Polymer Testing 25 (2006) 7582

www.elsevier.com/locate/polytest

Property Modelling

Superposition of heat release rate curves

for combustibles with bench-scale tests
W.K. Chow *, S.S. Han
Areas of Strength: Fire Safety Engineering, Research Centre for Fire Engineering, The Hong Kong Polytechnic University, Hong Kong, China
Received 29 June 2005; accepted 2 September 2005

Abstract: In fire hazard assessment, the resultant heat release rate of burning different combustibles has to be known. The principle
of superposition is commonly applied to estimate the total heat release rate from the individual curves measured for single items.
Accuracy of such an approach will be studied with bench-scale tests in this paper.
The heat release rate curves of burning each sample cube of polymethylmethacrylate (PMMA), polyvinyl chloride (PVC),
polycarbonate (PC) and wood were first measured individually by a cone calorimeter. Radiative heat fluxes of 50 and 70 kW mK2
were applied. After that, a PMMA cube was burnt with a cube of another material under the same heat flux. The resultant heat
release rate curves of burning these two cubes were measured. Heat release rate curves of burning the single cube were used to
estimate the resultant curves. The technique of fundamental analysis will be applied for comparing the predicted curves with the
experiments. Importance of the parameter s for estimating the secant inner product cosine specifying the data points intervals will
also be discussed.
For the samples tested, it is observed that superposition gives good estimations of the total heat released curve if those for
individual items were measured under the same radiative heat fluxes. However, the results will not be so good where the curves
for each combustible were measured at different heat fluxes. This point is very important in estimating the possible heat release rate
for a design fire.
q 2005 Elsevier Ltd. All rights reserved.
Keywords: Functional analysis; Superposition; Bench-scale test; Heat release rate

1. Introduction
The heat release rate in a room fire has to be
understood [1,2] in hazard assessment. This will give
key information on the size of the fire; the rate of fire
growth and, consequently, the release of smoke and
toxic gases; the time available for escape or fire
suppression and the type of suppressive action that is
likely to be effective. Other attributes that define the fire
hazard, such as the possibility of having a flashover fire,
can be estimated.
* Corresponding author. Tel.: C86 852 2765 7198.
E-mail address: bewkchow@polyu.edu.hk (W.K. Chow).

0142-9418/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.polymertesting.2005.09.016

Different combustibles, including both movable and

fixed fuel load [3], are stored in a building. The rate of heat
release in burning these materials together should be
estimated. Mostly likely, only the heat release rates of
individual materials or single items are available. How the
curves can be combined to estimate the resultant heat
release rate curve [4,5] should be understood.
The heat release rate per burning surface area of a
material can be measured by a cone calorimeter [4,6].
Models based on the cone calorimeter results have been
developed to predict the heat release rate for burning
those materials in bigger rooms [5], and applied for
different applications such as for train compartments
[79].

76

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

The combined heat release rate Q 0 T for burning two

samples A and B with heat release rates Q 0 A and Q 0 B is
suggested to be [10]:
Q0 T Z Q0A C Q0B

(a)
Thermal radiation

(1)

This was tested before with a cone calorimeter

[11,12].
The heat release rate per unit area Q 0 c t curves of
different materials measured with bench-scale tests in a
cone calorimeter, time to ignition (TTI, in s), peak heat
release rate per unit area (pkRHR, in kW mK2) and
total heat released per unit area (THR, in MJ mK2) are
estimated. Note that THR is given by:

Cone tray
20 mm cube

100 mm
Test A

(b)

Tests B, C, D, E

THR Z Q 0 c tdt

100 mm

(2)

(c)

Rfc

In this paper, how the resultant heat release rate

of two different polymeric materials can be
combined is explored. Polymethylmethacrylate
(PMMA), unplasticized polyvinyl chloride (PVC),
polycarbonate (PC) and oak wood widely used in
the market as consumer products and construction
materials are taken as examples. Samples of those
materials were selected and cut into small cubes of
size 20 mm. The samples were exposed to the same
conditions in a cone calorimeter for measuring the
heat release rates under heat fluxes of 50 and
70 kW mK2. The samples were tested individually
by themselves first. A sample PMMA cube was
then burnt with another sample for comparing with
the calculated heat release rate using Eq. (1).

2. Samples tested
Experimental measurements on one or two sample
cubes of PMMA, PVC, PC and wood were conducted
in a cone calorimeter. The samples were placed at the
side of the cone tray as shown in Fig. 1. The radiative
heat flux of the cone was set at 70 or 50 kW mK2. In
following the procedures in the standard tests [13],
samples were placed at 25 mm below the cone. In this
paper, some samples were also tested by moving
down to 50 mm below the cone [14]. This will give
different ventilation conditions as in real fire
scenarios. Different heat release rate curves can then
be achieved.
The testing arrangements as shown in Table 1 are
summarized in the following:

Rff
Rfa

Schematic view of heat fluxes

Fig. 1. Cone calorimeter tests.

Test A: Two sample cubes at two sides of the tray

with arrangements:
A1: PMMA and PC at 70 kW mK2, 25 mm
under the cone
A2: PMMA and PVC at 70 kW mK2, 25 mm
under the cone
A3: PMMA and PVC at 70 kW mK2, 50 mm
under the cone
A4: PMMA and PVC at 50 kW mK2, 50 mm
under the cone
A5: PMMA and wood at 50 kW mK2, 50 mm
under the cone
Test B: PMMA cube only at one side of the tray with
arrangements:
B1: PMMA at 70 kW mK2, 25 mm under the
cone
B2: PMMA at 70 kW mK2, 50 mm under the
cone
B3: PMMA at 50 kW mK2, 50 mm under the
cone
Test C: PC cube only at one side of the tray with only
one test:

Table 1
Functional analysis on the superposition results

Q 0 c t

THR

Parameters

Norm
Cosine
(sZ1)
Cosine
(sZ2)
Cosine
(sZ3)
Cosine
(sZ4)
Cosine
(sZ5)
Norm
Cosine
(sZ1)

A1: PMMACPC

A2: PMMACPVC

A3: PMMACPVC

A4: PMMACPVC

A5: PMMACwood

Same
heat flux

Under different
heat fluxes

Same
heat flux

Under different
heat fluxes

Different
heat flux

Same
heat flux

Different
heat flux

Under different
heat fluxes

Same
heat flux

B1CC1

B2CC1

B3CC1

B1CD1

B2CD1

B3CD1

B1CD2

B2CD2

B3CD2

B1CD3

B2CD3

B3CD3

B1CE1

B2CE1

B3CE1

0.13
0.86

0.21
0.83

0.46
0.56

0.17
0.83

0.31
0.53

0.55
0.16

0.19
0.44

0.23
0.72

0.51
0.14

0.70
K0.07

0.5
K0.02

0.22
0.16

0.82
0.11

0.52
0.16

0.15
0.60

0.90

0.88

0.61

0.89

0.59

0.18

0.49

0.78

0.16

K0.07

K0.01

0.18

0.12

0.19

0.65

0.91

0.91

0.63

0.91

0.62

0.20

0.54

0.80

0.17

K0.07

K0.01

0.20

0.12

0.21

0.70

0.92

0.93

0.65

0.93

0.65

0.21

0.59

0.82

0.19

K0.07

0.01

0.22

0.13

0.23

0.75

0.93

0.93

0.66

0.94

0.68

0.23

0.63

0.83

0.20

K0.06

0.02

0.25

0.15

0.25

0.80

0.08
0.99

0.16
0.99

0.30
0.90

0.16
0.99

0.25
0.97

0.39
0.84

0.06
0.98

0.14
0.97

0.31
0.86

0.05
0.06

0.51
0.07

0.04
0.97

0.61
0.77

0.41
0.89

0.11
0.99

Under different
heat fluxes

Same
heat flux

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

Curves
tested

77

Each testing arrangement was tested several times to

check the repeatability [15]. Only one typical set of
results was used to study the superposition. Results of
the heat release rate per unit area Q 0 c t measured in the
cone and the total heat released per unit area THR for
each test are shown in Figs. 26.

Total heat released per unit area THR / MJm2

Heat release rate per unit area Q'c / kWm2

C1: PC at 70 kW mK2, 25 mm under the cone

Test D: PVC cube only at one side of the tray with
arrangements:
D1: PVC at 70 kW mK2, 25 mm under the
cone
D2: PVC at 70 kW mK2, 50 mm under the
cone
D3: PVC at 50 kW mK2, 50 mm under the
cone
Test E: Wood cube only at one side of the tray with
only one test:
E1: Wood at 50 kW mK2, 50 mm under the
cone.

Heat release rate per unit area Qc' / kWm2

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

Total heat released per unit area THR / MJm2

78

(a)

2500

A1
B1
C1
S

B1
A1

2000

PMMA+PC
PMMA at 70 kWm2
PC at 70 kWm2
Calculated

1500

C1

1000

(a)

B1

2500

A2

2000

A2
B1
D1
S

PMMA+PVC
PMMA at 70 kWm2
PVC at 70 kWm2
Calculated

1500

D1

1000
500
0

100

200
300
Time / s
Heat release rate per unit area

400

(b)
B1

500

A2

400
S

300
D1

200
100
0

100

200
300
Time / s
Total heat released per unit area

400

Fig. 3. Test A2.

500
0

100

200
300
Time / s
Heat release rate per unit area

(b)

400

B1

500
A1

400

300

C1

200
100
0
0

100

200
300
Time / s
Total heat released per unit area
Fig. 2. Test A1.

400

In a real fire, combustibles placed together are

exposed to different radiative heat fluxes. Therefore,
different external radiative heat fluxes and separation
distances among them should be assessed. For the two
samples as tested in this paper, there was a constant heat
flux emitted from the cone Rfc , a heat flux from the
adjacent burning sample Rfa and a heat flux feedback
from the flames Rff acting on the burning surface of the
sample, as shown in Fig. 1c. The distances between the
two samples might become shorter than 6 cm (even
2 cm for PVC) due to melting and swelling upon
burning. There might be stronger interaction between
the two combustibles due to the shorter distance.
Although Rfa might be higher than Rfc , both heat fluxes
are likely to be less than Rff . Except at very high values,
external heat fluxes such as Rfc would only be important
in ignition. Effects of these couplings should be further
studied quantitatively but are not included in this paper.
PVC samples were difficult to ignite under lower
heat fluxes. Therefore, higher heat fluxes of 50 and

(a)

B2
A3

A3
B2
D2
S

PMMA+PVC
PMMA at 70 kWm2
PVC at 70 kWm2
Calculated

1500
S

1000

D2

500

0
0

100

200
300
Time / s
Heat release rate per unit area

400

(b)
500

B2

400

A3
S

300
D2

200
100
0
0

100

200
300
Time / s
Total heat released per unit area

400

Heat release rate per unit area Q'c / kWm2

2000

Total heat released per unit area THR / MJm2

Total heat released per unit area THR / MJm2

Heat release rate per unit area Q'c / kWm2

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

2000 (a)

1500

79

B3

A4

A4
B3
D3
S

PMMA+PVC
PMMA at 50 kWm2
PVC at 50 kWm2
Calculated

1000
D3

500

0
0

200

400
Time / s
Heat release rate per unit area
(b)

600

B3

500

A4

400

300
D3

200
100
0

200

400
Time / s
Total heat released per unit area

600

Fig. 5. Test A4.

Fig. 4. Test A3.

70 kW mK2 were used. As observed in the tests, PVC

samples melted quickly and evaporated into fuel
vapour. A large quantity of smoke with irritating
smell was liberated upon exposure to the heat fluxes.
Among the four samples tested, PVC was the most
difficult to burn with the longest ignition time. Further,
PVC cubes did not burn steadily with charring. Taking
test A4 as an example, the heat release rate oscillated
with several peaks.
There was smouldering at first in burning the wood
samples. Char was formed later at the burning stage.
PMMA was ignited easily and burnt completely with a
steady rate. The PC samples were also difficult to burn.
However, once ignited, they burnt vigorously with
smoke liberated. Except for PMMA, some residues
were left after burning.
Rate of heat transfer inside the sample from the
external heat flux depends on the effective exposure
area and the distance away from the conical heater.

The heat flux might be reduced by 515 kW mK2 when

the distance of the sample from the heater was moved
from 25 to 50 mm when set at 50 and 70 kW mK2.
There were differences in the exposure areas for the
different samples upon burning. The melted PVC is an
obvious example. Results of heat release rate per unit
area deduced from the cone would be affected. The
exposure areas for all sample cubes were taken to be the
same upper surface area of 4 cm2. The accuracy of the
heat release rate would be affected by the above factors.
However, those effects should be the same to all
samples, giving very little deviations in judging the
superposition principle.
3. Superposition of heat release rate curves
Whether the individual heat release rate curves of
two different materials can be added together (i.e.
superposition) to give the resultant heat release rate
while burning both of them will be assessed by the
measured results [2,1012,16,17].

80

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

Heat release rate Q'c / kWm2

2000 (a)

A5
B3
E1
S

PMMA+Wood
PMMA at 50 kWm2
Wood at 50 kWm2
Calculated

1500
S

1000
A5

500
E1

0
0

Total heat released per unit area THR / MJm2

B3

200

400
Time / s
Heat release rate per unit area
(b)

600

B3

500

4. Functional analysis

400
S

300
A5

200
100
0
0

E1

200

400
Time / s
Total heat released per unit area

600

Fig. 6. Test A5.

From the heat release rate per unit area curves Q 0 CA

and Q 0 CB of two samples A and B, the transient heat
release rate per unit area estimated Q 0 CTE is:
Q 0 CTE t Z

release rate for one sample by burning it alone will be

the same as that from burning it together with other
combustible under the same conditions. Therefore, the
upper limits of predicted results can be determined by
applying the results tested under the same standard
external heat fluxes. Test results of each sample
exposed to the same conditions could be added, though
more tests should be carried out to confirm this.
Another method is to calculate the increase in heat
flux by the adjacent burning item. The geometries of the
two combustibles and the relevant flames, the variable
heat flux and other potential factors should be
considered. These aspects will be further reported
separately later.


1 0
Q CA t C Q 0 CB t
2

(3)

Experimental curve on the heat release rate per unit

area for burning the samples A and B Q 0 CT t under the
same heat flux and the calculated curve Q 0 CTE t from
Eq. (3) are compared in Figs. 26.
Adding the heat release rate curves of two
combustibles by superposition is useful in predicting
real fire scenarios when the combustibles are not placed
too close to each other. The results can at least be taken
as a minimum estimation.
There might be some other effects which are more
obvious for bigger fires. A correction factor might be
required in using superposition. Assuming these effects
can be neglected under a specified standard external
heat flux higher than normal heat fluxes, the net heat

Instead of comparing the estimated results from

superposition with the experiments in qualitative terms
such as good, satisfactory or bad, functional
analysis proposed for evaluating fire models by
Peacock et al. [18] is used. Fire model predictions
have been compared with test data by Friday et al. [19]
with such approach.
As both experimental and predicted data can be
described by transient curves, functional analysis
would quantify the difference between two curves in
terms of magnitude and shape. The data points within
each curve are described by vectors, summing them up
would give a resultant single vector for each curve. The
distance between the resultant vectors for the predicted
and measured curves is the error. This error can be
normalized to provide a relative difference, or norm,
between the curves. The following parameters will be
calculated:
The parameter norm is a measure of the relative
difference in magnitude of the two curves.
The parameter inner product or cosine describes the
angular difference between the resultant vectors to
provide a quantitative measure on the similarity of
the curve shape.
For good agreement between the experimental and
predicted curves, the value of the norm is expected to be
close to zero, and the value of the cosine is expected to
be close to one.
Following the recommendation by Peacock et al.
[18], the Euclidean norm is calculated by the ith
experimental and model values Ei and mi at the ith time
increment ti with s data points to be considered at each

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

increment inside as:

s
n
P
Ei Kmi 2
iZ1
jEKmj
Z s
jEj
n
P
Ei 2

81

under higher heat fluxes, under which the combustion

was more complete.

(4)

6. Curves under different radiative heat fluxes

For real fire scenarios, the heat release rate of burning a
combustible will be affected by the total heat feedback
and the total external heat fluxes, which might be varied

iZ1

The secant inner product cosine is:

n
P

Ei KEiKs mi KmiKs =s2 ti KtiK1 

hE; mi
iZsC1
Z s
jEj,jmj
n
n
P
P
Ei KEiKs 2 =s2 ti KtiK1 ,
mi KmiKs 2 =s2 ti KtiK1 
iZsC1

(5)

iZsC1

The parameter s R1 would smoothe the results to

give better estimates of large-scale differences. Higher
value of s might not overcome the effects of small-scale
noise between dense data, depending on the shapes of
the curves. Values of s will be varied as 1, 2, 3, 4 and 5
in this paper to investigate its effect on the secant inner
product cosine.

5. Superposition of curves under the same radiative

heat fluxes
Norms and inner product cosines were calculated for
the curves of Q 0 c t and THR for each case under the
same radiative heat fluxes. The values were computed
over the burning duration period of 400 and 600 s for
radiative heat fluxes of 70 and 50 kW m K2,
respectively.
Functional analysis results of the point-to-point
comparison are presented in Table 1. It is observed
that the values of the norm are lower than 0.23,
indicating very good predictions on curve magnitude.
The shapes of the curves are in fact very close, as shown
in the figures.
For s equal to 1, the computed values of cosine at
higher heat fluxes are higher than 0.72. However, the
values of cosine for Q 0 c t are 0.16 for test A4 and 0.60
(with wood at 50 kW mK2) for test A5. A possible
reason might be because wood and PVC samples did
not burn steadily under the lower heat flux of
50 kW mK2.
Values of cosine would be higher for higher values
of s. For example, the values of cosine for tests A4 and
A5 would be changed to 0.25 and 0.80, respectively, by
taking s as 5.
Results on comparing the curves with functional
analysis suggested that superposition is better for tests

with the distance and exposure area. The result of one

sample tested in other conditions can be taken as a
reference curve, which might be varied under the same
initial conditions. Deviation of the calculated results by
superposition can be estimated by functional analysis.
If the curves under different heat fluxes are added
together, say B3 (under 50 kW mK2) with C1 (under
70 kW mK2) instead of B1 with C1, both the norm and
cosine compared with test A1 deviated from the
matching value of 0 and 1.0. The values of norm and
cosine are 0.46 and 0.56 respectively for B3 with C1
when s is 1. The value of cosine only increased to 0.66
when s is 5.
For curves under the same heat flux but at different
distances below the cone, say combining B2 of
70 kW mK2 for 50 mm and C1 for 25 mm, the values
of norm and cosine are 0.21 and 0.83, respectively. The
value of cosine increased up to 0.93 when s is 5.
Therefore, combining the curves measured under the
same heat flux but at different distances below the cone
would not give results deviating so much from the
experiment.

7. Conclusions
Samples of PMMA, PVC, PC and wood in different
arrangements were tested with a cone calorimeter under
different heat fluxes and distances from the conical
heater. Several tests were repeated for each testing
arrangement to ensure its repeatability. One typical set
of results was used to study the superposition.
From the above study, the heat release rate of
burning two material samples together can be estimated
by simple addition, i.e. by superposition of the curves
measured for each sample under the same heat flux.
Values of norm and cosine gave better agreement for

82

W.K. Chow, S.S. Han / Polymer Testing 25 (2006) 7582

the higher heat flux of 70 kW mK2. Examples of

burning PMMA with PVC, PC or wood demonstrated
this. Superposition can further be applied to burning
multiple combustibles. The ignition time, burning time
and peak heat release rate are key points to be
considered for superposition. The resultant curve can
be put into a computer model for simulating the fire
environment in hazard assessment.
Functional analysis suggested that the predicted
curves agreed better with the measured curves for the
tests under higher heat fluxes. Results of superposition
would be better when different combustibles are
exposed to higher external heat fluxes, such as a
flashover fire with relatively less interaction, or when
they are placed not so closely together with relatively
independent burning.
However, the results of superposition might be
underestimated for real fires when the combustibles are
placed close to each other and with lower external heat
fluxes from the ceiling, walls and smoke layer. High
external thermal radiation might give more complete
combustion. The results can be taken as a minimum
estimation. The coupling effects between the combustibles, external heat and ambient conditions should also
be considered.
Many more tests, especially full-scale burning
tests [16,20,21] should be carried out on those
samples to support the application of the superposition principle.
Acknowledgements
This paper is supported by the RGC project
Determination of the concentration needed for extinguishing fires with clean agent heptafluoropropane
(FM200) under Grant No. B-Q669.

References
[1] C. Huggett, Estimation of rate of heat release by means
of oxygen-consumption measurements, Fire Mater. 4 (1980)
6165.
[2] R.D. Peacock, R.W. Bukowski, W.W. Jones, P.A. Reneke, V.
Babrauskas, J.E. Brown, Fire safety of passenger trains: a review
of current approaches and of new concepts, NIST Technical
Note 1406, National Institute of Standards and Technology,
Maryland, USA, 1994.
[3] W.K. Chow, C. Cheung, Aspect of fires for factories in Hong
Kong, J. Appl. Fire Sci. 5 (1) (1996) 1732.
[4] V. Babrauskas, S.J. Grayson, Heat Release in Fires, Elsevier,
London, 1992.

[5] U. Goransson Model, based on cone calorimeter results, for

explaining the heat release rate growth of tests in a very large
room, Interflam93Proceedings of Sixth International Interflam Conference, Interscience, London, UK, 1993, pp. 3947
[6] ISO 5660-1: 2002(E), Reaction-to-Fire TestsHeat Release
Rate, Smoke Production and Mass Loss RatePart 1: Heat
Release Rate (Cone Calorimeter Method), second ed.,
International Organization for Standardization, Switzerland,
2002.
[7] G.J. Duggan, Usage of ISO 5660 data in UK railway standards
and fire safety cases, Paper 3, Fire Hazards, Testing, Materials
and Products, Proceedings of a One Day Conference, 13 March
1997, Rapra Technology Ltd, Shawburry, Shrewsbury, Shropshire, UK, 1997.
[8] W.K. Chow, Fire safety in train vehicle: design based on
accidental fire or arson fire? The Green Cross March/April
(2004) 7.
[9] V.P. Dowling, N. White, Fire sizes in railway passenger saloons,
in: E.S. Kim et al. (Ed.), Proceedings of the Sixth AsiaOceania
Symposium on Fire Science and Technology, 1720 March
2004, Daegu, Korea, Korean Institute of Fire Science and
Engineering, 2004, pp. 602611.
[10] F.W. Mowrer, B. Williamson, Methods to characterize heat
release rate data, Fire Safety J. 16 (1990) 367387.
[11] W.K. Chow, H.W. Au Yeung, On the superposition of heat
release rate for polymeric materials, Arch. Sci. Rev. 46 (2)
(2003) 145150.
[12] W.K. Chow, Assessment on heat release rate of furniture foam
arrangement by a cone calorimeter, J. Fire Sci. 20 (4) (2002)
319328.
[13] ASTM E 135404a, Standard Test Method for Heat and
Visible Smoke Release Rates for Materials and Products Using
an Oxygen Consumption Calorimeter, ASTM International,
USA, 2004.
[14] R. Vanspeybroeck, P. Van Hees, P. Vandevelde, Combustion
behaviour of polyurethane flexible foams under cone calorimeter test conditions, Fire Mater. 17 (1993) 155166.
[15] ISO 3534-1: 1993(E/F), StatisticsVocabulary and Symbols
Part 1: Probability and General Statistical Terms, first ed.,
International Organization for Standardization, Switzerland,
1993.
[16] ISO 9705: 1993(E), Fire TestsFull-scale Room Test for
Surface Products, International Standards for Organization,
Geneva, Switzerland, 1993.
[17] D.A. Smith, K. Shaw, The single burning item (SBI) test, the
Euro classes and transitional arrangement, Proceedings of
Interflam99, 29 June1 July, 1999, Edinburgh, UK,
Interscience Communications, London, UK, 1999, pp. 19.
[18] R.D. Peacock, P.A. Reneke, W.D. Davis, W. Jones, Quantifying
fire model evaluation using functional analysis, Fire Safety J. 33
(1999) 167184.
[19] P.A. Friday, F.W. Mowrer, Comparison of FDS Model
Predictions with FM/SNL Fire Test Data, NIST GCR 01-810,
National Institute of Standards and Technology, US Department
of Commerce, USA, 2001.
[20] N. White, V.P. Dowling, Conducting a full-scale experiment on
a railway passenger car, in: E.S. Kim et al. (Ed.), Proceedings of
the Sixth AsiaOceania Symposium on Fire Science and
Technology, 1720 March 2004, Daegu, Korea, Korean Institute
of Fire Science and Engineering, 2004, pp. 591601.
[21] W.K. Chow (Ed.), Special issue on full-scale burning tests,, Int.
J. Eng. Performance-Based Fire Codes 6 (3) (2004).