation under cross-wind conditions and without the worst situation
consideration includes:

(a) Temperature calculation till Ddrag 2, using Eqs. (13)–(16) along with simulta-
neous calculation of the incident radiation heat flux with Fst = 1. This calcu-
lation only estimates the surface temperature (refer the Appendix) of the tilted
and asymmetric flame, which is in the downward direction of the cross-wind.
It is evident that the surface temperature of the tilted flame facing the wind
will be different from that in the downward direction of the wind [7], thus
making the flame asymmetric. 
(b) Followed by the temperature calculation starting from Ddrag 2, using
Eqs. (19), (20), (14) and (15) along with the simultaneous calculation of inci-
dent radiation heat flux with Fst< 1. Thus this calculation of incident radia-
tion heat flux at any separation distance only takes the input (at the initial
boundary condition) of the tilted and asymmetric flame surface temperature in
the downward direction of the cross-wind. This surface temperature of a tilted
flame can be obtained either from accurate experimental observations or by
solving the simplified Eqs. (13)–(16), which takes into account the asymmetric
nature of the flame under wind conditions.

A maximum of 35.4% deviation in the calculated safe separation distance is


noted at uwind = 0.99 m/s from that under worst condition, which decreases to
3% till uwind = 2 m/s (this corresponds to the first region of variation for the
flame tilt; refer to Table 2 and Fig. 3). In the second region of variation for the
flame tilt, i.e. 2.5 m/s £ uwind £ 6 m/s, we note that the percentage deviation in the
calculated safe inter-tank distances(at the conditions Fst = 1 and Fst< 1) fluctu-
ates within a narrow range of 2.96% to 2.78% for all wind speed considered. Fur-
thermore, from Tables 1 and 3 we find that for Fst< 1, the safe inter-tank
Fire Technology 2019

distances increase with an increase in the wind speed from 0 m/s to 6 m/s; unlike
the worst situation case where there is an initial decrease in the maximum safe
inter-tank distance for the wind speed 0 m/s to 1.5 m/s. This continuous increase
in the inter-tank safe distance for Fst < 1, is due to the constant decrease in the n
value with increasing wind speed (refer Table 3). It is also noted that the introduc-
tion of Fst < 1 at the flame and surrounding air interface results in a drastic
reduction  in the incident radiation heat flux at the flame boundary (i.e.
x0 ¼ Ddrag 2), as shown in Fig. 4b, for various wind velocities. This is unlike that
observed for various wind conditions with Fst = 1; were there is a gradual
decrease in the incident radiation heat flux across the flame boundary. From
Fig. 4b a monotonic decrease in the q_ 00st values with the separation distance
between the tanks has also been observed, which is similar to that seen for all the
no-wind conditions and for various cross-wind conditions with Fst = 1.

3.3. Introduction of Water Radiation Shield


Addition of the water radiation shield at a distance from the source tank (as
shown in Fig. 1a) provides a thermal resistance to radiation heat transfer, thus
reducing the transfer of the incident radiative heat flux from the source to the tar-
get tank. From the above discussion it is evident that under the worst situation
consideration, both in presence and absence of wind; maximum separation dis-
tance between the tanks are required for safe layout of a tank farm exposed to a
fire domino effect. Also the use of a view factor causes simplification of the sce-
narios and thus resulting in a 20% safety factor consideration on the minimum
safe separation distance calculations [25]. Hence reduction of the maximum safe
separation distance, which corresponds to the calculations with Fst = 1, by add-
ing a radiation shield between the tanks seems more practical approach from
safety and economic point of view. Therefore, in the present work we have exten-
ded the developed 1-D model to reduce the maximum safe separation distance
(corresponding to Fst = 1) to an optimal safe inter-tank distance, by adding a
water radiative shield between the tanks.
D A water-shield
 of 1 cm thickness (dopt) has been introduced at a distance of
2 þ 3:33 m from the centre of the source tank in case of no-wind pool fire sce-
nario, with D = 10 m. This resulted in a decrease of the safe inter-tank separa-
tion distance (calculated using Eqs. 26–28) and measured from the centre of the
source tank, Lsafe,centre) by 60.9% as compared to no-wind condition without
water radiation shield (refer Tables 1 and 4). This Lsafe,centre value close to 8.34 m
under the no-wind condition and with the water-shield is maintained to prevent
the domino effect caused by the pool fire. In presence of various cross-wind veloc-
ities, the optimal thickness (dopt) of the radiation shield (reported in Table 4) is
adjusted so that the critical radiation heat flux (q_ 00st ¼ 4:732 kW/m2 ) is attained
within the thickness of the water-shield. Thus, a water-shield thickness of 1.21 cm
has been calculated to be the optimal thickness. In the presence of cross-wind, this
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

calculated dopt values has been noted to vary due to the simultaneous increase and
decrease in the flame tilt and in the k value (refer Tables 2 and 3), respectively.
Interestingly, we also observe from Table 4 that the variation of dopt value and
hence the dopt/Lsafe,centre value follows a relationship with the wind speed. First,
with the onset of the flame tilt at uwind = 0.99 m/s (h = 9.65 and
n = - 3.03 9 10-4) the dopt value shows an increase, which remains constant till
uwind = 1.5 m/s. In the second variation zone of dopt values, where the optimal
safe inter-tank separation distance from the centre of the source tank is approxi-
mately 8.34 m and the uwind values increase from 1.8 m/s to 4 m/s, the dopt value
decreases and dopt/Lsafe,centre ratio remains constant to a value of 1:39  10 3 .
Thereafter, the dopt values further decrease over the wind speed of 5 m/s and 6 m/
s due to the decrease in n value (refer Table 3), which results in greater reduction
in the fraction of heat radiated at these wind conditions, thus decreasing the dopt/
Lsafe,centre to 1.36 9 10-3. The change in the dopt values, as reported in Table 4, is
because of less water requirement in the water-shield due to low fire impingement
in absence of wind; followed by an increase in the water radiation shield thickness
to reduce fire impingement caused by the onset of wind effect, which gradually
decreases as fraction of heat radiation and hence the q_ 00r value diminishes with the
increasing wind speed. However, in all the cases considered we note that dopt/
Lsafe,centre 1 and that the dopt values under cross-wind condition are always
greater than the dopt value in the absence of wind.
Figure 5 shows the variation of incident radiation heat flux through the water-
shield, calculated as q_ 00st = Fst rew Tw4 (with Fst = 1, the worst situation possible)
from Eqs. (26)–(28), in both the absence and presence of cross-wind. It has been
observed that there is a steep decrease of the incident radiation heat flux inside
the water-shield as compared to the decrease in the radiant heat flux through the
surrounding air, both in absence and presence of cross-wind (Figs. 2 and 4a).

Table 4
Safe Inter-Tank Distances Under No-Wind and Cross-Wind Conditions
for Pool Fires in the Fuel Storage Tanks with 10 m Diameter and
Containing Gasoline

uwind Safe distance from the centre of source tank, Lsafe,centre (m) dopt 9 103 (dopt/Lsafe,cen-
3
(m/s) with Fst = 1 (m) tre) 9 10

0 8.3382 10.00 1.20


0.99 8.3360 12.10 1.45
1.30 8.3412 12.10 1.45
1.50 8.3416 12.10 1.45
1.80 8.3408 11.60 1.39
2.00 8.3392 11.60 1.39
2.50 8.3395 11.60 1.39
3.00 8.3382 11.60 1.39
4.00 8.3351 11.60 1.39
5.00 8.3391 11.30 1.36
6.00 8.3390 11.30 1.36
Fire Technology 2019

uwind = 0 (no-wind condition)


uwind = 0.99 m/s
50
uwind = 1.5 m/s
uwind = 2.0 m/s
uwind = 3.0 m/s
40 uwind = 4.0 m/s
uwind = 5.0 m/s
uwind = 6.0 m/s
q r (kW/m )
2

30
"

20

10

8.330 8.332 8.334 8.336 8.338 8.340 8.342


Distance from leading surface within water-shield (x, m)

Figure 5. Radiation heat flux with Fst = 1, along the thickness of the
water-shield placed at a distance from the centre of the source tank in
the absence and presence of wind, plotted in orthogonal x–y plane.

4. Conclusion
The present 1-D model, which has been obtained from the steady state energy bal-
ance over the parallelepiped geometry with plane surfaces, is used to predict the
safe inter-tank distances between class-I fuel storage tanks of 10 m diameter; both
in presence and absence of wind. Major contributions of this work include:

(a) Development of a deterministic 1-D model, which provides an excellent pre-


diction of the safe inter-tank distances in a tank farm exposed to fire both
under no-wind and cross-wind conditions. Under the worst situation consider-
ation the present 1-D model predictions are in excellent agreement with the
Shokrie-Beyler’s prediction (which gives better predicted Lsafe values in no-
wind condition [1, 4] and for the value q_ 00r  5 kW/m2 [40]) and with the
PHAST software predictions (which provides better approximate of the Lsafe
values under worst conditions [1]).
(b) Under cross-wind condition the increasing flame tilt for a large pool fire(D =
10 m) cause flattening of the flame (with the centroid of parallelepiped geom-
etry falling outside the base of tilted flame) at wind speeds greater than 6 m/s,
which therefore reduces the fire impingement onto target tanks and hence pro-
vides an upper limit of 28.42 m to the Lsafe value. However, within the wind
speed ranging between 0.99 m/s and 6 m/s the Lsafe values between the storage
tanks is consistently governed by the n value (defined by Eq. 31),which simul-
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

taneously measure an increase in the flame tilt and also the increase in the
change of k value. Also under the wind effect elongation of the flame base is
noted along with an observable flame tilt that increases with an increase in the
wind speed.
(c) Insertion of a water radiation shield under worst situation consideration (i.e.
Fst = 1), with an optimal thickness (dopt) is capable of drastically reducing the
maximum safe inter-tank distance(Lsafe,centre) close to 8.34 m (measured from
the centre of the source tank) both in presence and absence of wind; such that
dopt/Lsafe,centre 1. It is noted that the dopt values in the presence of wind are
larger than the dopt value under no-wind condition; indicating that an optimal
of 2589 kg-water/m/s of water loading (C-value corresponding to
dopt ¼ 12:10  10 3 ) in the water-shield is sufficient in mitigating the domino
effect caused by the large pool fire of 10 m diameter.

Thus, maintaining the optimal Lsafe,centre value between the fuel storage tanks as
has been calculated from the present 1-D model would reduce the establishment,
safety and risk assessment costs incurred due to the large inter-tank safe separa-
tion distances in absence of a water-shield.

Appendix
Within radiative regime, i.e. transfer of heat transport fluxes from the flame-front
surfaces at x and x + Dx (refer Fig. 1b) are dominantly by radiation; while the
heat losses from the other surfaces of the elemental control volume can still be via
convection and radiation (Sects. 2.1 and 2.2), the flame can be optically thick or
thin depending on the pool diameter [33]. For pool diameter > 1 m with radiative
flame give rise to optically thick flame [33]. Koseki [29] via experiments on large
pool fires had concluded that large unconfined pool fires are predominantly in
radiative regime. In the present scenario, we have considered a large 10 m diame-
ter pool fire which is therefore, assumed to be in the radiative regime. The mean
optical thickness of a flame is defined in terms of jb, where j is the absorption
extinction coefficient and b is the mean beam length corrector [26, 27, 33]. While j
value for most of the fuel is documented in literature for pool fires with
D ‡ 0.2 m, it had been argued that a ‘‘common b value does not emerge’’ for a
given fuel and pool diameter and therefore, the jb value is not unique [33].
Hence, while Babrauskas [33] reported the jb value for gasoline pool fires of
D ‡ 0.2 m to be 2.1 m-1; Sudheer et al. [26] reported jb = 1.5 m-1. So we have
calculated the jb value for the present pool fire of D = 10 m containing gasoline;
using the following relation, where m_ 00 is the mass burning rate of the fuel (which
can be obtained from pool diameter relationship [4, 27, 29, 33]) and m_ 001 is a con-
stant for a given fuel [33].

m_ 00 ¼ m_ 001 ð1 expð jbD)) ð32Þ


Fire Technology 2019

For gasoline (m_ 001 ¼ 0:055 kg/m2 =s, j = 2 m-1 [33]) we have reported the jb and
the b values in Table 5 under both no-wind and windy conditions. The mass
burning rate for each of the cases are also reported in the same table and are
obtained from pool diameter relationship given in literature [27] under the no-
wind condition; or by using Eq. (8) for windy conditions. Furthermore, the calcu-
lation of the optical depth (s) have been performed using the Equation (33) [27,
44] and these values are also reported in Table 5:

3:6Vf
s ¼ jb ð33Þ
Af

where Vf is the volume of the flame geometry and Af is the surface area of the
flame geometry. It is noted that the jb value and the optical depth under no-wind
condition for the present case is 0.1595 m-1 and 1.595, respectively. Following the
argument on radiative regime pool fires with D > 1 m [33], s = 1.595 is taken to
be the optical thin limit for the gasoline pool fire and we also observe that for all
windy conditions considered the corresponding optical depth values are greater
than this limit. Therefore, the flame surface temperatures of the pool fires have
been used to estimate the safe inter-tank distance in a gasoline tank farm for both
no-wind and windy conditions. We also note an increase in the s value with the
increasing wind speed.

Table 5
Effect of Wind Velocity on Optical Thickness of Pool Fires in the Fuel
Storage Tanks with 10 m Diameter and Containing Gasoline
 
3:6V
uwind (m/s) m_ 00 (kg/m2/s) jb (m-1) b s ¼ jb Af f

0 0.0438 [4, 27] 0.1595 0.0797 1.5950


0.99 0.0445 0.1655 0.0828 1.8074
1.30 0.0447 0.1675 0.0837 1.8974
1.50 0.0448 0.1688 0.0844 1.9508
1.80 0.0450 0.1707 0.0854 2.0229
2.00 0.0452 0.1720 0.0860 2.0696
2.50 0.0455 0.1754 0.0877 2.1772
3.00 0.0458 0.1790 0.0895 2.2763
4.00 0.0465 0.1864 0.0932 2.4677
5.00 0.0471 0.1944 0.0972 2.6536
6.00 0.0478 0.2031 0.1016 2.8438

Vf is the volume of the flame geometry and Af is the surface area of the flame geometry
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

References
1. Abbasi MH, Benhelal E, Ahmad A (2014) Designing an optimal safe layout for a fuel
storage tanks farm : case study of Jaipur Oil Depot. Int J Chem Mol Nucl Mater
Metall Eng 8:147–155
2. Bariha N, Srivastava VC, Mishra IM (2016) Theoretical and experimental studies on
hazard analysis of LPG/LNG release: a review Center for Chemical Process Safety.
Rev Chem Eng 33:387–432. https://doi.org/10.1515/revce-2016-0006
3. Necci A, Argenti F, Landucci G, Cozzani V (2014) Accident scenarios triggered by
lightning strike on atmospheric storage tanks. Reliab Eng Syst Saf 127:30–46. https://
doi.org/10.1016/j.ress.2014.02.005
4. Sengupta A, Gupta AK, Mishra IM (2011) Engineering layout of fuel tanks in a tank
farm. J Loss Prev Process Ind 24:568–574
5. Benedetti RP (2016) Flammable and combustible liquids code. National Fire Protection
Association, Quincy
6. Gorjipour S, Anvaripour B (2014) Safe distance from pools fires in NGL storage tanks
in stagnant Air. Int J Basic Sci Appl Res 3:777–780
7. Beer T (1991) The interaction of wind and fire. Bound-Layer Meteorol 54:287–308
8. Lautkaski R (1992) Validation of flame drag correlations with data from large pool
fires. J Loss Prev Process Ind 5:175–180. https://doi.org/10.1016/0950-4230(92)80021-Y
9. Jiang P, Lu SX (2016) Pool fire mass burning rate and flame tilt angle under crosswind
in open space. Procedia Eng 135:260–273. https://doi.org/10.1016/j.proeng.2016.01.122
10. Lam CS, Weckman EJ (2015) Wind-blown pool fire, part I: experimental characteriza-
tion of the thermal field. Fire Saf J 75:1–13. https://doi.org/10.1016/j.firesaf.2015.04.009
11. Lam CS, Weckman EJ (2015) Wind-blown pool fire, part II: comparison of measured
flame geometry with semi-empirical correlations. Fire Saf J 78:130–141. https://doi.org/
10.1016/j.firesaf.2015.08.004
12. Hu L, Liu S, Xu Y, Li D (2011) A wind tunnel experimental study on burning rate
enhancement behavior of gasoline pool fires by cross air flow. Combust Flame
158:586–591. https://doi.org/10.1016/j.combustflame.2010.10.013
13. Hu L, Wu L, Liu S (2013) Flame length elongation behavior of medium hydrocarbon
pool fires in cross air flow. Fuel 111:613–620. https://doi.org/10.1016/j.fuel.2013.03.025
14. Hu L, Liu S, Wu L (2013) Flame radiation feedback to fuel surface in medium ethanol
and heptane pool fires with cross air flow. Combust Flame 160:295–306. https://
doi.org/10.1016/j.combustflame.2011.11.008
15. Hu L, Liu S, De Ris JL, Wu L (2013) A new mathematical quantification of wind-
blown flame tilt angle of hydrocarbon pool fires with a new global correlation model.
Fuel 106:730–736. https://doi.org/10.1016/j.fuel.2012.10.075
16. Hu L, Hu J, Liu S, Tang W, Zhang X (2015) Evolution of heat feedback in medium
pool fires with cross air flow and scaling of mass burning flux by a stagnant layer the-
ory solution. Proc Combust Inst 35:2511–2518. https://doi.org/10.1016/
j.proci.2014.06.074
17. Hu L, Kuang C, Zhong X, Ren F, Zhang X, Ding H (2017) An experimental study on
burning rate and flame tilt of optical-thin heptane pool fires in cross flows. Proc Com-
bust Inst 36:3089–3096. https://doi.org/10.1016/j.proci.2016.08.021
18. Hu L (2017) A review of physics and correlations of pool fire behaviour in wind and
future challenges. Fire Saf J 91:41–55. https://doi.org/10.1016/j.firesaf.2017.05.008
19. Fan CG, Tang F (2017) Flame interaction and burning characteristics of abreast liquid
fuel fires with cross wind. Exp Therm Fluid Sci 82:160–165. https://doi.org/10.1016/
j.expthermflusci.2016.11.010
Fire Technology 2019

20. Khakzad N, Khan F, Amyotte P, Cozzani V (2014) Risk management of domino


effects considering dynamic consequence analysis. Risk Anal 34:1128–1138. https://
doi.org/10.1111/risa.12158
21. Khakzad N, Reniers G (2015) Using graph theory to analyze the vulnerability of pro-
cess plants in the context of cascading effects. Reliab Eng Syst Saf 143:63–73. https://
doi.org/10.1016/j.ress.2015.04.015
22. Santos S, Landesmann A (2014) Thermal performance-based analysis of minimum safe
distances between fuel storage tanks exposed to fi re. Fire Saf J 69:57–68. https://
doi.org/10.1016/j.firesaf.2014.08.010
23. Rebec A, Koľ J, Plě P (2016) Fires in storages of LFO : Analysis of hazard of struc-
tural collapse of steel-aluminium containers. J Hazard Mater 306:367–375. https://
doi.org/10.1016/j.jhazmat.2015.12.006
24. Sullivan AL, Ellis PF, Knight IK (2003) A review of radiant heat flux models used in
bushfire applications. Int J Wildl Fire 12:101–110. https://doi.org/10.1071/WF02052
25. Zárate L, Arnaldos J, Casal J (2008) Establishing safety distances for wildland fires.
Fire Saf J 43:565–575. https://doi.org/10.1016/j.firesaf.2008.01.001
26. Sudheer S, Prabhu SSV (2013) Characterization of open pool fires and study of heat
transfer in bodies engulfed in pool fires. https://www.iitb.ac.in/sudheer/files/thesis_ssr_
2013.pdf. Accessed 3 Jan 2018
27. Drysdale D (1999) An introduction to fire dynamics, 2nd edn. Wiley, Hoboken
28. Bariha N, Mani I, Chandra V (2016) Fire and explosion hazard analysis during surface
transport of liquefied petroleum gas (LPG): a case study of LPG truck tanker accident
in Kannur, Kerala, India. J Loss Prev Process Ind 40:449–460. https://doi.org/10.1016/
j.jlp.2016.01.020
29. Koseki H (1999) Large scale pool fires: results of recent experiments. In: Fire safety
science sixth international symposium, pp 115–132
30. Heskestad G (1984) Engineering relations for fire plumes. Fire Saf J 7:25–32. https://
doi.org/10.1016/0379-7112(84)90005-5
31. Cengel YA, Klein S, Beckman W (2002) Heat transfer: a practical approach. McGraw
Hill Education, New York
32. Churchill SW, Chu HHS (1975) Correlating equations for laminar and turbulent free
convection from a vertical plate. Int J Heat Mass Transf 18:1323–1329. https://doi.org/
10.1016/0017-9310(75)90243-4
33. Babrauskas V (1983) Estimating large pool fire burning rates. Fire Technol 19:251–261.
https://doi.org/10.1007/BF02380810
34. Kodur VKR, Harmathy TZ (2016) Properties of building materials. In: Hurley MJ
et al (eds) SFPE handbook of fire protection engineering. Springer, New York. https://
doi.org/10.1007/978-1-4939-2565-0_9
35. Thomas PH (1963) The size of flame from natural fires. Proc Combust Inst 9:844–859
36. Wade DD (2013) Flame descriptors. SFE fact sheet 2013–6, Southern fire exchange, 2 p
37. Bejan A (2004) Convection heat transfer, 3rd edn. Wiley, Hoboken
38. Chun KR, Seban RA (1971) Heat transfer to evaporating liquid films. J Heat Transfer
93:391. https://doi.org/10.1115/1.3449836
39. Geankoplis CJ (1993) Transport processes and unit operations, 3rd edn. Prentice-Hall
International Inc, Upper Saddle River
40. Hurley MJ, Gottuk D, Hall JR, Harada K, Kuligowski E, Puchovsky M, Torero J,
Watts JM, Wieczorek C (2016) SFPE handbook of fire protection engineering, 5th edn.
Springer, New York, pp 1–3493. https://doi.org/10.1007/978-1-4939-2565-0
Optimal Safe Layout of Fuel Storage Tanks Exposed to Pool Fire

41. Feingold A (1966) Radiant-interchange configuration factors between various selected


plane surfaces. Proc R Soc A Math Phys Eng Sci 292:51–60. https://doi.org/10.1098/
rspa.1966.0118
42. Holman JP (2010) Heat Transfer, 10th edn. McGraw Hill, New York. https://doi.org/1
0.1115/1.3246887
43. Koseki H, Yumoto T (1988) Air entrainment from heptane pool fires.pdf. Fire Technol-
ogy
44. De Ris J (1979) Fire radiation—a review. In: Symposium (international) on combus-
tion, pp 1003–1015. https://doi.org/10.1016/s0082-0784(79)80097-1

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

You might also like