Content uploaded by Vladimir Shigunov
Author content
All content in this area was uploaded by Vladimir Shigunov on Feb 23, 2014
Content may be subject to copyright.
Fire Investigation in a Container
By Ould El Moctar, Daniel Povel, Vladimir Shigunov & Alexander Tide
ABSTRACT
Container carriers are in general a safe ship type. Nevertheless, risk from some types of accidents can
be reduced in a cost-efficient way. An example is fire and explosions, which contribute to about third of
recorded fatalities onboard container ships and provide the second-largest contribution to the overall cost of
accidents. Fire and explosions in the cargo area are especially difficult to handle, because both fire detection
and, especially for deck cargo fires, fire fighting are difficult.
A series of full-scale fire tests on two loaded 20-ft sea containers was carried out in order to estimate cargo fire
risks for container ships, assess the behaviour of unprotected steel construction in fire, and provide data for the
validation of numerical methods. Gas temperatures in the containers, temperatures of the steel construction,
and concentrations of gases were measured. The air supply was varied in the tests by partial elimination of the
bottom and opening the door in order to model a partial loss of integrity. The conditions varied from a closed
container with a strongly under-ventilated fire to flashover with strong convective flames and high temperatures.
In addition, numerical simulations were carried out in order to enhance the understanding of the phenomena.
The results show that in none of the tests a destruction or some damage of the tested containers occurred,
which can compromise container integrity or stacking safety.
Key words: fire, model test, simulation
1 Introduction
Casualty data, e.g. LMI (2004), show that container ships are in general a safe ship type for
crew and cargo. Nevertheless, risk from some types of accidents can be reduced in a cost-efficient
way. For example, fire and explosions contribute to more than third of fatalities and injuries onboard
container vessels and provide the second-largest contribution to the overall incident costs, Tossevikeu
and Bergmann (2003).
Although the majority of fire and explosion events originate in the main engine room and
machinery spaces (51 % compared to 24 % in the cargo areas, MSC (2007)), there is a significant
risk-reduction potential related to cargo areas, especially taking into account the tendency towards
larger container carriers. Most cargo area fires onboard container ships originate in the hold (62 %)
and only about 19 % on the deck, MSC (2009). Modern container carriers are equipped with fixed
fire extinguishment systems for enclosed cargo holds, usually
CO2
systems. Such systems operate
satisfactorily if the hold remains sealed. However, analysis of the largest fire accidents onboard
container ships, e.g. Tossevikeu and Bergmann (2003), shows that all of them led to an explosion above
or below the hatch, which damaged the hatch cover and compromised the efficiency of the fixed fire
extinguishment system, initiating simultaneously strong fires, which made access to the scene with
manual equipment difficult. Fires on open decks are difficult to control because of the lack of thermal
or structural subdivision. In addition to access problems, resulting from container stacking and small
40
clearances, the fire hoses and portable extinguishers used on open decks are not suitable for high
stacks of containers and require application from a close distance, which may be impossible.
The development of cost-efficient fire detection and fire-fighting systems requires deeper
understanding of the processes of burning in a single container, as well as mechanisms of fire
propagation from one container to another. Previously, Dausendschön et al. (2007) reported the results
of a series of numerical simulations for a 40 ft sea container carried out in the project SAFEDOR.
In the simulation for an intact container, the openings were represented only by side vents and
gaps around the doors and between the floor panels, insufficient for continuous burning. 360 s
after ignition, the fire spread over the top of the cargo blocks and the maximum temperature of the
unexposed side of the roof achieved its peak of about 770
◦
C. The mass fraction of the oxygen in the
upper half of the container was below 5 %, therefore burning starts decreasing and extinguishes 380
s after ignition. However, the achieved maximum temperatures of the unexposed side of the roof
were sufficient for the ignition of the wooden floor of the container on top. Because this can lead to
collapse of the floor and to further propagation of fire, simulations were also carried out with one
floor panel removed, as well as without the wooden floor altogether, in order to study the possibility
of fire escalation in container stacks.
In these two simulations, the initial fire development was similar to the first one with the
intact floor: the initial fire spread lasted for about 360 s. Unlike in the simulation with the intact floor,
the fire did not extinguish after the initial spread but achieved a quasi-steady state, because of the
sufficient supply of the fresh air through the broken floor. Burning took place near the opening in the
floor; the oxygen concentration apart from this region was almost zero. The maximum temperatures
of the unexposed side of the roof in both simulations were over 1000
◦
C, and above 500
◦
C on the
steady-burning phase, thus indicating the possibility of further fire escalation. A more detailed study
of such possibility is proposed in this paper.
2 Tests and measurements
In the project SAFEDOR, a series of full-scale fire tests was carried out with two loaded 20 ft
sea containers (external length 6.058 m, width 2.438 m and height 2.591 m; internal dimensions
5.898 ,2.350 and 2.394 m, respectively) in order to investigate the fire development in a single with
varying degree of integrity, assess the fire effect on the container structure, estimate the likelihood of
fire propagation to other containers, and provide data for the validation of numerical methods.
The containers were made of steel plates of the thickness 1.6mm on the walls and 2.0mm on
the door and roof; the floor was fabricated of 28 .0mm thick wooden beams. The containers were
mounted on a concrete surface at a height reproducing the clearance in a container stack. The front
of each container was equipped with a double door; its perimeter was sealed with rubber. Under the
roof, two ventilation openings (50.0by 100 .0mm) were made in each longitudinal wall. Fire load
was produced by 10 stacks of 14 wooden euro-pallets each (a euro-pallet has a length 1.2m, width
0.8m and height 0.144 m) as shown in Fig. 1; the total load was 3500 kg. For ignition, 1.08 l of
isopropanol was used in two trays, put under the forward left stack of pallets.
The following measurements were carried out (see Fig. 10 for the locations):
•
gas temperatures at 4locations distributed along the container, with 4vertically distributed
measurement points per location - 16 measurement points altogether (A1 to A4, B1 to B4, C1 to
C4 and E1 to E4), and at 5measurement points distributed in the container 0.1m below the
roof (D1 to D5)
•
temperatures on the unexposed side of the container structure and steel plates: 9measurement
points on the roof (DA1 to DA9), 7on the right longitudinal wall (AR1 to AR7), 4on the door
(T1 to T4) and 3on the right forward pillar (ST1 to ST3)
•
concentrations of the molecular oxygen
O2
, carbon dioxide
CO2
and carbon monoxide CO at
the middle of the container at the heights 0.1m under the roof and 0.8m above the floor
(points G1 and G2, respectively)
41
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
5.9 m
2.35 m
door
ignition
tray
Fig. 1: Distribution of the stacks of pallets in the container. The dashed line shows opening in the floor in test 2.
Temperatures were measured with NiCr-Ni mantle thermocouples and registered with an
interval of 10 s. Measurement locations G1 and G2 were connected to gas sensors via two thin
probe pipes of different lengths. This caused time lag between the temperature and gas contents
measurements, therefore, in the presentation of the time histories, the time for the plots of gas
concentrations was shifted so that time histories for G1 and G2 are synchronised with the temperature
sensors B1 and B3, respectively. In addition, video and photographs (normal and infra-red) were taken
during the tests.
Ignition was initiated with an open flame, after which the door was closed. The air supply was
varied by partial elimination of the floor, in order to model a partial loss of integrity. In addition,
the door was open in test 3after 132 minutes of the test in order to initiate a natural fire. For
the conditions listed in Table 1 (with the exception of the second part of test 3), one would expect
strongly under-ventilated fires, which reduce to either a quasi-steady state, pulsating flame, or totally
extinguish after the consumption of the initial oxygen contents. Test conditions are shown in Table 2.
Tab. 1: Tests.
container 1 container 2
test 1 closed container closed container
test 2 removed floor 1.1×1.1m forward left removed floor 1.1×1.1m forward left
test 3 removed 3.0×2.4m of floor forward
open door after about 132 minutes
test 4 entire floor removed
Tab. 2: Test conditions.
test 1 2 3 4
ambient temperature, ◦C 4643
atmospheric pressure, ·10 5Pa 1.03 1.03 1.08 1.012
relative air humidity, % 83 83 96 90
mass humidity of wood pallets, % 18 18 17 17
3 Numerical method
Numerical simulations were carried out with the program Fire Dynamics Simulator (FDS)
version 5.2, McGrattan et al. (2009). FDS solves conservation equations of fluid dynamics and heat
transfer with an emphasis on fire development and smoke propagation. Conservation equations of
42
mass, momentum and energy for a mixture of ideal gases are solved in the limit of low Mach number
thermally driven flow, thus the method cannot consider acoustic or detonation waves. On the other
hand, the time step of numerical solution is bounded by the flow speed rather than sound speed.
Turbulence is modelled using large eddy simulation with Smagorinsky subgrid stress model.
Spatial derivatives of the conservation equations are discretised with second-order finite
differences on a Cartesian grid (one or more grids with different cell size can be used in different flow
regions). Scalar variables are assigned to cell centres, while vector components to the corresponding
normal cell faces. Time derivatives are approximated with an explicit second-order predictor-corrector
method. The Poisson equation for pressure is derived by taking divergence of the momentum
equations and solved with a direct Fast Fourier Transform method. The robustness of the staggered
grid discretisation and the use of fast direct solver for the pressure lead to highly efficient numerical
solution. However, using Cartesian grids can be a limitation when flow boundaries do not align with
a rectilinear grid or boundary layer resolution is necessary.
For thermally thick solid boundaries, heat conduction is modelled using one-dimensional
heat conduction equation in the normal direction to the surface. Otherwise, the temperature of a
solid boundary is assumed uniform throughout the thickness. Convective heat exchange across the
boundaries between solid bodies and gas is modelled using empirical correlations.
Combustion was modelled using a mixture fraction description. The mixture fraction is a
conserved scalar, equal for a given fluid particle to the fraction of the gas that originated as fuel.
The model assumes that combustion is controlled by the mixing rate between fuel and oxygen, and
once they are mixed, the reaction is instantaneous and complete for any temperature. This model
works rather well for large-scale well-ventilated fires. In underventilated fires, or if a suppression
agent is used, fuel and oxygen may mix without burning. Therefore, the model is supplemented
with empirical corrections for oxygen-limited burning, which specify the minimum concentration
of oxygen required for the reaction as a function of the local temperature. The mass fractions of
gas components of the reactants and products are derived directly from the mixture fraction using
algebraic state relations, which are usually obtained from measurements for a particular reaction.
The heat release rate from the fire is also derived directly from the mixture fraction field.
Radiative heat transfer between the solid surfaces and gas is modelled using radiation transport
equation for a non-scattering grey gas. Because soot contributes most to thermal radiation in large-
scale fires, and because the radiation spectrum of soot is continuous, the approximation of grey gas
is acceptable. However, this model is sensitive to the computed soot contents in the combustion
products and thus can lead to inaccuracies.
FDS was originally intended for simulations with the specified heat release rate from the fire
as a function of time; the aim of such simulations is the computation of the transport of the heat and
smoke taking into account the influence of the actual ventilation conditions. For computations where
the heat release rate is predicted rather than specified, pyrolysis from the fuel bed is controlled by a
threshold pyrolysis temperature (300
◦
C in the present simulations): once the solid surface achieves
this temperature, it starts to eject a specified combustible gas (propan in the present simulations) at a
specified mass rate per surface area. This approach requires the modelling of the heat conduction in
the solid fuel bad, as well as the definition of the heat flux to its surface due to radiation (especially
from the fire) and convective heat exchange with the gas phase. The uncertainty brought in by this
approach is rather high, because the properties of real materials and fuels are often unknown and
difficult to obtain. Besides, the models of combustion, radiative feedback, convective heat exchange
and heat conduction in the solid phase are simplified in FDS, and the results are sensitive to both
modelling and discretisation errors.
The simulation domain is extended beyond the container walls; the bottom boundary is treated
as wall, while all the others as pressure boundaries. In all simulations shown here, the numerical
grid consisted of 1.8
·
10
6
cells. The grid in the forward part of the container around the area of the
initial ignition consists of the cells of the size 25 mm, while in the other part of the container the
uniform cell size is 50 mm. Grid refinement studies show that the used cell size is not sufficient for a
quantitative assessment, but leads to qualitatively grid-independent results.
43
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
4 Fire in a container with intact floor
The aim of this test was to observe fire development in an intact container. The test was
repeated for two containers with similar results; only those for the second container are reported.
4.1 Measurements
Fig. 2 shows time histories of the measured temperatures and volume concentrations; Table 3
compares measured maximum temperatures with the other tests.
In this test, only a short rise of temperatures was registered locally in the area around the
initial ignition: after the initial fire development, the oxygen in the container was consumed and the
fire extinguished. The initial development lasted for about 230 s. The maximum gas temperature
was about 240
◦
C at measurement point D1, located 0.1m under the roof above the initial ignition;
lower maximum temperatures of 124 ,175 and 161
◦
C were registered at the neighbour measurement
points A1, C1 and D2, respectively, located 0.1 m under the roof. At the distance 3m from the door
into the container (point B1), the maximum gas temperature was 85
◦
C, and 5.4m from the door
(point E1), only 67
◦
C. The maximum gas temperatures decreased quickly also in the vertical direction
with distance from the roof: at the point C2 (0.8m under the roof), the maximum temperature was
36
◦
C, and at C3 (1.5m under the roof) only 21
◦
C. These temperatures were achieved later than the
maximum temperature at the point C1 – at about 274 s from the start.
The volume concentration of oxygen achieves its minimum of 10 % for the fully developed fire
at about 240 s from the start; the corresponding maximum volume concentrations of carbon dioxide
and carbon monoxide were 6.7and 1.5%, respectively. After this, the fire extinguishes because
of insufficient oxygen. After 20 minutes, all registered temperatures dropped back almost to the
initial temperature. After extinguishment, the oxygen concentration leveled over the entire container
volume due to diffusion and stayed constant at 13 % (volume concentration) until the end of the test.
After the opening of the container door, there were no signs of burning and no re-ignition of fire.
The maximum temperatures on the unexposed sides of the steel plates required more time
for development than gas temperatures because of the thermal inertia of the container structure –
about 310 s – and never exceeded 100
◦
C in this test. Post-test inspection revealed no damage to the
container. The wooden pallets showed local and superficial traces of the flame only in the area of the
direct influence of the initial fire.
4.2 Computations
Fig. 3 shows time histories of computed temperatures and gas concentrations. Note that the
fuel used in the computations (propan) does not produce carbon monoxide, therefore only the volume
concentrations of oxygen and carbon dioxide are shown.
The computed durations of the initial fire development (about 270 s) and fire self-extinguish-
ment correlate well with the measurements, although maximum computed temperatures are higher.
The temperatures rise quicker in the simulations, and the oxygen concentration also drops quicker
than in the test. This indicates quicker fire development, possibly caused by the used pyrolysis model.
Fig. 11 (top) shows measured and computed temperature distributions at 304 s after the start,
indicating qualitative agreement.
4.3 Fire in a container with a small opening
In this test, a small, 1.1m by 1.1m opening was cut in the wooden floor of the container
under the first left stack of pallets, see Fig. 1. Tests with the both containers showed similar results,
therefore only those for the first container are shown.
4.4 Measurements
Fig. 4 shows time histories of the measured temperatures and volume concentrations of gas
components; Table 3 compares maximum temperatures with the other tests.
44
time, s
gases G2, vol%
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
2
4
6
8
10
12
14
16
18
20
22 O2
CO2CO
time, s
temperatures A, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
20
40
60
80
100
120
140
A4
A1
A2 A3
time, s
temperatures B, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
20
40
60
80
100 B1
B2 B3
B4
time, s
temperatures C, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
20
40
60
80
100
120
140
160
180 C1
C2
C3 C4
time, s
temperatures E, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
10
20
30
40
50
60
70
E1
E2
E3
E4
time, s
temperatures D, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
50
100
150
200
250 D1 D2
D3
D4
D5
time, s
temperatures AR, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
5
10
15
20
25
30
35 AR1
AR6
AR5 AR7
AR3 AR2
time, s
temperatures T, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
5
10
15
20
25
30
T3
T2
T1
T4
time, s
gases G1, vol%
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
2
4
6
8
10
12
14
16
18
20
22 O
CO CO
2
2
time, s
temperatures DA, deg.
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
0
20
40
60
80
100 DA9 DA2
DA1 DA5
DA8
DA6 DA7
DA4
DA3
Fig. 2: Measurement results for test 1.
Similarly to the previous test, only a short rise of gas temperatures was registered, limited
to the domain above the initial ignition. Measured time histories show initial fire development,
lasting for about 270 s, followed by fire extinguishment. Similarly to the first test, the largest
maximum temperatures were achieved at the point D1 above the ignition location 0.1m under
the roof; this temperature (571
◦
C) is significantly higher than in the first test. The same relates to
45
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
time, s
volume fraction, %
0 50 100 150 200 250 300 350 400 450 500
0
2
4
6
8
10
12
14
16
18
20
22
O
CO2
2
time, s
steel temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500
0
50
100
150
200
250
T2
DA6
DA2 AR1
time, s
gas temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
700 C1
E1
A1
B1
Fig. 3: Temperatures of gas in the container (top) and on unexposed sides of steel
plates (middle), and volume concentrations of gases (bottom) computed for test 1.
the neighbour measurement locations under the container roof A1 and C1, where the maximum
registered temperatures were 198 and 308◦C, respectively.
The temperatures away from the door are also higher than in test 1 and lower than under the
roof above the ignition: maximum temperatures registered at the points B1 and E1 are 118 and 76
◦
C,
respectively; the same relates to the measurement points in the area of ignition further below the
roof: maximum temperatures at the points C2 and C3 are 116 and 23 ◦C, respectively.
The supply of oxygen through the floor opening appears insufficient to support burning; one of
the reasons is the observed emergence of the combustion products through the floor opening, which
reduced fresh air supply. During the initial fire development, the volume concentration of oxygen
dropped to about 6.8% – lower than in the first test; after the extinguishment of the fire, it increases
continuously until the end of the test due to the diffusion of the fresh air through the floor opening,
but does not lead to re-ignition. Until the 20th test minute, all temperatures reduced back to the
almost initial temperature.
Similarly to the first test, maximum temperatures on the unexposed sides of the steel plates
were achieved later than the maximum gas temperatures; they are slightly higher than in test 1:
151
◦
C on the roof (point DA6 above the ignition), 36
◦
C on the right wall (point AR2) and 29
◦
C on
the door (point T2). Post-test investigation revealed no damage to the container. The wooden pallets of
the load indicated only local and superficial traces of flame in the area of the direct influence of the
initial fire.
4.5 Computations
Fig. 5 shows time histories of computed temperatures and volume concentrations of gas
components. The computed temperatures for this test are slightly higher and fire development is
slightly longer than for the previous configuration because of the supply of the fresh air; however,
the results are very similar. Although there is an increased oxygen supply, it is insufficient to
support burning after the initial oxygen contents in the container is used up. The investigation of the
computed flow field in the container shows that effective gas exchange is blocked by the close ground.
Besides, the simulation shows that the combustion products eject from the container mostly through
the opening in the floor, because the ventilation openings are too small. This prevents efficient supply
of the fresh air. Fig. 12 (top) shows computed distribution of the molar concentration of oxygen for the
developed fire in the vertical longitudinal section at the distance 1.1m from the right wall. The flow
of the fresh air through the opening is clearly seen, as well as the low-oxygen air with combustion
products which flows first to the aft part of the container, is turned down at the aft wall, and then
flows towards the opening.
In general, the simulations reproduce fire development and extinguishment rather well.
However, similarly to the results for the first configuration, initial fire development is quicker,
46
time, s
temperatures A, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
20
40
60
80
100
120
140
160
180
200
A4
A1
A2 A3
time, s
temperatures B, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
20
40
60
80
100
120
B1
B2
B3
B4
time, s
temperatures C, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
50
100
150
200
250
300
350 C1
C2 C3 C4
time, s
temperatures E, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
10
20
30
40
50
60
70
80 E1
E2
E3
E4
time, s
temperatures D, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
100
200
300
400
500
600 D1
D2D3
D4
D5
time, s
temperatures AR, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
5
10
15
20
25
30
35
40
AR1
AR6
AR4
AR7
AR3
AR2
AR5
time, s
temperatures DA, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
20
40
60
80
100
120
140
160
DA9 DA2
DA3 DA1
DA8
DA6 DA7
DA4 DA5
time, s
temperatures T, deg.
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
5
10
15
20
25
30
T3
T2
T1
T4
time, s
gases G1, vol%
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
2
4
6
8
10
12
14
16
18
20
22 O
CO CO
2
2
time, s
gases G2, vol%
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000
0
2
4
6
8
10
12
14
16
18
20
22 O2
CO2CO
Fig. 4: Measurement results for test 2.
and the temperatures are higher than in the measurements. The differences are presumably due to
modelling and discretisation errors in the computation of pyrolysis.
47
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
time, s
volume fraction, %
0 50 100 150 200 250 300 350 400 450 500
0
2
4
6
8
10
12
14
16
18
20
22
O
CO2
2
time, s
steel temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500
0
50
100
150
200
250
T2
DA6
DA2 AR1
time, s
gas temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
700 C1
E1
A1
B1
Fig. 5: Temperatures of gas in the container (top) and on unexposed sides of steel
plates (middle), and volume concentrations of gases (bottom) computed for test 2.
4.6 Fire in a container with a large opening
In this configuration, the forward part of the floor area (about 7m
2
) near the door was removed.
In addition, after about 133 minutes of the test, the door was completely open in order to produce a
natural fire. In the simulations, only the first part of the test (with the closed door) was computed.
4.7 Measurements
Fig. 6 shows time histories of the measured temperatures and volume concentrations of the
gas components; Table 3 compares maximum temperatures with the results of the other tests. Tests
3a and 3b refer to the parts of the test with the closed and open door, respectively.
During the initial fire development in the first part of the test, significantly higher gas tempera-
tures were registered than in the previous tests. The initial fire development lasted for about 500
s and followed into a slowly declining ventilation-controlled fire. The maximum gas temperatures
after the initial fire development were achieved under the container roof above the ignition location,
e.g. 625
◦
C at the point D1 and 608
◦
C at the point C1. Away from this domain, the achieved peak
temperatures under the roof are progressively lower: 317
◦
C at the point D2, 246
◦
C at the point B1
and 184
◦
C at the point E1 (1.8,3.0and 5.4m from the door, respectively). The peak temperatures
also quickly decrease with the increasing distance from the roof: the maximum temperature at the
point C2 0.8m below the roof is 97 ◦C, and 1.5m below the roof (point C3) it is only 34◦C.
The volume concentration of oxygen drops to about 1.8% at the end of the initial development;
the corresponding concentrations of carbon dioxide and carbon monoxide achieve maxima of 14.1
and 3.2%, respectively.
Observers reported strongly oscillating smoke emergence from the ventilation openings and
from the opening in the floor.
After the initial development, a ventilation-controlled fire established, with slowly declining
intensity of burning on the average: at the 120 th test minute, the maximum temperatures 0.1
m below the roof above the ignition location (points A1, C1 and D1) were below 140
◦
C, while at
the locations B1 and E1 deeper into the container, only about 100 and 94
◦
C, respectively. The
corresponding volume concentration of oxygen was about 14 %. This declining behaviour interrupted
with rather short (about 4min duration) fire escalations about every 18 -20 min. of the test, with
maximum gas temperatures up to about 400◦C. The mechanism of these outbursts is unclear.
During the first part of the test, also significant rise of the temperatures on the unexposed
side of the roof was registered – e.g. 509
◦
C at the point DA6; on the right wall and the door, the
surface temperatures however have never exceeded 100
◦
C. Observers report local discolouration
and burning off of the paint layer on the roof and on the left wall. In the 120 th test minute, the
maximum roof temperature was only 101◦C.
48
time, s
temperatures A, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
700
800
900
A4
A1 A2 A3
time, s
temperatures B, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
700
B1 B2 B3 B4
time, s
temperatures C, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
700
800
900
1000
C1 C2 C3 C4
time, s
temperatures E, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
E1 E2 E3 E4
time, s
temperatures D, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
700
800
900
1000
D1 D2
D3
D4
D5
time, s
temperatures AR, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
700 AR1
AR6
AR4
AR7
AR3
AR2 AR5
time, s
temperatures DA, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
DA9
DA2
DA3
DA1
DA8
DA6
DA7
DA4
DA5
time, s
temperatures T, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
100
200
300
400
500
600
T3
T2 T1
T4
time, s
gases G1, vol%
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
2
4
6
8
10
12
14
16
18
20
22 OCO
CO
2
2
time, s
gases G2, vol%
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500
0
2
4
6
8
10
12
14
16
18
20
22 O2CO2
CO
Fig. 6: Measurement results for test 3; the dashed lines mark opening of the door.
After 133 minutes of the test, the door of the container was opened. The fire developed within
two minutes into a strong convective flame, ejecting from the container and achieving height of 2m
above the container roof. Gas temperatures under the roof achieved almost 900
◦
C at the points D1
and C1. Away from the door, maximum temperatures were significantly lower: 609
◦
C at the point B1
(3.0m from the door) and 402
◦
C at the point E1 (5.4m from the door). The volume concentration of
49
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
oxygen drops to zero during the flashover, while the volume concentrations of carbon dioxide and
carbon monoxide achieve maxima of 18.4and 6.9%, respectively. Significant rise of temperatures
was registered on the unexposed sides of the steel plates, achieving maxima of about 531
◦
C on the
roof (point D2) and 607 ◦C on the right wall (point AR4).
The fire was extinguished with water after about 10 minutes. Inspection of the container after
the test revealed damage to the paint layer and slight deformations of the steel plates of the roof,
walls and door. The rubber sealing of the door completely burned up. The elements of the container
structure (frames and pillars) showed no visible deformations.
4.8 Computations
Fig. 7 shows time histories of the computed temperatures and volume fractions of the gas
components. Only the first part of the test, with the closed door, was reproduced.
time, s
volume fraction, %
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
2
4
6
8
10
12
14
16
18
20
22
O
CO2
2
time, s
steel temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
50
100
150
200
250
300
T2
DA6
DA2
AR1
time, s
gas temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
100
200
300
400
500
600
700 C1
E1
A1
B1
Fig. 7: Temperatures of gas in the container (top) and on unexposed sides of steel
plates (middle), and volume concentrations of gases (bottom) computed for test 3.
Until the initial oxygen contents in the air is used up, the simulated fire development is similar
to the previous cases. The calculated gas temperatures are similar to the measured; however, the
initial fire development (lasting for about 280 s) is again much quicker than in the tests, which
can be related to the used pyrolysis model. Another reason of the differences relates to the oxygen
concentration necessary for the burning reaction: the measurements show that oxygen concentration
reduces to almost zero after the initial fire development in the first part of the test. For the combustion
model used, such concentrations are too low for burning reaction. Therefore the simulated fire
quickly reduces to a steady-state burning after the initial development. Because of the too quick fire
development and extinguishment, the temperatures of the unexposed surfaces are significantly lower
than in the measurements.
In this configuration, the floor opening is sufficiently large for both inflow of the fresh air and
removal of the combustion products. Fig. 12 (middle) shows the computed distribution of the molar
concentration of oxygen in the vertical longitudinal section 1.1m from the right wall of the container.
It shows the inflow of the fresh air in the forward part of the opening and outflow of the combustion
products in the aft part. Similarly to the measurements, the simulations show some oscillations of the
temperatures and gas concentrations in the quasi-steady phase of burning.
4.9 Fire in a container without floor
In this configuration, the wooden floor of the container was completely removed. It was
expected, that a prolonged fire would develop due to the sufficient gas exchange with the ambient
atmosphere. However, only short burning was registered after the ignition: already after 25 minutes
of the test, there was no indications of fire. Therefore, the test was stopped, the ignition trays filled
again and re-ignited in the 30 th minute of the test. Only the second part of the test (referred to as
test 4b) is discussed here.
50
4.10 Measurements
After the second ignition, the fire fully developed by the 33 rd minute of the test. Fig. 8 shows
time histories of the measured temperatures of the gas and unexposed sides of the steel plates, as
well as volume concentrations of gas components; Table 3 compares maximum temperatures with the
other tests (tests 4a and 4b refer to the first and second ignition attempts, respectively).
time, s
temperatures A, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
100
200
300
400
500
A4
A1
A2 A3
time, s
temperatures B, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
50
100
150
200
250
300 B1
B2
B3
B4
time, s
temperatures C, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
100
200
300
400 C1
C2 C3 C4
time, s
temperatures E, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
50
100
150
200
250 E1
E2
E3
E4
time, s
temperatures D, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
100
200
300
400
500 D1
D2
D3
D4
D5
time, s
temperatures AR, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
20
40
60
80
100
120
140 AR1
AR6
AR4
AR7
AR3
AR2 AR5
time, s
temperatures DA, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
100
200
300
400 DA9
DA2
DA3
DA1
DA8 DA6
DA7 DA4
DA5
time, s
temperatures T, deg.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
100
200
300
400
T3
T2 T1
T4
time, s
gases G1, vol%
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
2
4
6
8
10
12
14
16
18
20
22 OCO
CO
2
2
time, s
gases G2, vol%
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
0
2
4
6
8
10
12
14
16
18
20
22 O2
CO2
CO
Fig. 8: Measurement results for test 4; dashed lines indicate the second ignition.
51
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
After the initial fire development, intensive ventilation-controlled fire established, slowly
declining until the end of the test. The initial fire development after the second ignition lasted for
about 200 s and lead to the maximum gas temperature of 476
◦
C at the point A1 (0.1m under the
roof near the ignition). Again, maximum gas temperatures reduce with the distance from the ignition
location (228
◦
C at the point B1 at 3.0m from the door and 162
◦
C at the point E1, 5.4m from the
door) and with the decreasing height of the measurement point (160
◦
C at the point A2 and 74
◦
C
at A3, 0.8and 1.5m below the roof, respectively). The volume concentration of oxygen reduced to
3.0% during the initial development; the corresponding maximum concentrations of carbon dioxide
and carbon monoxide were 13.8and 3.1%, respectively.
After the initial development, intensive ventilation-controlled fire established, with oscillating
gas temperatures and concentrations of the gas components, with the oscillation period of about 5
min. Observers reported significant escape of smoke through the ventilation openings, floor opening,
and through the gaps along the door perimeter, where the rubber sealing has burnt up. On the
average, the intensity of burning slowly decreased until the end of the test, and after opening of the
container door in the 180 th test minute, there was no re-ignition despite strong local smoke sources.
Similarly to the previous tests, the maximum temperatures on the unexposed sides of the steel
plates follow maximum gas temperatures with some delay due to the thermal inertia of the container
structure: the maximum temperature on the unexposed side of the roof of 356
◦
C (point DA6) was
achieved at about 18 min., on the right wall (128
◦
C at the point AR1) about 28 min., and on the door
(347◦C at the point T2), about 16 minutes after the second ignition.
Post-test investigation revealed local damages to the painting layer, especially on the left wall
near the initial ignition and on the almost entire roof. In these areas, also insignificant deformations
of the steel plates were found. The frame and pillars showed no visible deformations. The wooden
loading pallets showed significant damage due to flame in the area around the initial ignition.
4.11 Computations
Fig. 9 shows time histories of the computed temperatures and volume concentrations of the
gas components. The computed time histories qualitatively reproduce the measured ones. The initial
fire development lasts for about 290 s, although the computed maximum temperatures are higher
than the measured ones. Similarly to the measurements, the fire reduces after the initial development
to oscillatory burning. The speed of oxygen consumption is also similar during the fire development;
however, the oxygen concentration achieves its minimum of 8.2% at the instant 290 s after the
start in the simulation, while the measured oxygen concentration continues decreasing further to
3.0%. The reason for this difference is the minimum oxygen contents required for burning in the
combustion model – in the test, lower oxygen concentration was sufficient. This might explain the
quicker fire decrease after the initial development in the simulation.
time, s
volume fraction, %
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
2
4
6
8
10
12
14
16
18
20
22
O
CO2
2
time, s
steel temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
50
100
150
200
250
300
T2
DA6
DA2
AR1
time, s
gas temper.,deg.C
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900
0
100
200
300
400
500
600
700 C1
E1
A1
B1
Fig. 9: Temperatures of gas in the container (top) and on unexposed sides of steel
plates (middle), and volume concentrations of gases (bottom) computed for test 4.
52
Fig. 12 (bottom) shows the computed distribution of the molar concentration of oxygen at the
time instant of the maximum fire development in the vertical longitudinal section at the distance 1.1
m from the right wall. The fresh air flows into container from the outside in the forward part of the
container, while combustion products leave container predominantly in the aft part. Fig. 11 (bottom)
shows measured and computed wall and roof temperatures at the 333th second of the test.
5 Conclusion
The study investigated the influence of the container integrity on the fire development and the
behaviour of the container structure in fire, in order to evaluate the likelihood of fire propagation to
adjacent containers.
For the intact container and a small floor opening (tests 1 and 2), the fire self-extinguished
within about 20 minutes due to lack of oxygen. There was only small local damage to the load; steel
plates of the walls and roof, wooden floor and container frame were not damaged.
The container with half floor removed (test 3) featured a slowly declining ventilation-controlled
fire. The temperatures inside the container and on its surface were higher than in the first two tests,
leading to local damage to the paint layer and small local deformations of the steel plates; however,
the container integrity was not compromised. When the door was opened (test 3b), the incoming
fresh air quickly led to the growth of available hot spots and to flashover within two minutes with
massive flames ejecting through the open door. The fire was extinguished with water after about
10 minutes. Despite local deformations and paint damage on the walls and roof, the integrity of the
container and the supporting structure was not compromised.
For the totally removed wooden floor (test 4), after two ignition attempts (tests 4a and 4b in
Table 3), a ventilation-controlled fire developed, reducing to a slowly decaying oscillatory burning.
After the opening of the container after 180 minutes, there were no visible flames and no re-ignition.
The paint layer of the unexposed surfaces showed some local damages, especially on the roof, and
the steel plates were locally slightly deformed; the container structure showed no damages.
None of the tests led to a damage which could compromise container integrity or stacking safety.
Burning through the floors or roof of stacked containers is unlikely for the moderately combustible
loading used in the tests; thus fire propagation to the neighbour containers is also unlikely. A
summary of the maximum measured temperatures is shown in Table 3.
Tab. 3: Maximum temperatures in ◦Cmeasured in all tests.
test 1 test 2 test 3a test 3b test 4a test 4b
D1 (under the ceiling) 240 571 655 899 399 385
C1 (under the ceiling) 176 308 647 896 368 388
DA6 (ceiling, unexposed side) 82 151 509 372 272 356
AR1 (right wall, unexposed side) 33 29 101 534 58 128
T2 (door, unexposed side) 26 29 96 539 108 347
Simulations with FDS reproduced the general development of the fire and differences between
the tested configurations well. The differences between simulations and measurements (quicker initial
fire development, higher temperatures, and quicker decay) can be attributed to the inaccuracies of the
pyrolysis (too strong) and combustion (performance at low oxygen concentrations) models. A further
factor may be insufficient grid resolution, leading to inaccurate prediction of the inflow of the fresh air
in the container and outflow of combustion products. In addition, some results are very sensitive to
the exact location of the observation points with respect to the ignition source and loading: because
of the heat shadows, small changes in the position can lead to significant differences in the results,
which could also contribute to some differences between the computations and measurements. More
detailed numerical studies are required, regarding required grid resolution, influence of material
properties, and pyrolysis and combustion models.
53
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010
E4
DA5
C1-C4 B1-B4 E1-E4
E1
E2
E3
A4
A3
A2
A1D1
C3
C2
C1
B4
B3
B2
B1
C4
D2
D1
D3
D2
D4-D5
D3
D4
D5
A1-A4
DA1
DA7 DA3 DA4DA2
DA6
AR6
DA9
G1
AR1 AR2
AR4
AR7
AR5
AR3
ST3
T1 T4
T2 T3
ST1
ST2
x
x
x
DA8 G1-G2
G2
1.3m
0.588m
0.7m0.7m0.7m
0.5m 0.5m
0.8m
0.1m
1.2m
0.588m
0.1m 0.5m 0.4m 0.8m 1.2m 1.2m
Fig. 10: Measurement points.
Fig. 11: Measured (left) and computed (right) temperature distribution on the unexposed sides of the container
walls and roof at 304 s from the star t of test 1 (top) and 333 s from the start for test 4 (bottom).
54
Fig. 12: Computed distribution of the molar concentration of oxygen for the developed fire in the vertical
longitudinal section 1.1 m from the right container wall for test 2 (top), 3 (middle) and 4 (bottom); the
door is on the left-hand side of the pictures.
References
Dausendschön, K., Owens, M., Povel, D. and Sinai, Y. (2007). Cargo safety - Integration of results. Rep. No. D2.5.7, SAFEDOR
LMI (2004). Casualty database. Lloyd Maritime Information
McGrattan, K., Hostikka, S., Floyd, J., Baum, H., Rehm, R., Mell, W. and McDermott, R. (2009). Fire Dynamics Simulator Version 5:
Technical Reference Guide. Volume 1: Mathematical Model. NIST Special Publication 1018-5
MSC (2007). FSA-Container vessels. Details of the Formal Safety Assessment). MSC 83/INF.8, IMO, submitted by Denmark
MSC (2009). FSA-container fire on deck. Details of the Formal Safety Assessment. Submitted by Germany
Tossevikeu, A. and Bergmann, J. (2003). Cargo fires on container ships. DNV Technical Paper Paper Series No. 2003-P013, DNV
55
Ship Technology Research Schiffstechnik VOL. 57 / NO. 1 January 2010