You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/289080866

Fire modelling and cooling of storage tanks in case of radiation from large
pool fires

Article  in  Proceedings of the IASTED International Conference on Modelling, Simulation and Optimatization · January 2005

CITATIONS READS

0 1,212

2 authors:

Francesco Devia M. Fossa


Università degli Studi di Genova Università degli Studi di Genova
37 PUBLICATIONS   360 CITATIONS    92 PUBLICATIONS   960 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Developing a dynamic model of Phase Change Material View project

shallow vertical GSHP View project

All content following this page was uploaded by M. Fossa on 28 October 2016.

The user has requested enhancement of the downloaded file.


8:07f=WðJul162004Þ Prod:Type:FTP ED:AjeeshVV
þ model
FISJ : 1614 pp:128ðcol:fig::16;811Þ PAGN:Anvar SCAN:

ARTICLE IN PRESS

3
Fire Safety Journal ] (]]]]) ]]]–]]]
www.elsevier.com/locate/firesaf
5

7
A model for radiation evaluation and cooling system design in case of
9
fire in tank farms
11
Marco Fossa, Francesco Devia
13
Dipartimento di Ingegneria della Produzione Termoenergetica e Modelli Matematici (DIPTEM), University of Genova, via all’Opera Pia 15/a, I-16145
Genova, Italy
15
Received 17 May 2005; received in revised form 12 January 2007; accepted 17 January 2007

17

19
Abstract

21 The effects on surrounding tanks of the heat radiated by a pool fire occurring over a storage tank of liquid hydrocarbons are studied by
means of a solid flame model. The irradiation to the surroundings is calculated with a set of empirical correlations for flame parameters
23 and by means of a complete 3D view factor model, implemented in a self-built code, which solves the view factor problem regarding the

F
infinitesimal surfaces that describe the source and target surfaces. The calculation of irradiation on target surfaces is then applied to
25 O
evaluate the local and overall amount of cooling water needed to prevent the fire from propagating to the surroundings. The cooling
water, which protects the tanks facing the flame, is expected to be distributed as a film moving downward from a pipe around the top of
O
27 the vessel, while tank top is equipped with foam systems. Once validated in terms of available literature data, the present model is applied
to a series of test cases regarding different separation distances between tanks in order to evaluate the mass flow rates of water needed for
PR

29 cooling the target tank side for a variety of operating conditions.


r 2007 Elsevier Ltd. All rights reserved.
31 Keywords: Pool fire modelling; Radiation heat transfer; View factors; Cooling design
D

33
1. Introduction 10 m. In such a situation, it is top priority to confine the fire
TE

35 and to protect adjacent tanks from ignition (to avoid the 59


Among fires occurring in industrial sites, pool fires from domino effect) by means of wall cooling with a proper
37 storage tanks of hydrocarbons probably represent the most amount of water. In order to evaluate the hazards
EC

61
dangerous situation for surrounding structures and per- associated with such a scenario and plan all the actions
39 sons. Such pool fires, in fact, are highly localised and necessary to reduce fire effects, a reliable estimation of the 63
intense emitters of heat and smoke. Radiation and heat radiated by the flame is mandatory. This task can be
R

41 convection are the main mechanisms for transferring heat accomplished by different approaches. The usual strategy 65
from a fire to the surroundings. Radiation is usually the is based on the assumption that the flame is a stable surface
R

43 dominant mode of heat transfer, although convective heat whose parameters (temperature, emissivity and so on) do 67
O

transfer becomes an important mode for structures, which not vary in space and time. This approach needs
45 are directly engulfed by the fire. correlations for flame parameters and a proper model for 69
C

In the region close to the fire, at distances comparable to radiative heat propagation (including estimation of config-
47 the flame length, radiation can be intense enough to uration factors for all the area elements involved). In this 71
N

threaten the structural integrity of constructions, as well as paper a review of existing correlations for pool fire
49 the safety of personnel and firefighters. The occurrence of a parameters is presented together with an analysis of
U

73
fire in a storage tank farm is a situation where structures available closed form formulae for view factor estimation.
51 are irradiated by flames at short distances, of the order of The paper presents the results of a numerical solution of 75
the view factor problem with reference to the heat radiated
53 Corresponding author. Tel.: +39 0103532198; fax +39 010311870. by a tilted cylindrical flame (of assigned temperature and 77
E-mail address: marco.fossa@unige.it (M. Fossa). geometry) to cylindrical tanks.
55 79
0379-7112/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
57 doi:10.1016/j.firesaf.2007.01.005

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
2 M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]]

1
Nomenclature Re Reynolds number, Re ¼ wD/u 59
3 w wind velocity, m/s
A area, m2 x x coordinate or distance, m 61
5 c specific heat, kJ/kg K y y coordinate, m
D diameter, m z z coordinate, m 63
7 Ef flame average emissive power, kW/m2
FdA1–A2 view factor with respect to flame surface A2 Greek letters 65
9 Fr Froude number, Fr ¼ w2/gD
Frad irradiated energy fraction f flame tilt angle, rad 67
11 G irradiance, kW/m2 e luminous fraction
g acceleration of gravity, m/s2 t air transmissivity 69
13 H tank height, m y angle between normal and vector connecting
Hi lower heating value, kJ/kg elementary surfaces, rad 71
15 km constant for emissive power, m1 u kinematic viscosity, m2/s
kb extinction coefficient, m1 DTnom water nominal temperature difference, K 73
17 L flame length, m
m_ 00 mass burning rate per unit area, kg/sm2 Subscripts 75
19 m_ dimensionless mass burning rate
m_ 00L water mass flow rate per unit area, kg/sm2 max maximum value 77
21 R radius, m soot relative to soot
79
23

F
The model presented here has the purpose of evaluating and compare literature correlations regarding either pool 81
25 fire effects on the surrounding tanks. In particular, it is
assumed that the tanks are protected by water cooling and
O
or jet fires. All the above models consider the flame as a
uniform radiation emitter, while the recent model by Rew 83
O
27 that the water is spread over the tank vertical surface from et al. describes the flame as a two-layer emitter, with a
a pipe ring located close to the top edge of the side surface hotter base and a less radiant upper part. Measurements on 85
PR

29 of the tank. Here it is considered that the tank top is emissive power distribution along the flame surface have
equipped with foam systems, but no calculations are been recently performed by Munoz et al. [5] and Koseki et 87
31 performed to evaluate the thermal behaviour of such al. [6], but no general correlations have been proposed to
medium when exposed to radiation. This paper is aimed at model the non-uniform emission from a flame. 89
D

33 demonstrating that if the liquid is not distributed properly The solid flame, describing a pool fire, is generally
in the circumferential direction, areas of water dryout can assumed to be a tilted cylinder. The geometry of the flame 91
TE

35 occur on the tank surface and conditions for hydrocarbon is hence defined by the diameter of the pool D, by the
vapour formation inside the tank may be created. cylinder length L (measured along cylinder axis) and by a 93
37 shape factor, to take into account the effect of wind, which
EC

2. Pool fire modelling tends to elongate the cylinder base from a circular to an 95
39 elliptical contour. The relevant properties of the flame are
The basic strategy for calculating the amount of the burning rate of the liquid m _ 00 and the surface emissive 97
R

41 radiation received from a flame by a target surface is the power Ef, both depending on the fuel properties and pool
point source. In this case the assumption is that the flame dimensions. Finally a suitable model is required for 99
R

43 radiates either from a single point or from a number of radiation propagation in terms of air transmissivity and
points along the central axis of the flame. The single point calculation of the view factors. In what follows there is a 101
O

45 emitter model is rather suitable for large distances from the summary of the correlations adopted in this work.
C

heat source, but the method does not provide accurate In the scenario here considered (fire over a storage tank) 103
47 results for short separation distances. Another possibility is the pool diameter may be assumed coincident with the tank
N

to model the flame as a solid surface, whose shape and diameter. 105
49 thermal characteristics do not vary in time. This method is The estimation of the burning rate for hydrocarbon
U

based on the choice of a proper set of empirical formulae liquids is given by the following relation, which is a 107
51 for flame parameters and this is the strategy usually simplified formula derived from the studies by Blinov and
adopted in scientific literature devoted to fire hazard Khudianov [7] and by Hottel [8]: 109
53 predictions or flame experimental data reduction. Many  
m_ 00 ¼ m
_ 00max 1  ekbD , (1)
authors have proposed empirical correlations to describe 111
55 the behaviour of buoyant flames, even in the presence of where m_ 00max is the burning rate of an infinite dimension
wind. The works by Mudan [1], Shokri and Beyler [2], by pool, kb is the extinction coefficient [9,10]. The above 113
57 Rew et al. [3] and by Guedes Soares [4] extensively present formula applies for flame diameters larger than, tipically,

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]] 3

1 1 m. Values for the maximum burning rate and extinction produced by combustion. On the other hand, available
coefficient for liquid fuels are given by the formulae and experimental data on irradiated fraction refer to a limited 59
3 tables provided in [3,10,11]. set of pool diameters and fuel types.
The flame length L is here calculated by the correlation A refined version of formula (6) is that proposed by 61
5 by Thomas [12], as a function of the dimensionless mass- Mudan [15], which was based on extensive measurements
burning rate m_ : of radiation regarding kerosene, JP5 and gasoline: 63
7  0:61
 
L=D ¼ 42ðm _ Þ , (2) E f ¼ 140e0:12D þ 20 1  e0:12D , (8)
65
9  pffiffiffiffiffiffiffi where constants are expressed in kW/m2.
m ¼ m_ 00 =ra gD , (3) Formula (8) is not suitable to evaluate the heat radiated 67
11 where ra is the air density and g the acceleration of gravity. by fire of low smoking fuels like LNG and LPG. In this
Eq. (2) yields similar results to those provided by the paper experimental data regarding fires on land of such 69
13 more recent correlation by Binding and Pritchard [13], that liquified gases [15] has been employed to calculate the
takes into account a weak effect of the wind velocity on the constants Ef, max and km for emissive power functions 71
15 flame length, by means of an exponent equal to 0.03. having the form:
  73
To characterise the solid flame in the presence of a wind, E f ¼ E f;max 1  ekm D . (9)
17 information on the flame tilt angle f and on the flame
elongation ratio is required. Binding and Pritchard [13] The emissive power exponential function has been 75
19 correlated literature data in terms of proper Reynolds and selected according to the peculiar data trend with respect
Froude numbers based on the properties of air, on wind to pool diameter, as reported in literature. For LNG and 77
21 velocity and on pool diameter: LPG the above constants turned out to be (225 kW/m2,
0.372 m1) and (157 kW/m2, 0.351 m1), respectively. The 79
23 tan f= cos f ¼ 0:666 Fr0:33 Re0:117 . (4) irradiation G on a target surface exposed to a fire is given

F
Worthwhile noticing, the above formula quite well by the product of the flame emissive Ef, of the air 81
25 agrees with the correlation proposed by the American transmissivity t and of the view factor Fd1–2.
O
Gas Association [14]. Air transmissivity with respect to air temperature and 83
O
27 The wind causes the base of the flame to be dragged and relative humidity is here calculated by means of the
deformed into an elliptical shape [15]. The correlation by correlation proposed by Wayne [19]. 85
PR

29 Lautkaski [16] provides the elongated diameters D0 as G ¼ E f tF dA1 A2 , (10)


87
0
D =D ¼ 1:5 Fr 0:069
. (5) Z
31 cos y1 cos y2
F dA1 A2 ¼ dA2 . (11) 89
A crucial point is the evaluation of the heat radiated by pS 2
D

33 the flame. It depends on many factors, including the fuel


The view factor formula (Eq. (10)) refers to a generic 91
type and the pool diameter. The flame can be modelled as a
TE

35 target infinitesimal surface dA1, which faces the A2 finite


grey medium whose emissivity depends mainly on the soot
surface of the tilted flame; y represents the angle between 93
component and, in minor part, on gaseous elements
37 the normal to the surface element and the vector of length
(carbon dioxide and water vapour). Since it is difficult to
EC

S connecting the emitting and target surfaces. Available 95


estimate the contributions to emissivity due to soot and to
39 literature correlations for estimating the view factor (cf.
gases separately, empirical correlations have been pro-
next paragraph) unfortunately refer to a limited set of 97
posed. These correlations are based on the assumption that
target and cylinder mutual disposition. Therefore, the view
R

41 the real flame area is constituted by a luminous fraction e


factor problem was solved by means of a direct algorithm 99
and by a less radiating part of flame, which is partially
R

43 implemented in a code developed at Diptem, named Fast


obscured by soot smoke [1].
View Factor Solver (FVFS). The code subsequently solves 101
O

E f ¼ E max  þ E soot ð1  Þ. (6) the problem of the view factor from differential elements
45
(Eq. (11)) and performs an optimised integration to
C

In this paper the emissive power of the black soot is 103


47 assumed to be equal to 20 kW/m2 [1]. The emissive power account for the presence of the finite surface of the flame
N

of the luminous part of flame Emax can be calculated and for the mutual disposition of the flame with respect to 105
49 according to the method suggested in [17,18], in terms of any user defined finite target area in a 3D domain.
U

the irradiated energy fraction Frad: The algorithm calculates the view factor of the dA1 area 107
51  1 (target) with respect to A2 area (the flame), which is
_ 00 H i 1 þ 4L=D .
E max ¼ F rad m (7) constituted by N2 elementary dA2 areas: 109
53 Values for Frad range from 0.1 to 0.4 depending on the 1XN2
cos y1i cos y2i
fuel type and are available in [16]. In this study the values F d12 ¼ A2i . (12) 111
p 1 S 2i
55 adopted for e and Esoot are 0.2 and 20 kW/m2, respectively,
as suggested by Mudan [1]. Eq. (7) has physical consis- Since the flame is modelled as a cylinder tilted downwind 113
57 tency, since it relates the radiated heat to the overall heat by an angle F with respect to z-axis Rx and Ry the radii of

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
4 M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]]

1 the elliptical base of tilted cylinder, (x0, y0, z0) the location
of base centre of cylinder, and the flame surface can be 59
3 analitically expressed as
 2   2 61
5 ðx  x0 Þ  ðz  z0 ÞtgF y  y0  ðz  z0 ÞtgF
þ ¼ 1.
Rx Ry 63
7 (13)
65
9 From the above formula, the coordinates and the
orientation (in terms of cosine directors nx, ny, nz) of each 67
11 elementary dA2 area can be calculated. Finally, for each
small surface dA1, located externally to the flame, whose 69
13 location and orientation are expressed by [x1; y1; z1] and
[nx1; ny1 nz1], it is possible to calculate the distance Si 71
15 between dA1 and the element dA2i, and the angles y1i and
y2i. 73
17
75
19 3. Model validation and case studies
77
21 In order to validate the algorithm developed at Diptem
and aimed at view factor evaluation, extensive comparisons 79
23 have been carried out with respect to analytical solutions

F
proposed in the literature. The FVFS code has been 81
25 demonstrated to provide reliable results in agreement with
all reference formulae [20–23]. The cross comparison
O 83
O
27 among literature solutions revealed some remarkable
Fig. 1. View factors calculation comparison. Filled symbols: dimension-
discrepancies in the predictions provided by different 85
less flame length L/R ¼ 2. Empty symbols L/R ¼ 6. Continuous lines:
PR

29 correlations applied to assigned geometrical configura- valued by Rein et al.


tions; this matter will be discussed in depth in a future 87
31 paper. As an example of the comparisons carried out, Fig.
1 shows the view factor values calculated by means of 89
D

33 FVFS and those provided by Rein et al. [21]. The related R


geometry is that sketched in Fig. 2 and refers to a surface in 91
TE

35 the downwind direction with respect to the tilted flame.


The validation against Rein data was carried out with 93
L
37 reference to different cylinder tilt angles in the range
EC

15–601. A2 95
39 The reliability of the overall model (radiation propaga- φ
tion together with correlation data set for flame parameter dA1 97
R

41 evaluation) has been confirmed by extensive comparison


with literature experimental data on irradiation at target 99
R

x
43 from pool fires. The comparisons refer to gasoline, Jp4,
crude oil, kerosine fuels and pool diameters from 6 to 60 m, 101
O

Fig. 2. Tilted cylinder facing a vertical differential element.


45 as reported in [25–27]. Additional data have been taken
C

from Munoz et al. [5] and refer to gasoline and diesel fires 103
47 from pools having diameters between 3 and 6 m. Further Figs. 3 and 4 refer to measurements in the absence of
N

experimental data refers to LNG on land, in the presence wind: good agreement between predictions and measure- 105
49 of wind, for a 6 m pool diameter [28]. The comparisons ments can be observed, even if with a general over-
U

between measured and calculated values have been estimation, especially for smoky hydrocarbons and large 107
51 performed in terms of either Eqs. (7) or (8) emissive power distances ((x/D41). At short distances (x/Do1), the
values. Mudan Eq. (9) provided better agreement between present model seems to underestimate the heat radiated 109
53 predictions and measured values, and hence is here by the flame, probably due to the strong contribution (here
adopted as reference formula for emissive power estimation not considered) that the luminous flame base exerts 111
55 for non-boiling fuels. towards close targets, as discussed in detail by Rew et al.
Figs. 3–5 show the predicted and measured values of [3]. Fig. 5 refers to less sooty flames (LNG on land) in the 113
57 irradiation on vertical targets at flame base elevation. presence of wind (w ¼ 4 m/s). In this comparison, the

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]] 5

1 4 12
x/D=5 LNG 59
3 D=3~60m D=6m Measured
10 Calculated 61
Estimated irradiation [kW/m2]

Irradiation on target [kW/m2]


5 3

8 63
7
Gasoline 65
9 2 6
kerosine
Crude oil 67
11 [24]
Heptane 4
69
1 Hexane
13
gasoline [5] 2 71
15 diesel [5]
0
73
0
17 0 1 2 3 4 4 5 6 7 8
Measured irradiation [kW/m2]
75
19 x/D
Fig. 3. Calculated and measured irradiance values on vertical target at
Fig. 5. Calculated and measured irradiance values on vertical target at 77
21 flame base. Commercial hydrocarbons and crude oil, data by Koseki [26]
flame base. LNG on land, data by Johnson [28].
and Muñoz [5].
79
23

F
30 x 81
25 JP4
D=10m Measured
O Wind
y 83
O
27 25 Calculated direction
x
85
Irradiaton on target [kW/m2]

PR

29 Tank 1 Tank 2
20
87
31
15
89
D

33 Φ L
Flame 91
TE

35 10 (tilted cylinder)
93
37 2R
EC

5 H z 95
39 x
97
R

0 Tank 1 Tank 2
41
0.50 0.75 1.00 1.25 1.50
Fig. 6. Tank configuration and related lengths considered in the case 99
R

43 x/D study.
101
O

Fig. 4. Calculated and measured irradiance values on vertical target at


45 flame base. JP4 fuel, data by Hagglund and Persson [26].
C

values 2.5ox/Ro10. All the characteristic dimensions and 103


47 distances of the problem are explained in Fig. 6. The tank
N

model seems again to overestimate the irradiation even radius and tank height were chosen equal to 7.5 and 14 m, 105
49 though a proper, fuel based, correlation (Eq. (10)) is respectively, the typical tank dimensions in Italian tank
U

employed for emissive power evaluation. farms. 107


51 The procedure described in the previous paragraph was
applied to study the risks associated with an event of fire 4. Irradiation results and cooling system sizing 109
53 over a storage tank of gasoline. The burning tank radiates
toward a second one, which is located in the downwind The model described through Eqs. (1)–(9) allows the 111
55 direction and has the same dimensions as the burning one. radiation problem described in Fig. 6 to be solved in terms
A parametric analysis was carried out by locating the target of local irradiation on the target surface. The wind speed at 113
57 tank at different distances from the flame, in the range of a height of 10 m was fixed at 5 m/s. Emissive power is

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
6 M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]]

1 calculated by means of Eq. (8). The calculations were 3000 30


performed by subdividing either the solid flame (tilted φ = 52° tank top 59
3 cylinder) or the target cylinder into 103 elements and then

Radiative heat transfer rate [kW]


2500 tank side 20

Maximum heat flux [kW/m2]


performing the integration. 61
5 The distribution of the heat flux over the target surface is Overall
2000 10
shown in Fig. 7. Here the separation distance is quite small 63
7 (x/r ¼ 4) and it can be observed that, due to this fact and
as a consequence of the uniform temperature of the flame 1500 0 65
9 cylinder, the radiative heat flux is mainly concentrated over
the tank top (i.e. the roof). In the wind conditions here 1000 -10 67
11 imposed, the flame is tilted toward the target tank with an
inclination of about 521. The calculated maximum irradia- 500 -20
69
13 tion is about 10.1 kW/m2 (at tank top), while the overall
heat flux over side and top surfaces is about 640 and 71
0 -30
15 940 kW, respectively.
0 2 4 6 8 10 12
The effect of distance on the irradiation on the target 73
Dimensionless distance x/R
17 tank is shown in Fig. 8, as a function of the dimensionless
distance x/R. Interesting to notice is how the heat transfer Fig. 8. Heat transfer rate and maximum heat flux on target tank as a 75
19 rate is distributed over the tank surface. At short distances, function of the dimensionless distance.
the irradiation is mainly concentrated over the tank top, as 77
21 a consequence of the flame tilt and flame height. At
distances larger than x/R ¼ 4, the irradiation mainly hits 3000 79
23 the tank side. The same figure shows the profile of the top, x/R = 2.5

F
maximum heat flux at different distances. top, x/R = 4 81
Radiative heat transfer rate [kW]

2500 side x/R = 2.5


25 Further calculations were carried out in order to assess
the influence of the flame tilt angle on the irradiated energy.
O side, x/R=4 83
O
27 Fig. 9 shows the results of the above analysis for three 2000 overall, x/R = 2.5
values of the tilt angle (01, 301, 451, 521), with the overall x/R = 4 85
PR

29 dimensionless tank distance as a parameter. It can be 1500


observed that, at short separation distances, the heat 87
31 transfer rate is nearly proportional to the flame tilt angle.
1000
Based on the above results on heat transfer rates, the 89
D

33 mass flow rates of water needed to cool the target tank


were calculated by means of steady state energy balances. It 500 91
TE

35 was assumed that cooling water is supplied uniformly by a


0
93
37
EC

0 10 20 30 40 50 60
Flame tilt angle [°]
95
39
14
12 Fig. 9. Heat transfer rate as a function of the flame tilt angle and 97
R

41 10 dimensionless distance.
8 99
z

43 6 ring at tank side upper part, as usually done in most


4
practical applications. The other assumption is that the 101
O

2
45 maximum allowed temperature increase DTnom is fixed at
C

30 55 1C (according to Agip 20244.VON.SAF.SDS norm): if 103


47 water is injected at 25 1C, the maximum allowed tempera-
N

24 ture increase due to irradiation is therefore 80 1C. The 105


x

49 -6 overall mass flow rate of water is hence calculated as the


U

0 ratio of the heat transfer rate (tank side) and of its enthalpy 107
18 y
51 6 variation in terms of nominal temperature variation,
Heat Flux [W/m2] DTnom (Eq. 14). The mass flux of water m _ 00L is finally 109
53 calculated with reference to irradiated side area Arad, which
corresponds, for most of the cases here considered, to 111
55 0 1000 2000 3000 4000 5000 6000 7000 8000 9000
about half the tank side area. Only at the shortest distance
Fig. 7. Local irradiation (W/m2) from solid flame over target surface, tank (x/R ¼ 2.5) and maximum tilt (521) does the flame 113
57 2. Dimensionless distance x/r ¼ 4. overcome the target tank and irradiate toward the overall

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]] 7

1 tank side area A. 2.4


59

Safety coefficient for water flow rate


Z
3 1 G
_ 00L
m ¼ dA. (14) 2.2
Arad A cDT nom 61
5
The overall mass flow rate of water can thus be 63
2
7 calculated by multiplying the water mass flux by the
overall area. Furthermore, it is also possible to calculate 65
9 the local water temperature variation along the tank side. 1.8
Water is distributed at 25 1C, at a vertical position near 67
φ = 52°
11 tank top, and it runs down under gravity. If the heat flux
1.6 φ = 30°
were uniform at tank side, water would reach the tank base 69
φ = 0°
13 always at a temperature of 80 1C. Due to the non-
uniformity of the radiative heat flux, certain regions (the 1.4 71
15 most exposed to the flame) are more irradiated and the
0 2 4 6 8 10
water temperature can exceed the maximum allowed value 73
Dimensionless distance x/R
17 and even reach the boiling condition.
Fig. 10 shows the locus of circumferential positions Fig. 11. 75
19 where the water temperature attains the maximum allowed
temperature (i.e. 80 1C, continuous lines). Here the results 77
21 are organised in a graph where the abscissa is the angular even boiling conditions can occur (regions below the dotted
distance from downwind direction (origin corresponds to lines, 100 1C isothermal profile). 79
23 the position on the x-axis of Fig. 6) and the ordinate is the Calculations were finally performed by varying either the

F
dimensionless tank height z/H. Together with the 80 1C flame tilt angle or the distance between tanks. It was thus 81
isothermal profile, the 100 1C isothermal curve is depicted possible to infer the extra amount of water (with respect to
25
as dotted line. When the isothermal function reaches the z/
H elevation equal to zero (tank side bottom), all the tank
O
that calculated as an average value, Eq. (14)) needed to 83
O
27 cover the tank side without exceeding the maximum water
height is locally protected by water at temperatures below temperature value (80 1C). This extra amount of water 85
PR

29 the assigned threshold values. A value of the isothermal (with respect to that calculated according to overall heat
function different from zero means that only a fraction of transfer rate) was evaluated in terms of safety coefficients 87
31 the tank side height is protected by water at temperatures to be applied to overall water flow rate. The results are
below the maximum allowed values. sketched in Fig. 11, where the safety coefficients, as defined 89
D

33 Fig. 10 shows the results for two parametric distances, x/ above, are expressed in terms of the dimensionless tank
R ¼ 3 and 10; both profiles refer to a flame tilt angle of 521. distance and flame tilt angle. 91
TE

35 It can be observed that the regions of the tank side facing It can be noticed that the nominal temperature condition
the flame are not properly shielded by the water film, and for water all over the tank side can be achieved by roughly 93
37 doubling the overall mass flow rate calculated according to
EC

1 the overall heat transfer rate: typical values turned out to 95


39 Side TOP x/R=3 be in the range of 1.8–2.2.
352 97
R

41 0.8 5. Conclusions
x/R=10
99
R

100°C 1052
43 An integrated analysis aimed at the evaluation of
0.6
hazards related to pool fire events in tank farms of liquid 101
O

T=80°C
z/H

45 hydrocarbons have been presented and discussed. In


C

0.4 particular, a model for pool fire description was developed, 103
47 where the flame is a solid surface at lumped parameters and
N

the thermal radiation is calculated by an algorithm for the 105


49 0.2 exact numerical solution of the view factor problem in
U

three-dimensional environments. The radiation model was 107


51 validated by means of extensive comparison with available
0
literature data on view factors from tilted cylinders and 109
-90 -60 -30 0 30 60 90
53 with respect to field measurement on radiation from large
angular distance from wind direction [°] diameter pool fires. The model was applied to a test case 111
55 Fig. 10. Isothermal water profiles along tank side surface. Flame tilt angle
where a burning tank radiates to another of similar
521. Parameters: dimensionless distance and water temperature (contin- geometry. The heat transfer rate values were calculated 113
57 uous lines, T ¼ 80 1C, dotted lines T ¼ 100 1C). for different separation distances between tanks in order to

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
FISJ : 1614
ARTICLE IN PRESS
8 M. Fossa, F. Devia / Fire Safety Journal ] (]]]]) ]]]–]]]

1 evaluate the mass flow rates of water needed for cooling the [8] Hottel HC. Certain laws governing diffusive burning of liquids. Fire 43
target tank side. The amount of water was calculated Res Abstr Rev 1959; 1–41.
[9] Attalah S, Allan DS. Safe separation distances from liquid fuel fires.
3 according to a steady state energy balance under usual 45
Fire Technol 1971;7:47–56.
assumptions on maximum allowed temperature increase of [10] Babrauskas V. Estimating large pool fire burning rates. Fire
5 cooling fluid. Although the test cases here considered refer Technology 1983; 251–61. 47
to a well defined geometry and cannot be considered a [11] Burgess DS, Strasser A, Grumer J. Diffusive burning of liquid fuel in
7 comprehensive description of a fire event in a tank farm, open trays. Fire Res Abstr Rev 1961;3:177. 49
some general conclusions and design criteria can be drawn. [12] Thomas PH. The size of flames from natural fires. In: Ninth
international combustion symposium. New York: Academic Press;
9 In particular it is demonstrated that, due to non-uniform 1963. p. 844–59. 51
heat flux distribution on tank side, even conditions of [13] Binding TM, Pritchard MJ. Fire 2: a new approach for predicting
11 incipient water boiling can occur. Moreover, it was thermal radiation levels from hydrocarbon pool fires, IChem E Symp 53
demonstrated that, in most cases where the flame is Ser, 1992; 3(130).
13 considered at uniform temperature, a considerable part of [14] American Gas Association, LNG Safety Program, Interim Report on 55
Phase II Work. Report IS 3-1, 1974.
radiation is directed toward tank the top, where usually [15] Mudan KS, Croce PA. Fire hazard calculations for open hydro-
15 foam protection only is applied. Further work on this carbon fires. SFPE handbook of fire protection engineering. 2nd ed., 57
subject could consider the effects of the radiation from a Quincy: National Fire Protection Association; 1995. pp. 197–240.
17 non uniform emission flame, as proposed in the work by [16] Lautkaski R. Validation of flame drag correlations with data from 59
large pool fires. J Loss Prev Process Ind 1992;5(3):175–80.
Rew et al. Finally, in order to assure proper and safe tank
[17] Burgess DS, Hertzberg MG. Radiation from pool flames. Heat
19 cooling, safety coefficients (to be applied to water overall transfer in flames, Afgan and Beer, 1974. 61
flow rates calculated according to the overall heat transfer [18] In: van den Bosh CJH, Weterings RAPM, editors, Methods for the
21 rate) were inferred also in terms of tank mutual disposition. calculation of physical effects—yellow book. The Hague (NL); 1997. 63
[19] Wayne FD. An economical formula for calculating atmospheric
23 Uncited references infrared transmissivities. J Loss Prev Proc Ind 1991;4:86–92. 65

F
[20] Hamilton DC, Morgan WR. Radiant-interchange configuration
factors, NASA TN 2836, 1952.
25 [24]. O
[21] Rein RG, Sliepcevich CM, Welker JR. Radiation view factors for
tilted cylinders. J Fire Flammability 1970;1:140–53.
67
O
27 References [22] Guelzim A, Souil JM, Vantelon JP. Suitable configuration factors for 69
radiation calculation concerning tilted flames. J Heat Transfer
1993;115(2):489–92.
PR

29 [1] Mudan KS. Thermal radiation hazards from hydrocarbon pool fires. 71
Prog Energy Combust Sci 1984;10:59–80. [23] Heskestad G. Luminous heights of turbulent diffusion flames. Fire
Saf J 1983;5:103–8.
[2] Shokri M, Beyler CL. Radiation from large pool fires. J Fire Prot Eng
31 1989;1:141–50. [24] Heskestad G. Fire plumes. In: SFPE Handbook. 2nd ed. National 73
[3] Rew PJ, Hulbert WG, Deaves DM. Modelling of thermal radiation Fire Protection Association; 1995.
D

33 from external hydrocarbon pool fires. J Process Saf Environ Prot [25] Koseki H. Radiation properties and flame structure of large 75
hydrocarbon pool fires. In: 30th UJNR meeting, March 13–30,
1997;75:81–9.
TE

[4] Gaudes Soares C, Teixera AP. Probabilistic modelling of offshore Gaithersburg, MD, 1996.
35 [26] Koseki H. Combustion properties of large scale liquid pool fires. J 77
fires. Fire Saf J 2000;34(I):25–45.
[5] Muñoz M, Arnaldos J, Casal J, Planas E. Analysis of the geometric Fire Technol 1996; 241–55.
37 [27] Hagglund B, Persson LE, The heat radiation from petroleum fires, 79
and radiative characteristics of hydrocarbon pool fires. Combust
EC

Flame 2004;139:263–77. FOA Report, Stockholm, 1976.


[6] Koseki H, Iwata Y, Natsume Y, Takahashi T, Hirano T. Tomakomai [28] Johnson AD. A model to predict thermal radiation hazards from
39 large-scale pool fires. Icheme Symp Ser 1992; (130): 507–24. 81
large scale crude oil fire experiments. Fire Technol 2000;36(1):24–38.
[7] Blinov VI, Khudianov GN. The burning of liquid pools. Dokl Acad
R

41 Nauk, SSR 1957;113:1094. 83


R
O
C
N
U

Please cite this article as: Fossa M, Devia F. A model for radiation evaluation and cooling system design in case of fire in tank farms. Fire Safety J
(2007), doi:10.1016/j.firesaf.2007.01.005
View publication stats

You might also like