Fire Safety Journal 89 (2017) 16–21

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Total radiative heat loss and radiation distribution of liquid pool fire flames MARK
a b,⁎ a b
Liang Zhou , Dong Zeng , Dongyang Li , Marcos Chaos
a
Chinese People's Armed Police Academy, Department of Fire Protection Engineering, 220 Xichang Rd, Langfang, Hebei 065000, People's Republic of China
b
FM Global, Research Division, 1151 Boston-Providence Turnpike, Norwood, MA 02062, USA

A R T I C L E I N F O A BS T RAC T

Keywords: The radiative characteristics of laboratory-scale pool fire flames have been studied in detail. Experiments were
Pool fire conducted in the ASTM E2058/ISO 12136 Fire Propagation Apparatus (FPA). Eleven liquid fuels with different
Radiative power distribution sooting propensities, including alcohols and alkanes, burning in a 9.5 cm diameter quartz dish were considered.
Radiation fraction Radiative power distribution (along the flame axis) and global radiant emission were measured for all the fuels
Multiple-point source radiation model
by using slit and wide-view-angle radiometers, respectively. The effects of measurement location and fuel type
on the measured data were investigated. Radiation distribution profiles for a given fuel, when adequately
normalized, show little sensitivity to the horizontal separation distance of the slit radiometer. Fuels with similar
chemical structures exhibit similar distributions, consistent with flame image analyses. The radiative power
distributions along with the wide-view-angle radiometer data were used to derive radiant fractions for the pool
fires studied by applying a multiple-point source (MPS) radiation model. To examine the sensitivity of the
calculated radiant fractions to the measurement location, the position of the wide-view-angle radiometer was
considerably varied both vertically and horizontally. The results show that the radiant fractions derived based
on the measured radiative power distribution are independent of the location of the wide-view-angle radiometer
and consistent with literature values. Therefore, the approach developed in this study presents a flexible
methodology apt for the accurate determination of radiation properties of diffusion flames in a laboratory
setting.

1. Introduction
Radiation is the dominant heat transfer mode in fires of consider-
able scale [1] and it plays an important role in fire growth and spread
mechanisms. Understanding the radiative characteristics of fires
provides a fundamental basis for fire growth modeling [2]. One such
characteristic is the radiant fraction (χr), defined as the ratio of the
radiative power (Q̇ ̇ r) to the theoretical heat release rate (Q̇ ̇ ):
Q̇
χr = r
Q̇
where Q̇ ̇ can be expressed as the product of the mass burning rate, ṁ ,
and the net calorific value of the fuel, ΔH0cn. For jet flames, some
studies showed that radiant faction can be affected by the jet exit
velocity [3,4] and/or the flame residence time [5,6]. On the other hand,
it is generally accepted that for buoyant turbulent diffusion flames,
which are relevant to fires, the radiant fraction is fuel specific and
remains relatively constant over a range of combustion conditions [7–
12].
Since the total radiative power, Q̇ ̇ r, cannot be easily measured, a
common practice is to measure the incident radiation at one location
near the fire, and then derive the total radiative power using a model.
The accuracy of the derived radiant fraction is therefore affected by the
model. The most widely used approach, due to its simplicity, is the
single point source (SPS) model [13]. In this model, the flame is
approximated by a point usually located at the centroid of the flame
and the radiation from the point source is assumed to be isotropic.
Then, Q̇ ̇ r can be calculated from the following equation:
Qṙ τ
qg″̇ = cos ϕ
4πS 2 (2)
(1) where q̇ ″g is the heat flux measured by a gage (e.g., a radiometer) at a
distance S from the point source, ϕ is the angle formed by the normal to
the gage surface and the line of sight to the point source, and τ is the
atmospheric transmissivity over the distance S.
The nature of the SPS model makes it only applicable to measure-
ments performed in the far field of a given fire. Modak [13] showed that
the radiative power output (and, thus, χr) of pool fires derived from the
SPS model underestimated the true value by over 25% for measure-
ments taken within one pool radius of the fire; such estimates improved
considerably for measurements taken beyond ten pool radii. Hamins
et al. [14] found that the SPS model yielded radiant fractions accurate

⁎
Corresponding author.

http://dx.doi.org/10.1016/j.firesaf.2017.02.004
Received 26 April 2016; Received in revised form 1 December 2016; Accepted 13 February 2017
0379-7112/ © 2017 Elsevier Ltd. All rights reserved.
L. Zhou et al. Fire Safety Journal 89 (2017) 16–21

considered here. Therefore, rather than presuming a profile, slit

radiometer measurements of radiation distribution are part of the
present study.

2. Experimental

All the tests were conducted on steady pool fires established in an

un-cooled quartz dish of 9.5 cm inner diameter placed in a Fire
Propagation Apparatus (FPA) [21,22]. This apparatus provides a
calorimetry measurement in an environment with a controlled air co-
flow. Dry air was supplied to the system, using a mass flow controller at
a flow rate of 100 L/min, through a bed of glass beads to ensure flow
uniformity, yielding an upward co-flow of approximately 0.1 m/s. This
flow was used to stabilize and anchor the base of the flames and did not
alter their buoyant turbulent nature [23,24]. This assumption was
verified by varying the flow rate of air without appreciable change (less
than 5%) in measured burning rates and radiation profiles.

2.1. Fuel selection

Fig. 1. Geometry setup for the MPS radiation model.

Since sooting propensity is related to the radiant fraction of fuels

to within 13% for small to medium sized pool fires for measurement (e.g., [25]), eleven liquid fuels with different sooting propensities,
locations at separation distances between five and sixteen pool radii. including alcohols and alkanes, were selected in the present study. The
However, it should be noted that measurements at far field locations normalized smoke point (NSP) [26] and the net calorific value [27] of
may not be practical in a laboratory setting and can be affected by each fuel are listed in Table 1. All the fuels used were HPLC grade.
resolution issues, due to the marked decrease in heat flux with
increasing separation, as well as atmospheric attenuation effects [15].
2.2. Fuel supply and level control
Hankinson and Lowesmith [16] proposed a weighted multi-point
source (MPS) model to determine the radiant fraction of jet flames.
The adopted fuel supply system in the present study is similar to
This approach was shown to yield reasonably accurate radiant fractions
that used by Ditch et al. [24]. During the experiment, fuel was
based on both near-field and far-field measurements [16]. Recently,
continuously fed and controlled with two needle valves connected to
this model has been successfully applied to measure the radiant
a 250 ml burette, as shown in Fig. 2. Fuel dripped through the first
fraction of buoyant turbulent diffusion line flames [17]. Fig. 1 shows
needle valve into a burette, which was used to dampen the influence of
a schematic of the MPS approach, which has been adapted in this study
fuel elevation on the dripping velocity, and then through the second
to pool fires. The MPS methodology assumes that N point sources are
valve into a U-shaped tube connected to the bottom of the quartz dish.
evenly distributed along the flame axis and the radiation measured by a
Once steady-state conditions were achieved, the mass loss rate (ṁ ) was
gage can be calculated as the weighted sum of the radiation from each
calculated based on the average volume change measured by the
point source as follows:
burette over a given time interval which was usually on the order of
N 40 min. The uncertainty in ṁ is mainly attributed to the manual fuel
Qṙ τi
qg″̇ = ∑ wi cos ϕi level control; based on repeated tests, this uncertainty was within 5%
4πSi2 (3)
i =1 for all the used fuels.

where wi is the weight of the ith point source and Σwi=1, ϕi is the angle
between the normal to the gage surface and the line of sight to the ith 2.3. Vertical radiative power distribution measurement
point source, and τi is the transmissivity over the separation distance,
Si, from the ith point source to the gage. The weight profile of the point In order to quantify the weight of each point source for use in the
sources (shown schematically in Fig. 1) was approximated in [16] by a MPS model, the slit radiometer methodology developed by Markstein
scheme representing weights increasing linearly to a peak at some n th [20] was adopted. As shown in Fig. 2, a horizontal slit from two water-
point (where wn=nw1) and decreasing linearly thereafter so that
Table 1
wN=w1. It is noted that, for jet flames, the peak weight was positioned
Fuel properties.
at 0.75 Lf [16], where Lf is the flame height, based on measurements
made on large scale jet fires using a narrow angle radiometer reported Fuel NSPa (mm) ΔH0cn (kJ/g)b
by Cook et al. [4], and also the data of Sivathanu and Gore [18] as well
Methyl alcohol n.m.c 19.94
as Baillie et al. [19].
Ethyl alcohol n.m. 26.84
The present work applies the MPS model to derive radiant fractions n-Propyl alcohol n.m. 30.71
for pool fires burning several liquid fuels with special attention given to i-Propyl alcohol n.m. 30.47
the sensitivity of the model to measurement location. To use the MPS n-Butyl alcohol n.m. 33.12
model, the radiation distribution must be known so that weights can be n-Amyl alcohol n.m. 34.79
n-Hexane 149 ± 24 44.74
assigned (see Fig. 1). Considering the different fire dynamics of jet n-Heptane 139 ± 15 44.56
flames as compared to buoyant turbulent flames, the assumed dis- n-Octane 137 ± 28 44.42
tribution of jet flames [16] might not be applicable. The only available n-Decane 122 ± 14 44.24
measured radiation distribution of buoyancy-controlled turbulent Cyclohexane 82 ± 16 43.42
diffusion flames reported by Markstein [20] suggests that the distribu- a
Li and Sunderland [26].
tion is characterized by a slightly asymmetric bell-shaped curve with b
Values listed are for liquid state and calculated based on standard heat of formation
the maximum occurring at about 0.45 Lf. However, the gaseous from [27].
c
propane fuel used in his study [20] differs from the liquid fuels Not measurable.

17
L. Zhou et al. Fire Safety Journal 89 (2017) 16–21

Fig. 2. Modified FPA setup with fuel supply, control system, and slit radiometer.

cooled, insulated plates was installed on a water-cooled housing. A
water-cooled wide-view-angle Schmidt-Boelter radiometer with a range
of 0–2 kW/m2 (Medtherm Corp., model #64-0.2-20, serial #187061)
was installed inside the housing. Therefore, the radiometer could only
see part of the flame through the slit. The effective flame height (Δz) at
the flame axis viewed by the radiometer could be varied by changing
the slit opening and the position of the radiometer relative to the slit.
Various configurations were considered, all leading to similar results.
Ultimately, an arrangement providing Δz=4.8 cm was chosen, which
provided optimum measurement resolution. During all the tests, the
slit-radiometer was traversed vertically, at discrete intervals of 1.3 cm,
from the fuel surface to a height where the radiometer reading reduced
to near zero. At each position, readings were taken for 60 s and
averaged. With these measurements, the weight distribution could be
calculated with Eq. (4):
″
qsṙ , i
wi = N method described in ASTM E2058/ISO 12136 [21,22] using the CO
∑i =1 qsr″̇ , i
where q̇ ″sr, i is the measured heat flux by the slit radiometer at the ith
point, the Nth point is the highest point measured before the
measurement drops to near zero and is approximately located at the
peak visible flame height. Since the interval between the adjacent
points was constant at 1.3 cm, the total number of point sources, N,
varies for different fuels due to the variation of flame height. Therefore,
in order to compare the weight distributions of different fuels, Eq. (4)
may be modified as follows:
wnorm, i = Nwi
where wnorm, i can be thought of as a normalized weight.
1394b-M) at a frame rate of 10 Hz with variable exposure settings to
facilitate high contrast between flame and background. The instanta-
neous grey-scale flame image in each of the recorded frames was
converted to a binary black/white image with white pixels representing
the visible flame and black pixels corresponding to the absence of
flame. The processed images were averaged over the steady burning
period and converted to probability (i.e., intermittency) maps; the
average flame height (Lf) could then be determined by the 50%
intermittency location [28].
3. Results and discussion
For all the tested fuels, the steady-state mass loss rate (ṁ ), the
theoretical heat release rate (Q̇ ̇ ) calculated based on ṁ and ΔH0cn, as
well as the chemical heat release rate (Q̇ ̇ ch) calculated based on the
(4) and CO2 generation rates measured by FPA are listed in Table 2
together with the errors of ṁ based on three repeated tests. Both the
mass loss rate and generation rate of CO and CO2 are averaged values
during a period of steady burning.
3.1. Radiative power distribution
In order to better understand the radiation distribution, the
(5) influence of the horizontal distance between the slit radiometer and
the flame axis as well as the fuel type on the measured profiles were
studied in detail.

2.4. Global radiation measurements Table 2

Mass loss and heat release rates of the fires.
A 150° wide-view-angle Schmidt-Boelter radiometer with a range of
Fuel ṁ (g/s) Q̇ ̇ ch (kW) Q̇ ̇ (kW)
0–20 kW/m2 (Medtherm Corp., model #64-10SB-18, serial #154822)
was used to measure radiation received from the whole flame. This Methyl alcohol 0.069 ± 3.9% 1.37 1.38
radiometer was positioned with its surface parallel to the axis of the Ethyl alcohol 0.064 ± 4.3% 1.71 1.76
flame. In order to verify the applicability of the MPS model, the n-Propyl alcohol 0.062 ± 4.0% 1.89 2.03
i-Propyl alcohol 0.070 ± 4.5% 2.12 2.24
radiometer was mounted on a slide so that its location could be varied
n-Butyl alcohol 0.058 ± 4.0% 1.91 2.03
in both horizontal and vertical directions. Similar to the slit radiometer n-Amyl alcohol 0.054 ± 4.9% 1.89 2.03
measurements, readings were taken for 60 s and then averaged. n-Hexane 0.108 ± 5.9% 4.85 5.47
n-Heptane 0.082 ± 6.2% 3.66 4.11
n-Octane 0.062 ± 5.7% 2.75 3.04
2.5. Flame image analysis
n-Decane 0.057 ± 5.2% 2.52 2.86
Cyclohexane 0.099 ± 7.8% 4.28 4.99
Flame images were recorded using a CCD camera (ISG-LW-5-S-

18
L. Zhou et al.
Fig. 3. Measured radiative heat flux (upper panel) and normalized weight distribution
(lower panel) of n-hexane flames for different separation distances.
3.1.1. Effect of horizontal distance of slit radiometer measurement
In this section, n-hexane is used as an example; similar results were
obtained for all fuels considered. The radiation distribution was
measured with the slit radiometer separation set at 20.3 cm
(x1=17.8 cm; x2=2.5 cm), 25.4 cm (x1=22.2 cm; x2=3.2 cm), and
30.5 cm (x1=26.7 cm; x2=3.8 cm) while keeping Δz at 4.8 cm by
adjusting the slit opening height. Measured average radiation over
60 s at different heights along with the calculated weight distributions
are shown in Fig. 3. The error bars in Fig. 3a denote the standard
deviation of three repeated tests at each location. The errors in the heat
flux measurement are associated to the expanded uncertainty of the slit
radiometer, which was ± 3 and the uncertainty of the height of the slit
radiometer, which was 0.16. Fig. 3b plots the results of Eq. (5) as a
function of height normalized by the peak flame height. From the
results shown in Fig. 3b, it can be seen that with a change of the
horizontal distance between slit radiometer and flame axis, the relative
radiation weight distribution is virtually unchanged. In fact, for all the
fuels examined, the measured radiation distributions were highly
repeatable (to within 3%). It can be noticed that the standard deviation
of the signal increases as the sensor is closer to the flame. On the other
hand, the measurement cannot be too far from the flame due to the
decreased signal to noise ratio.
3.1.2. Effect of Fuel Type
In order to compare the radiation distribution of all the 11 fuels, a
series of burning tests were carried out with the slit radiometer
positioned at 25.4 cm (x1=22.2 cm; x2=3.2 cm) from the flame axis.
The measured radiation distributions are shown in Fig. 4. These liquid
pool fire radiation distributions are qualitatively different from those of
jet flames [4,18] and buoyancy-controlled turbulent gas diffusion
19
Fire Safety Journal 89 (2017) 16–21
Fig. 4. Normalized radiation weight distribution of different fuels.
flames [20]. As described previously, for jet flames the distribution
peaks at about z=0.75 Lf. For the buoyancy-controlled turbulent
propane diffusion flames, the distribution is slightly asymmetric with
the maximum occurring at about z=0.45 Lf.
For the liquid pool fires studied in the current work, the radiation
distribution varies from fuel to fuel. At the same time, self-similar
distributions are observed for fuels with similar chemical structures
(see Fig. 4). As shown in Fig. 4a, methyl alcohol and ethyl alcohol have
only one peak in the radiation distribution and the peak is located close
to the fuel surface. All other alcohols, with larger carbon chains, have
two peaks of similar size (Fig. 4b). On the other hand, Fig. 4c shows
that all the alkanes have radiation distributions that exhibit two peaks:
a small peak located near the fuel surface and a larger peak located at
about 54–66% of Lf or 40–50% of the peak flame height.
According to test observations, one possible cause for the difference
and similarity in radiation distribution is the flame length and shape;
i.e., similar fuels usually have similar flame length and shape, and
similar distributions. Flame images of all the fuels were analyzed to
further demonstrate this point. Fig. 5 shows flame images and flame
height, Lf, averaged over a period of 1000 s for all the fuels except
L. Zhou et al. Fire Safety Journal 89 (2017) 16–21

Fig. 5. Average flame images and flame heights. The green, yellow, and blue dashed lines indicate the height of the first, second peak, and the location of the lowest radiant intensity
between peaks, respectively (see Fig. 4).

methanol, whose weak luminosity was not properly captured by the using the measured radiant power distributions described above along
camera used. The color scale in Fig. 5 indicates the probability of flame with wide-angle radiometer measurements. Since these latter measure-
appearance at a given location. Ethanol has the smallest flame height ments were acquired within a maximum separation distance, xg, of
with no apparent “necking-in” structure in the flame shape; the 0.46 m between the radiometer and the dish center line, atmospheric
remaining alcohols have similar flame height and shape with some radiation absorption was neglected and τj=1.
degree of necking-in. Similarly, all the alkanes have comparable flame The effect of varying the position of the wide angle radiometer on
shapes with noticeable necking-in structures. It can also be noticed that the calculated radiant fraction is shown in Table 3, using n-heptane as
all the fuels with necking-in structures have two peaks in their an example. The horizontal separation distance of the wide-angle
corresponding radiation profile (see Fig. 4). Flame necking-in close radiometer was varied between 0.23 and 0.46 m (~5–10 pool radii),
to the fuel surface is a fundamental behavior for buoyant diffusion whereas the vertical distance to the fuel surface varied from 0.08 to
flames caused by the entrained flow induced by buoyancy, which is the 0.38 m. The average radiant fraction calculated considering all the
major cause of flame periodic instability [29] and this instability measurement locations is 0.31. From Table 3, it is noted that the
governs the local volumetric heat release rate [30]. As shown in maximum deviation from this mean value is 4.6%, which shows the
Fig. 5, the necking-in structure separates the peaks observed in the MPS model provides consistent results at the different near-field
radiation distribution profiles. The chosen 1.3 cm vertical traverse
interval of the slit radiometer was sufficient to resolve the radiation
Table 3
profile and fully capture the first radiation peak near the fuel surface.
Radiant fraction of n-heptane calculated by the MPS model for different measurement
The second peak is associated with the increased flame volume and locations (numbers in parentheses denote the absolute deviation from the average value).
flame temperature as buoyant structures rise from the fuel surface. In
the case of alkanes, which generate more soot than the alcohols [31], zg (m) xg (m)
most of the flame luminosity occurs above the necking-in structure,
0.46 0.38 0.30 0.23
leading to the larger second peak in Fig. 4. Other factors such as gas-
phase radiation may also contribute to the shape of radiative profiles. 0.38 0.30 0.30 0.32 0.31
(2.0%) (1.7%) (2.5%) (0.3%)
0.30 0.31 (1.3%) 0.31 (0.2%) 0.32 (2.2%) 0.31 (1.6%)
0.23 0.31 (1.6%) 0.30 (1.9%) 0.30 (2.9%) 0.30 (1.8%)
3.2. Radiant fraction
0.15 0.32 (4.6%) 0.32 (2.1%) 0.30 (2.4%) 0.31 (0.4%)
0.08 0.32 (2.6%) 0.31 (1.5%) 0.32 (2.5%) 0.31 (1.0%)
The MPS model, Eq. (3), was used to calculate the radiant fraction

20
L. Zhou et al. Fire Safety Journal 89 (2017) 16–21

Table 4 point source (MPS) model based on the measured radiative power
Comparison with literature values. distributions and the total radiation measurement provided by a wide-
view-angle radiometer. Even though the MPS model was originally
χr
developed for jet flames, the present study has demonstrated its
Fuel N Present Literature applicability to buoyant turbulent flames. The radiant fractions calcu-
lated by the MPS model are consistent with literature values and have
Methyl alcohol 24 0.19 ± 0.02 0.20a
been shown to be insensitive to the location of the wide-view-angle
0.18b
0.15c radiometer. Therefore, the measurement methodology described in this
work presents researchers with an approach that is flexible and does
Ethyl alcohol 26 0.23 ± 0.02 0.20d not suffer from the inaccuracies of other single-location measurements.
0.17a
0.23c
Acknowledgements
n-Propyl alcohol 29 0.27 ± 0.02 0.27c
i-Propyl alcohol 32 0.30 ± 0.03 0.27c The authors foremost acknowledge the financial support and
n-Butyl alcohol 33 0.29 ± 0.01 0.28c resources provided by FM Global and would like to thank Mr.
n-Amyl alcohol 34 0.29 ± 0.03 0.29c
Stephen Ogden for carrying out all the tests; Mr. Steve D′Aniello and
n-Hexane 48 0.32 ± 0.02 0.30c
n-Heptane 48 0.31 ± 0.02 0.32d Mr. Robert Tabinowski for assistance in the building of the test facility;
0.29a Drs. Yi Wang, Francesco Tamanini, John de Ris, and Mr. Mohammed
0.31b Khan for valuable discussions and advice.
0.30c

n-Octane 46 0.32 ± 0.03 0.31c
n-Decane 36 0.31 ± 0.02 0.31c
Cyclohexane 50 0.34 ± 0.03 0.35c
a
7.1 cm diameter, constant level [14].
b
30 cm diameter, constant level [14].
c [4] D.K. Cook, M. Fairweather, J. Hammonds, D.J. Hughes, Trans. IchemE Part A 65
10 cm diameter without constant level [31] based on convective heat release rate
measurements.
d [5] W. Houf, R. Schefer, Int. J. Hydrog. Energy 32 (2007) 136–151.
4.6 cm diameter, constant level [14].
locations considered. A similar trend was also found for the other
studied fuels with maximum deviations always less than 8%.
The sensitivity of the measured χr to the number of point sources,
N, was explored by varying the vertical traverse resolution of the slit
radiometer. Decreasing this resolution below 1.3 cm did not change the
value of the calculated radiant fraction for all fuels. The average radiant
fractions based on the MPS model and measured radiant power
distributions are presented in Table 4 together with the value of N
used for each flame. Compared to literature data (see Table 4), the
average radiant fractions based on the MPS model and measured
radiant power distributions are consistent with reported values derived
from multi-location radiation measurements of pool fires with fuel level
control [14]. A similar consistency is evident between the current data
and those based on convective heat release rate measurements [31].
4. Conclusions
A slit radiometer methodology was implemented to measure the
radiation emission profiles along the flame height for pool fires of 11
different liquids. The radiation profiles of these fires were found to be
very different from those of jet flames and buoyancy-driven turbulent
gas diffusion flames. The measured profiles also showed noticeable
variation among fuels. However, similarities were found for fuels with
similar chemical structure, which have similar flame height and shape.
It was also verified that variations in measurement location have no
significant impact on measured radiation distributions. Image proces-
sing of these flames confirmed that the radiant power distributions
correlate closely with averaged flame shapes and, in particular, with
necking-in structures.
The radiant fraction for each fuel was determined using a multi-
References
[1] J. de Ris, Proc. Combust. Inst. 17 (1979) 1003–1016.
[2] P. Chatterjee, J.L. de Ris, Y. Wang, S.B. Dorofeev, Proc. Combust. Inst. 33 (2011)
2665–2671.
[3] G.A. Chamberlain, Chem. Eng. Res. Des. 65 (1987) 299–309.
(1987) 318–325.
[6] S.R. Turns, F.H. Myhr, Combust. Flame 87 (1991) 319–335.
[7] G.R. Kent, Hydrocarb. Process 43 (1963) 121–125.
[8] S.H. Tan, Hydrocarb. Process 46 (1967) 172–176.
[9] B.J. Lowesmith, G. Hankinson, M.R. Acton, G. Chamberlain, Process Saf. Environ.
Prot. 85 (2007) 207–220.
[10] J.C. Yang, A. Hamins, T. Kashiwagi, Combust. Sci. Technol. 96 (1994) 183–188.
[11] G.H. Markstein, Combust. Flame 27 (1976) 51–63.
[12] G.H. Markstein, J.L. de Ris, Proc. Combust. Inst. 20 (1985) 1637–1646.
[13] A.T. Modak, Combust. Flame 29 (1977) 177–192.
[14] A. Hamins, M. Klassen, J. Gore, T. Kashiwagi, Combust. Flame 86 (1991) 223–228.
[15] S. Fuss, A. Hamins, Fire Saf. J. 37 (2002) 181–190.
[16] G. Hankinson, B.J. Lowesmith, Combust. Flame 159 (2012) 1165–1177.
[17] J.P. White, E.D. Link, A.C. Trouvé, P.B. Sunderland, A.W. Marshall, J.A. Sheffel,
M.L. Corn, M.B. Colket, M. Chaos, H.-Z. Yu, Fire Saf. J. 76 (2015) 74–84.
[18] Y.R. Sivathanu, J.P. Gore, Combust. Flame 94 (1993) 265–270.
[19] S. Baillie, M. Caulfield, D.K. Cook, P. Docherty, Process Saf. Environ. Prot. 76 (1)
(1998) 3–13.
[20] G.H. Markstein, Proc. Combust. Inst. 16 (1977) 1407–1419.
[21] ASTM E2058-13a, Standard Test Methods for Measurement of Material
Flammability Using a Fire Propagation Apparatus (FPA), ASTM International,
West Conshohocken, PA, 2013.
[22] ISO 12136, Reaction to Fire tests – Measurement of Material Properties Using a
Fire Propagation Apparatus, International Organization for Standardization,
Geneva, Switzerland, 2011.
[23] H.C. Hottel, Fire Res Abstr, Rev 1 (1959) 41–44.
[24] B.D. Ditch, J.L. de Ris, T.K. Blanchat, M. Chaos, R.G. Bill Jr, S.B. Dorofeev,
Combust. Flame 160 (2013) 2964–2974.
[25] G.H. Markstein, Proc. Combust. Inst. 20 (1985) 1055–1061.
[26] L. Li, P.B. Sunderland, Combust. Sci. Technol. 184 (6) (2012) 829–841.
[27] P.J. Linstrom, W.G. Mallard (Eds.), NIST Chemistry WebBook, NIST Standard
Reference Database Number 69, National Institute of Standards and Technology,
Gaithersburg, MD, available at: 〈http://webbook.nist.gov/chemistry/〉.
[28] E.E. Zukoski, B.M. Cetegen, T. Kubota, Proc. Combust. Inst. 20 (1984) 361–366.
[29] L.H. Hu, J.J. Hu, J.L. de Ris, Combust. Flame 162 (2015) 1095–1103.
[30] J.L. de Ris, Procedia Eng. 62 (2013) 13–27.
[31] M.M. Khan, A. Tewarson, M. Chaos, M.J. Hurley, D.T. Gottuk, J.R. Hall, Jr,
K. Harada, E.D. Kuligowski, M. Puchovsky, J.L. Torero, J.M. Watts, Jr,
C.J. Wieczorek (Eds.), SFPE Handbook of Fire Protection Engineering5th ed.,
Springer, New York, 2016, pp. 1143–1232.

21