You are on page 1of 13

See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/256687835

The effect of pressure and oxygen


concentration on the combustion of PMMA
ARTICLE in COMBUSTION AND FLAME AUGUST 2013
Impact Factor: 3.08 DOI: 10.1016/j.combustflame.2013.02.019

CITATIONS

5 AUTHORS, INCLUDING:
James G. Quintiere

Richard E. Lyon

University of Maryland, College Park

Federal Aviation Administration

127 PUBLICATIONS 1,976 CITATIONS

114 PUBLICATIONS 1,842 CITATIONS

SEE PROFILE

SEE PROFILE

F. J. Diez
Rutgers, The State University of New Jersey
56 PUBLICATIONS 132 CITATIONS
SEE PROFILE

Available from: James G. Quintiere


Retrieved on: 12 October 2015

Combustion and Flame 160 (2013) 15191530

Contents lists available at SciVerse ScienceDirect

Combustion and Flame


j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e

The effect of pressure and oxygen concentration on the combustion


of PMMA
Mariusz Zarzecki a, James G. Quintiere b, Richard E. Lyon c, Tobias Rossmann a, Francisco J. Diez a,
a

Department of Mechanical Engineering, Rutgers University, Piscataway, NJ 08854, USA


Department of Fire Protection Engineering, University of Maryland, College Park, MD 20742, USA
c
Fire Safety Branch, Federal Aviation Administration, William J. Hughes Technical Center, Atlantic City International Airport, NJ 08405, USA
b

a r t i c l e

i n f o

Article history:
Received 30 May 2012
Received in revised form 25 October 2012
Accepted 12 February 2013
Available online 28 March 2013
Keywords:
Combustion of polymers
PMMA
Low pressure
Burning rate
Time to ignition
Flammability

a b s t r a c t
An experimental study of the ammability properties of PMMA at low pressures and oxygen concentrations was performed. The work was motivated by the importance of these effects on re safety in the aviation industry. Measurements were obtained in a mass loss calorimeter inside a large 10 m3 pressure
vessel capable of reaching pressures as low as 0.1 atm. The PMMA ammability was characterized by
measuring the burning rate and the time to ignition of small test samples. These were ignited and burned
under different external heat uxes, total pressures and oxygen concentrations. The combined effects of
pressure and oxygen concentration on the burning rate, combustion ow eld, and ignition were evaluated. Results showed that at low pressure, the burning rate was less intense with a decrease in the mass
loss rate. However, the reduction in pressure caused a shortened ignition delay time. Experimental measurements were compared with a simple analytical model showing good agreement. The results also
show how pressure and oxygen concentration contributed to the heat transfer from the ame. The model
revealed that a single function in oxygen and pressure could account for both ame radiative and convective effects. As a result, a power law t was obtained for the relation of the combined pressure and oxygen
effect on the burning rate. This correlation shows a good agreement with the measurements and predicts
the burning rate behavior for the full range of pressure and oxygen tests.
2013 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction
Pressure and buoyancy effects on the ammability and burning
characteristics of solid materials have been previously extensively
explored in low gravity and micro-gravity environments. These results have been directly applied to re prevention and re-ghting
strategies on spacecraft systems [1,2]. However, res can also occur in low pressure environments associated with high altitude
locations and inside sub-ambient pressurized compartments.
Burning characteristics and rates at high altitude locations (local
pressure is up to 35% less than sea level) and the pressurized cabin
of a cruising aircraft (at 10 km altitude the local ambient pressure
is 0.26 atm) play an important role in characterizing the ammability of solid materials. Similarly, the role of local oxygen concentration is extremely important to the burning of solid materials
due to their strong dependence on oxygen concentration at the
ame zone [3]. Since the burning of these solids occurs in a diffusion ame mode, oxygen transport is often one of the major factors
inuencing ignition criterion and the sustainment of combustion
[4,5]. Therefore, pressure and local oxygen concentration are both
expected to signicantly inuence the ammability of solid
Corresponding author.
E-mail address: diez@jove.rutgers.edu (F.J. Diez).

materials. For practical application, depressurization and nitrogen


enrichment are both possible re-combating procedures for cargo
planes cruising a high altitude. Although the combined effects of
pressure and local oxygen concentration on solid material ammability are not well understood, reduced burning rates under certain
depressurization conditions could allow for more time for the efcacy of other re-ghting techniques or to allow for safer landing
procedures.
Prior studies in reduced-pressure ammability of solid materials have focused on ignition criterion and burning rate. Experiments at high altitude conditions have shown a reduction in
ignition delay time of solid materials (wood) when compared to
similar experiments at sea level [6]. In addition, studies performed
in laboratory low-pressure environments have shown a tendency
for a sample to ignite earlier and at lower ignition energies and
lower pressure. Ignition studies using piloted ignition at reduced
pressure, performed by McAllister et al. [4] and Wang et al. [6],
showed a decrease in time to ignition with decreasing pressure.
It was observed that as the pressure is reduced, the ignition time
decreased, reached a minimum, and then increased until ignition
did not occur [4], suggesting a competition between transport
and chemical rates.
Another factor inuencing the ignition and burning rate of
solid combustibles is the interaction between the ame heat ux

0010-2180/$ - see front matter 2013 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.combustame.2013.02.019

1520

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

Nomenclature
A
Cp
CHF
D
Gr
hc
k
l
lm,e
L
Lf
LFL
_ 00ig
m
Nu
p
q_ 00e
q_ 00ext
q_ 00f ;c
q_ 00f ;r
q_ 00py
q_ 00rr
r
s

bounding surface of the gas


specic heat
critical heat ux
diameter
Grashof number
convective heat transfer coefcient
conductivity material burnt
characteristic length
mean beam length
heat of gasication
ame height
lower ammability limit concentration
mass ux at ignition
Nusselt number
pressure
incident external radiative ux from the heater
applied external heat ux
convective heat ux
incident ame radiative heat ux to the material
pyrolysis energy ux
re-radiation heat ux from the surface
stoichiometric fuel to oxygen mass ratio
length of one side of square sample

(radiation and convection) with the ame zone. Tewarson [3]


showed an increase in ame heat ux with a corresponding increase in pressure and oxygen concentration above ambient. Wang
et al. [6] on the other hand showed that the mass loss rate of wood
at high altitude (Lhasa, 3650 m) was greater than that at low altitude (Hefei, 50 m). In addition, the ame shape can be highly inuenced by the reduction in buoyancy for reduced pressure ames as
has been shown previously in microgravity environments [7,8] and
low pressure environments [9,10]. This alteration of the ame
shape can strongly affect both heat loss through convection and
heat feedback to the ame zone through radiation.
Another important parameter that characterizes material ammability is ame spread. Most of the research on ammability of
combustible solids in reduced pressure environments has involved
ame spread measurements [1115]. The results show that at reduced pressure the ame spread decreases, with the exception of
insulated wires, where the effects were negligible [13]. NASA
researchers studied the effects of velocity on the extinguishment
of PMMA samples, in a low pressure and low gravity environment
[15] and concluded that the rate of depressurization would affect
the ame spread and burning rate of the sample. Again, coupling
between the buoyancy induced ow and the kinetics gave rise to
a non-monotonic dependence of the burning rate with pressure.
Recently, NASA has developed an equivalent low stretch apparatus
(ELSA), which can be used to study the effects of buoyancy on convective cooling by varying the ame stretch rate [16] to examine
the effect of local quenching rates.
Soot radiation also plays an important role in the complex
feedback heat transfer of the ame. The production of soot in a
ame contributes to an increase in ame heat ux through the
re-radiation of energy from the soot particles back to the burning
sample. Studies done at pressures above atmospheric suggest an
increase in soot formation with an increase in total pressure
[6,17,18]. The studies also suggest that below atmospheric
pressure the ame might produce less soot, causing the radiative
component of the ame heat ux to decrease.
Prior studies on solid combustible ammability in low-pressure
environments have previously been limited to ignition delay and

Tf
Tf,crit
tig
Tv
T1
TRP
V
Xr
Y O2 ;1

ame temperature
adiabatic ame temperature at LFL
ignition time delay
material surface vaporization temperature
ambient temperature
thermal response parameter
volume of the gas
ame radiative fraction
ambient oxygen mass fraction

Greek symbols
Dhc
heat of combustion
Dhpy
energy per unit mass loss to vaporize the solid
ef
ame emissivity
jf
absorption emission coefcient
l
viscosity
q
density material burnt
r
StefanBoltzmann constant
s
transmissivity of the ame between the heater and the
material

ame spread experiments. Limited data is available for mass loss


measurements under steady burning in reduced pressure or reduced oxygen concentration environments. Prior experimental
work using mass loss calorimeters has shown good correlation between steady burning mass loss rates with piloted ignition and
ame-spread data [19,20]. In addition, external heat ux has also
been extensively used as a combustion promoter for solid materials to enhance gasication and ignition [21,22]. The purpose of the
current work is to experimentally obtain mass loss measurements
during the burning of a solid combustible (PMMA) at low pressures
and oxygen concentrations using a mass loss calorimeter. The data
are then correlated with a theoretical analysis to model the behavior of the res in the experiments. This coupling of the analysis
with the experimental data elucidates where the pressure and oxygen concentration effects contribute most strongly to the variation
of ignition and steady burning rates of the sample.

2. Experimental setup and procedure


All the tests were performed in the pressure modeling facility at
the FAA tech center (Atlantic City, NJ). This facility can simulate the
pressure observed at different altitudes up to 14,600 m. The facility
includes a 10 m3 pressure vessel, capable of reaching pressures of
0.1 atm, a data feedthrough system for both temperature and
imaging, and rapid internal access through a sealed end-cap door.
The internal pressure is monitored using a pressure transducer,
connected to a control module that can indenitely maintain a
constant pressure within 0.007 atm for a given test. The control
module uses a three-way proportional value which allows air to
be removed by a vacuum pump during the tests to maintain a constant system background pressure. Further, a constant air ow rate
through the vessel of 7 0.5 l/s (lps) was maintained for all the
tests to provide an independent air source for the combustion
experiments. Therefore, the environment in the test chamber could
be independently controlled through the gasses supplied to the
mass loss calorimeter and the externally applied pressure set in
the pressure vessel.

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

1521

Fig. 1. (a) Front and (b) side view of the calorimeter and enclosure (used to control the amount of oxygen). Air is fed in through the bottom of the enclosure and exhausted out
through the top.

To closely control the amount of oxygen available to the polymer sampling during the burning, all the measurements are performed inside a smaller container placed inside the pressure
vessel. This container shown in Fig. 1 is 76.2 cm tall with a crosssectional area 62.5  62.5 cm2. Through the use of the externally
applied constant air ow, the oxygen concentration can be varied
locally inside the container, without the need to ll the entire pressure vessel. The airow enters through the bottom of the container
via a 12.7 mm diameter pipe perforated every 1 cm to create a uniform updraft. The air exits the container through the top via a
7.62 cm exhaust port pipe, expelling the exhaust gas into the inside
of the pressure vessel. To avoid exhaust gas air recycling, a very
small positive pressure difference exists between the smaller
chamber and the pressure vessel during the tests to shield the test
environment from the conditions in the larger pressure vessel. The
supplied air, after it has passed through the smaller container, is
then removed via a vacuum pump to the outside. The air entering
the internal container is maintained at a specied oxygen concentration, Y O2 1 , that ranged from 12% to 21% depending on the particular tests as shown in Table 1. The required oxygen concentration
for each test is obtained by mixing the house air with nitrogen to
produce the required gas mixture. The system is capable of reliably
producing and maintaining oxygen concentrations in nitrogen
from 21% to 10% by volume. The diluent nitrogen was produced

Table 1
List of test conditions.
Test conditions
Pressure (atm)
Oxygen concentration (%)
External heat ux (kW/m2)
Inlet air temperature (C)
Air ow rate through pressure vessel (l/s)
Polymer sample thickness (mm)

Values

0.18, 0.37, 0.46, 0.68, 0.82, 1


12, 14, 16, 18, 21
10, 12, 16, 25, 50, 72
25
7
6 (10, 12, 16, 25 kW/m2)
24.4 (50 kW/m2)
28.58 (72 kW/m2)
Polymer sample area (m2)
0.01
Distance between heater and sample surface (cm) 6
Distance between igniter and sample surface (cm) 1.3

using a Floxal nitrogen generation system manufactured by Air


Liquide.
Several time-resolved measurements were made within the
smaller chamber to characterize the burning of PMMA. The most
important of these measurements made inside the pressure vessel
is performed with a mass loss calorimeter (GovMark cone calorimeter meeting ASTM E 1354 standards) [34]. This particular cone
calorimeter offers a high-accuracy load cell with 0.01 g resolution,
5 kW Inconel heater, spark igniter, and automated heat ux settings. The mass loss calorimeter was placed inside the container
and allows for the measurements of the instantaneous sample
weight during combustion in a controlled (xed pressure and oxygen concentration) environment. A cone heater in the calorimeter
is used to provide uniform radiant heat to the polymer sample surface. Considering that the heater takes some time to reach steady
state, a shutter built into the calorimeter between the sample
and the heater reduces the heat transfer to the sample before the
measurements through radiative shielding, allowing for a better
determination of initial test conditions. To further reduce this
pre-test radiative heat transfer for these experiments, the shutter
was moved one centimeter closer to the sample, and the sample
was held 6 cm below the cone heater. The calorimeter shutter
was operated remotely via a hydraulic actuator. The same actuator
was used to move the igniter into position at the beginning of each
combustion test.
The polymer material used for the burning test is optically clear
poly(methyl methacrylate) (PMMA) [Plexiglass-G, Modern Plastics]. The sample preparation included placing the PMMA into a
steel sample holder where the back-side of the sample is insulated
using a high-temperature thermal ceramic insulator (Kaowool
board, 50 mm). The samples were 10  10 cm2, with a 6 mm thickness. The samples were further wrapped in aluminum to prevent
any spillage of the melted PMMA pool, which would alter the
boundary conditions associated with sample heating. Initially,
the sample was placed on the load cell with the heater was turned
off. The pressure vessel was then closed and equilibrated to the set
pressure. At this point, the cone heater was then brought to the desired heat ux condition with the shutter closed. When a test is initiated, the shutter to the cone heater is opened, radiating the
appropriate surface heat ux onto the sample, simultaneously with
the locating of the igniter above the sample centerline. This

1522

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

synchronization of events allowed for accurate determination of


the sample ignition delay times. The recorded ignition times were
shown not to be very sensitive to igniter position through varying
the igniter position at several pressures.
Tests were performed at different applied external heat ux values, q_ 00ext , ranging from 10 to 72 kW/m2 as shown in Table 1. For
higher heat uxes, steady burning was not achieved for a 6 mm
sample thickness. In those cases, steady burning was obtained by
increasing the sample thickness to 25.4 mm for 50 kW/m2, and
31.8 mm for 72 kW/m2. Once the sample was placed on the load
cell, the system was inaccessible for the duration of the test. All
tests were monitored via a closed circuit TV system, and recorded
with a commercial FS100 Canon camera. The experimental setup is
shown in Fig. 1. For oweld characterization, both the still camera
(providing still color images) and the video system (providing
information about ow-eld instability and transients) proved extremely useful in delivering visual information which assisted in
the interpretation of the combustion oweld.
For each burn test condition discussed in this paper, the experimental condition was repeated six times in the mass loss calorimeter with identical system pressures and ambient oxygen
concentrations. It was determined that the repeated testing of a

single set of initial conditions resulted in a variation for the mass


ux measurements of 1 g/m2 s, and a variation for the time to
ignition of 30 s. The test conditions included external heat uxes
from 1072 kW/m2, oxygen concentrations from 1221%, and
pressures from 0.18 to 1 atm. A summary of test conditions is given
in Table 1.

3. Experimental results
Video recordings provided direct observation of the ame shape
and color during the PMMA ammability study. Some typical
images of burning PMMA samples are shown in Fig. 2 where no
external heat ux was applied and the oxidizer mole fraction
was maintained at 21% for the various atmospheric pressure conditions. The images show substantial differences in ame height,
ame color, and ame shape as a function of external pressure.
In terms of luminescence from the ame, the ame color changed
from a bright yellow at standard pressure to a dim blue color at
low pressures with an associated reduction in the ame luminosity. This change in ame color and overall luminous intensity is related to both the amount of soot generated from the PMMA

Fig. 2. Typical images of burning PMMA samples for 21% O2 with no external heat ux applied showing ames at (a) 1 atm, (b) 0.68 atm, (c) 0.47 atm and (d) 0.18 atm. The
images show substantial differences in ame height, ame color, and ame shape as a function of external pressure. (For interpretation of the references to color in this gure
legend, the reader is referred to the web version of this article.)

Fig. 3. Typical images of burning PMMA samples for 21% O2 and a large external heat ux applied of 50 kW/m2 showing ames at (a) 0.48 atm and (b) 0.18 atm.

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

oxidation and the features of the buoyancy induced oweld,


which changes as the overall gas density varies with pressure.
The images in Fig. 2 also show a clear decrease in ame length
with a decrease in pressure. This is primarily due to the reduced
mass ux in lower pressure environments (due to the lower oxidizer density) such that the ame can only be sustained close to
the surface. Since the ame evolution is a balance between the
mass transfer from the PMMA surface and the heat transfer from
the ame, as the mass ux is reduced due to lower densities, the
mass loss rate (and thus the ame length) is similarly affected.
The effects of a reduced pressure environment on the ame
can be modulated by exposing the sample to large external heat
uxes (effectively increasing the ame heat transfer back to the
PMMA, enhancing the overall mass loss from the surface). When
the sample was exposed to a large external heat ux, such as
50 kW/m2 in Fig. 3, the mass ux from the surface was markedly
increased. As this ux is increased, the aspects of the higher pressure ame (yellowish ame color, larger ame height, and buoyancy dominated ow) can be recovered at lower pressure
conditions. This effect can also be seen down to the lowest pressure used in this study (comparing Figs. 2d to 3b) where a weak,
blue triangular ame at low pressure is enhanced to a higher
luminosity, larger, and more buoyant ame by the application
of the large external heat ux. Figure 3 shows that ames with
applied external heat uxes are much wider and sootier than
those with no external heat ux, especially evident at low
pressures.
Using the pressure vessel system to control the external pressure and oxygen mole fraction, a typical test run involves measuring the mass loss rate from the burning sample with the mass loss
calorimeter as a function of time. Typical results are shown in
Fig. 4. The raw mass loss data was smoothed using a linear 21point SavitzkyGolay lter to reduce measurement and readout
noise. During the test runs, the mass loss rate can be classied into
three regions: the rst region is where the re is growing (as
shown by the increase in the mass loss rate); the second region
shows steady burning as the mass loss rate in constant in time;
the third region is where the sample is being consumed, and there
is a loss of available reactants to the ame zone, the ame decays
(as shown by the decrease in the mass loss). The critical mass ux
is obtained from an average of 20 unltered measurement points
corresponding to 20 s of data, just before the sample ignited. Also,
a steady burning region was reached for all test conditions. Early

Fig. 4. Typical measured mass loss rate as a function of time from the PMMA
burning sample for 0.68 atm and 21% O2 at 16 kW/m2 external heat ux. Results are
smoothed using a linear 21 point SavitzkyGolay lter to reduced measurement
and readout noise. Steady state is observed between t  400 s and t  700 s.

1523

Fig. 5. Ignition times normalized and plotted versus the external heat ux for 1 atm
and 21% O2. The experimental results are tted by the theoretical model in Eq. (4)
(CHF = 8.2 kW/m2).

results showed that for the PMMA sample thicknesses used, steady
burning could not be achieved, due to the very high mass loss rates
and limited sample sizes. Thus, for the larger external heat uxes,
thicker PMMA samples were used to ensure that steady burning
was achieved during the test time.
Using the mass loss rate curves for various environments, several key combustion parameters can be extracted. The time to ignition (or ignition delay time) is a relevant parameter for
determining the ammability of the sample. This parameter can
be considered a sum of three distinct steps that help contribute
to the delay of ignition. Initially the solid has to be heated to a high
enough temperature to cause the evolution of pyrolysis gases to be
given off. The second step involves the transport of the fuel
through the boundary layer and the mixing of the fuel with the oxidizer. The last step involves the time it takes for the mixture to sustain propagation from the electric arc pilot. Since the main
objectives of the study are to determine the effect of low-pressure
environments and normal to low oxygen concentrations, ignition
times were determined from the greatest change in mass loss rate
correlated with the observance of ame by recorded video data.
Once the individual ignition times were achieved, they were normalized and collapsed by plotting them versus the external heat
ux as shown in Fig. 5 for ambient pressure conditions. The time
1=2
to ignition should scale as  tig with increasing external heat ux
[25]. This decrease in ignition delay time is expected as an increase
in external heat ux enhanced the mass transfer at the PMMA surface, leading to quicker production of ammable mixtures at the
ignition location.
Time to ignition values were obtained experimentally for all the
test conditions in Table 1 (range of pressures, external heat uxes,
and oxygen concentrations). These results are shown as a function
of pressure in Fig. 6 for the ve external heat ux conditions imposed in this study. Each test condition was repeated six times to
reduce the experimental uncertainty with both the average value
and the standard deviation shown in Fig. 6. The results show a
weak dependence of time to ignition with pressure at the lower
range of the externally applied heat ux conditions. This weak
dependence then disappears as the heat ux is further increased,
and the time to ignition is essentially independent of pressure
above an external heat ux of 16 kW/m2. Due to the nonlinear effect of the external heat ux on the ignition delay time, the error
bars on the higher heat ux conditions are much wider, while
the repeated lower heat ux tests showed a much smaller variation
in ignition times. These experiments suggest that the effect of pressure on time to ignition is negligible at sufciently high heat uxes
and hence can be neglected.

1524

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

Fig. 6. Experimentally obtained time to ignition values as a function of pressure for


test conditions in Table 1 at 21% oxygen concentration. The analytical model , Eqs.
(4) and (12), shows a good t at higher heat uxes and can be used to explain the
experimental measurements.

Fig. 7. Experimentally measured mass loss rate as a function of external heat ux at


21% oxygen concentration for characterization of the pressure effect on steady
burning. Results are compared to the analytical model in Eq. (13) to help explain the
measurements.

For characterization of the pressure effects on steady burning,


the experimentally measured mass loss rate can be plotted
against the external heat ux applied as shown in Fig. 7. The results show the expected scaling that the surface mass ux is directly dependent on the external heat ux. The externally
applied heat ux increases the surface heat transfer rate, increasing the volatilization rate of the fuel and counteracting the radiative and convective heat losses from the ame oweld. The
measurements at low external heat ux up to about 16 kW/m2
have a stronger dependence on pressure than at higher external
heat uxes where the experimental data show little effect of
the pressure in the chamber. This may be explained by the competition between the heat losses and ame heat feedback mechanism at the polymer surface. At higher external heat uxes, the
heat transport to the surface is in excess of that which is naturally generated from the ame, resulting in larger vaporization
rates. This larger mass loss rate then feeds into stronger ame
behavior (as clearly shown in Fig. 3a and b) where a strongly
buoyant structure is noted in reduced pressure environments,
where before a weaker diffusional type ame is evident (Fig. 2c
and d). In addition, ame luminosity is also signicantly larger
suggesting changes in the soot production levels.

Fig. 8. Measured steady burning mass ux as a function of pressure for different


external heat uxes at 21% oxygen concentration. Results are compared to the
analytical model in Eq. (13) to help explain the measurements.

A clearer picture of the effect of pressure on the mass loss rate


can be gained by directly plotting mass ux versus external pressure, as shown in Fig. 8. The mass ux decreases with pressure as
was shown in Fig. 7, but now we can more clearly see its dependence. For a constant externally applied heat ux, the mass ux
decreases signicantly with decreasing pressure. This is due to
the reduction of two main ame heat feedback mechanisms with
decreasing pressure. The rst is the convective heat ux at the
surface. Here, the reduction in gas density and well as the shift
in the oweld from a buoyancy-induced ow to a diffusional
ow act to modulate the convective heat transfer back to the surface. Similarly, the radiant heat ux also decreases with decreasing pressure due to reduced soot production and ame
luminosity. Furthermore, the experimental results show that
dependence of the mass ux on pressure decreases at the higher
applied external heat uxes, with essentially no pressure dependence when the external heat ux approaches 75 kW/m2. At sufciently high external heat ux, the losses in convection and
radiative feedback from the natural ame at lower pressure conditions are mitigated.
Oxygen mass fraction was varied independently with system
pressure to determine the effect of a reduced oxidizer environment
on burning and mass loss rate. The experimental measurements
showing the effect of oxygen mass fraction on the mass loss rate
are presented in Fig. 9. The measurements show a decrease in mass

Fig. 9. Measured steady burning mass ux as a function of oxygen mass fraction for
different pressures at a constant heat ux of 16 KW/m2. Results are compared to the
analytical model in Eq. (13) to help explain the measurements.

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

ux for all pressures tested when the oxygen mass fraction is


decreased. The reduction in mass ux with oxygen mass fraction
is similar at all the pressures used in this study, with a slightly
stronger correlation at higher pressures. As the oxygen mass fraction is reduced, the convective heat transfer and radiative feedback
mechanisms to the ame surface are reduced due to the reduction
in fuel burned (leading to smaller ames and less ame luminosity). It should be noted that the results at P = 0.18 atm in Fig. 9
are only available for oxygen concentration at or above
Y O2 ;1 0:16 since sustained ignition was unattainable below that
value. Lower ammability limits were not thoroughly explored
due to the difculty of maintaining steady aming combustion at
these conditions, thus these data do not extend to their limiting
oxygen mass fractions.
Lastly, the critical mass ux for piloted ignition can be determined by combining the time to ignition data with the continuously monitored sample mass ux. The critical mass ux is then
found from the slope of the mass loss curve just before sample
ignition. This critical mass ux is shown in Fig. 10a and b for two
different external heat ux conditions. At decreased pressures,
the reduction in available oxygen lowers the amount of fuel
needed to reach the lower ammability limit, thereby resulting
in a lower critical mass ux at ignition as pressure is decreased,
as shown in Fig. 10. Similar observations were made by Fereres
et al. in a horizontal, forced convection apparatus [37], indicating
that the sample material is becoming more ammable at reduced
pressures.
The mass ux at ignition appears to have a slight
functional dependence on the external heat ux, with the higher
heat ux values increasing the amount of fuel available through
pyrolysis. The external heat ux also speeds the pre-heating of
the sample and production of pyrolysis gases effectively reducing
the mass ux required at ignition. In addition, all of the critical
mass ux data tends to follow the computed re point
associated with self-sustained aming combustion of the sample.
Typically, we see that the ashpoint conditions are those
where the pyrolysis products achieve a ammable concentration,
while the re-point corresponds to the condition where the
ame is self-sustaining. The re-point typically depends on
whether the heat ux from the ame (or externally applied) is
sufcient to produce enough pyrolysis gases to sustain aming
combustion.

1525

4. Analytical model
The theoretical analysis will address ignition and burning rate
as a function of external heat ux atmospheric pressure, and oxygen. The model specically addresses the burning in the cone calorimeter conguration, or one comparable in orientation and
sample size. However, the model can be extended to larger pool
re orientations. The framework of the modeling follows established theoretical based correlation techniques using property
parameters that are derived from data taken at several heat uxes.
We will use such data from normal atmospheric conditions of pressure (14.7 psi) and oxygen (21%) to derive results at other pressure
and oxygen conditions. The modeling approach is supported by
studies from Tewarson et al. [3,5,20,23] and Hamins et al. [24] on
predicting the burning rate of pool res. The basis of the approach
for ignition and burning rate is described in more detail in Quintiere [25]. Here, only steady burning will be considered, and kinetic
effects in the solid phase are not addressed.
4.1. Ignition
4.1.1. Ignition delay time
Ignition modeling follows the approach used in ASTM E-1321
with some modication. This is described in detail in Quintiere
[25, p. 176187]. It has been found that piloted ignition data for
real materials follows the following behavior, except near the critical heat ux (CHF). The ignition time delay can be related to the
applied external heat ux, q_ 00ext , as shown in the following equation.

tig


2
TRP
q_ 00ext

The Thermal Response Parameter (TRP), coined by Tewarson


[20], is a material property combining the ignition temperature
with the material properties conductivity, density and specic heat
as

TRP

p
4

12
kqcp T ig  T 1 ;

This result can be derived from the heat conduction in a semiinnite solid with linearized surface heat loss in the limit of small
time or high external heat ux [25].

Fig. 10. Measured mass ux at ignition as a function of pressure for 21% O2 and compared to the pressure-dependent sample ashpoint and repoint for (a) 10 W/m2 external
heat ux and for (b) 12 W/m2 external heat ux.

1526

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

This equation does not accurately represent experimental ignition data when the external radiant heat ux approaches the CHF.
The CHF can be estimated by computation through an energy balance at the surface of the material just before ignition to a sustained ame at very long time. At very long time, the external
heat ux will be the CHF with the material effectively at a uniform
temperature equal to its ignition temperature. There is no heat loss
by conduction into the material. Consequently the external heat
ux is equal to the radiative and convective surface losses and
the energy required to pyrolyze the material to achieve its lower
ammability limit concentration (LFL) at the pilot location. Here,
no inclusion of surface oxidation is taken, as we assume it will
be small for our application. However, at higher than ambient oxygen concentrations, it should be considered as a source of energy to
the surface. Under the conditions specied the CHF can be found by

CHF rT 4ig  T 41 hc T ig  T 1 q_ 00py

where r is the StefanBoltzmann constant, q_ 00py pyrolysis energy


ux, and hc is the convective heat transfer coefcient. The emissivity is taken as unity, as this property is nearly unity for these combustion materials and re conditions.
The challenge is to present a unied theory for predicting the
time to ignite that addresses both high and low external heat ux
conditions with respect to the CHF. As prior experimental studies
have shown, the TRP material property can be extracted from
experimental data at small ignition delay times and higher external
heat uxes as it correlates these data very well. Using the experimentally derived TRP and theoretically determined CHF, the following equation is proposed which can be a suitable correlating
function of ignition delay times over the entire heat ux range:



CHF
CHF p
t

1

exp

ig
q_ 00ext
TRP

This equation follows the scaling of ignition delay time in the


limits of both high and low heat ux behavior. However, it should
be realized that this equation does not account for kinetic gas
phase effects that would especially become signicant at low pressure and low oxygen concentrations. Additionally, the pilot must
be capable of launching a premixed ame to achieve sustained
ignition of the material. The reaction rate in the gas phase is monotonic with pressure and oxygen by typical Arrhenius kinetics, so at
critically low levels of each, ammability will not occur. These kinetic effects in the gas phase will retard the ignition time and ultimately stop the ignition process. We will not address kinetics
explicitly in the equation for the time to ignite, but can address
it by investigating the critical mass ux needed for ignition.

it is empirically well known that a critical ame temperature is


needed at the LFL mixture [25, p.104] to create ignition. Thus, the
LFL can be written as

R T f ;crit
LFL 

T1

cp dT cp T f ;crit  T 1

Dhc
Dhc

The critical temperature is commonly taken at 1300 C. The


lower limit as an oxygen mass fraction can be alternatively considered, with the heat of combustion per mass of oxygen taken as
13 kJ/g. Using ambient temperature as 25 C, and the specic heat
as 1.2 J/g K (1000 C), then

LFLDhc Y o2 ;lim Dho2  1:53 kJ=g

_ 00ig
m

hc
LFL
cp

Drysdale [27, p 88] indicates that the LFL is fairly constant below normal atmospheric pressure down to about 0.1 atm
(1.5 psi). He also shows that the LFL is fairly constant for
oxygen concentrations above that of a normal atmosphere
conditions. At lower fuel concentrations, piloted ignition will not
be possible. The lower limit can be estimated using the fact that

It should be recognized that there is a critical energy density


needed for ignition (1.53 MJ/kg  1.1 kg/m3) of 1.68 MJ/ m3. Drysdale [27, p. 84] cites 2.16 MJ/m3, and Lyon and Quintiere [26] nds
1.9 MJ/m3. The differences between these three estimations are in
the empiricism of the various models involved, but many fuels follow these typical scaling values. This result also yields an estimate
for the lower limit oxygen concentration, for all fuels, as approximately 0.12 or about 11% oxygen concentration by volume. We
would expect no piloted ignition below this limit. The other critical
energy density values would increase this oxygen limit to about
12.8 1% by volume.
From the above analysis it follows that the critical mass ux at
ignition can be estimated from

_ 00ig 
m

hc 1530 K
Dhc

As this is the condition associated with the ashpoint, using the


stoichiometric fuel to air concentration instead of the LFL would
give a result close to the re point or sustained ignition. Drysdale
[27] showed that the ratio of the LFL to the stoichiometric concentration is about 0.55 0.03 for many fuels giving the estimate that
_ 00ig . For this
the mass ux at sustained ignition would then be 1.8 m
analysis, we dene the ashpoint as conditions under which the
pyrolysis products achieve the LFL, while the re-point corresponds to the condition where the ame can sustain itself. The
re-point is dependent on whether the heat ux from the ame
is sufcient in raising the surface temperature, where enough
pyrolysis gases are produced to sustain aming combustion.
4.1.3. Pyrolysis energy ux
In order to compute the CHF, the pyrolysis energy ux is
needed. From the analysis to estimate the critical mass ux at ignition, this follows as

_ 00ig Dhpy
q_ 00py m
4.1.2. Critical mass ux for ignition
A theoretical analysis of the critical mass ux at ignition (ashpoint) and at extinction (re point) has previously been put forth
by Lyon and Quintiere [26] showing good agreement with polymer
data. The mass ux at ignition can be related to the LFL (as a mass
fraction concentration) using convective mass transfer theory with
equal Prandtl and Schmidt numbers applicable to air:

where Dhpy is the energy per unit mass loss to vaporize the solid.
4.1.4. Effect of pressure and oxygen
To bring closure to the analysis in predicting the time to ignite,
Eq. (4) can be used to correlate ignition delay times over a wide set
of conditions. However, there are several restrictions to its use that
should be noted. Eq. (4) will not hold below a pressure of approximately 0.1 atm due to kinetic rates that will signicantly slow at
low pressures, increasing the time to ignite or rendering the material nonammable. At oxygen concentrations below about 12%
(vol.) no ignition is likely. Additionally, two crucial parameters that
will affect the time to ignite are the heat of combustion and the
convective heat transfer coefcient. From Eq. (7), as long as the
LFL does not vary signicantly (above 0.1 atm and above 12% oxygen) the heat of combustion should be invariant. However, the heat
transfer coefcient will vary with pressure.

1527

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

In the present application the combustion environment is a


cone calorimeter as described in the experimental setup. Liu
et al. [28] have measured the convective heat transfer coefcient
in a similar cone calorimeter apparatus with varying fan speeds,
and have theoretically correlated the heat transfer coefcient satisfactorily by considering combined forced and free laminar conditions. They nd that hc is 12 2 W/m2 K over a range of heat ux of
1535 kW/m2 and airow rates of 1825 g/s at normal atmospheric pressure. These experimental results were supported by a
mix of low velocity forced and free convection theory.
Therefore from laminar free convection theory, e.g. [25, p 250],

Nu  Gr 1=4
 
1=4
2
q
hc l
DT 3

g
l
k
l
T

10

hc  q1=2  p1=2
This suggests a pressure dependence on the heat transfer coefcient without combustion.
Hence, for the same temperature conditions at ignition (or aming), the same size material, and with density proportional to pressure by perfect gas theory, it follow that the heat transfer
coefcient is proportional to pressure to the one-half power. If conditions were turbulent, the pressure scaling would shift to the twothirds power, and in forced ow this would vary as one-half power
for laminar conditions and one-third power for turbulent conditions. In the present application, the one-half power for free and
forced conditions per Liu et al. [28] is used to represent the dependence of the convective heat transfer coefcient with pressure:

hc 0:012


1=2
ppsi
0:00313ppsi1=2 kW=m2 K
14:7

11

It should be noted that the heat transfer coefcient occurring in


Eqs. (3), (5), (8), and (13) all represent the approximation by Eq.
(11), as they are the pure heat transfer value without mass transfer.
4.1.5. Properties for PMMA
The CHF is computed based on an ignition temperature for
PMMA of 275 C, as a best estimate from Babrauskas [29, p.
1068]. The heat of pyrolysis is taken from Stoliarov and Walters
[30] as 0.87 kJ/g, and the heat of combustion from Tewarson [20]
of 24.2 kJ/g. Substituting accordingly in Eq. (3) gives
1=2

CHF 4:68 0:934ppsi

kW=m

12

At atmospheric pressure, the CHF is determined to be 8.25 kW/


m2 (see Fig. (5)).
4.2. Steady burning rate
The burning rate model follows the modied B number approach for a stagnant layer with added radiative heat ux considerations as explained in Quintiere [25]. Such an approach was used
by Hamins et al. [24] to reasonably predict the burning rate of pool
res for a wide range of liquid fuels over diameters ranging from
about 0.022 m. They used a gray gas model for radiation from
the ame with a mean beam length to model emissivity, and we
will employ a similar model. In the current application, the conguration is a horizontal square sample exposed to radiation from the
cone heater. Radiation from the ame and the blockage of external
radiation by the ame to the material is considered. No effect of the
radiation blockage due to the fuel pyrolysis gases is considered, as
that effect is lacking in prior experimental or theoretical results to
address. While steady burning is only considered here, the gasphase heat ux model followed can provide input to a transient
pyrolysis model to address transient burning.

4.2.1. B-number theory


Under steady burning, the net surface heat ux is equal to the
energy ux to vaporize the fuel as modeled by the heat of gasication, L.

_ 00 L
m




hc
k
Y o2 ;1 Dhc =r1  X r  cp T v  T 1
cp ek  1
q_ 00f ;r sq_ 00ext  rT 4v  T 41

13

where q_ 00f ;r is the incident ame radiative heat ux to the material,


q_ 00ext is the incident external radiative ux from the heater, s is the
transmissivity of the ame between the heater and the material,
Yox,1 is the ambient oxygen mass fraction Dhc/r is the heat of combustion per unit mass of oxygen (13 kJ/g Xr) is the ame radiative
fraction, Tv is the material surface vaporization temperature,
_ 00 cp =hc .
km
The rst term on the right hand side is the convective heating
by the ame. The convective heat transfer coefcient is taken as
that measured in the cone calorimeter, as given by Eq. (11). It is
considered laminar in representation as the ame near its base is
laminar in this representation while above it makes a transition
to turbulent ow. The convective heat ux is dependent on both
oxygen concentration and pressure.
4.2.2. Radiation from heater
The radiation effects are modeled by a homogeneous gray-gas
representation for the ame. The entire gas phase above the material to the ame tip is considered. The blockage of the radiative ux
from the heater is determined from the transmissivity given in
terms of a mean beam length, lm,e.

s ejf lm;e

14

The mean beam length is given by [31, p 180].

lm 3:6V=A

15

where V is the volume of the gas, A is the bounding surface of the


gas,
Here the volume is considered to be a hemisphere with a base
diameter (D) as the effective diameter of the square (s  s) sample,
i.e. D = (2/p1/2)s where s is the length of one side of the square sample. Consequently,

pD=23
0:6D 0:677 s
lm;e 3:6
2pD=22
2
3

16

4.2.3. Radiation from ame


The radiant heat ux from the ame is modeled in a similar
fashion with the volume considered as a cylinder of diameter D
and height Lf, the ame height. In this case the mean beam length,
lm,f, is given as

lm;f

p D2 L
4

0:9

pDLf 2  p4 D2


Lf =D
1:01s
Lf =D 1=2


Lf =D
D
Lf =D 1=2
17

Here Lf cannot exceed the vertical duct height of the cone calorimeter, as that would limit the vertical ame height.
The incident ame heat ux is given as

q_ 00f ;r ef rT 4f  T 41

18

where the emissivity is given as

ef 1  ejf lm;f

19

1528

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

The ame height is computed as a turbulent ame [32]


although the ame may resort to laminar at low pressures. As there
is no clear way to differentiate in this study, we maintain the turbulent equation of Heskestad [32]
Lf
D


15:6 q

or
Lf
D

2=5 

Y o2 ;1 cp T 1 3=5
Q_
p 5=2
Dhc =r
gD

1 T 1 cp

0:23Q_ 2=5

14:7
p

2=5 

0:233
Y o2 ;1

3=5

20
 1:02

 2
p
po

21

where jo and po are the absorption coefcient and pressure at 1


atm (14:7 psi)
It is likely that the absorption coefcient also is affected by
ambient oxygen concentration due to the dependence of soot formation on local oxygen. The soot levels would be expected to drop
as the oxygen concentration is reduced from normal ambient, but
we will not speculate in including this likely effect.
To complete the model for ame radiation the ame temperature must be estimated. From Quintiere [25, p. 316] the temperature can be computed from
T f T 1
T1

C T;f 1  X r Y o2 ;1 DcphTc 1=r

where C T;f 0:50 for cp 1:0 kJ=kg K

22

T f  T 1 65001  X r Y ox;1 K
The ame radiative fraction likely is affected by pressure and
oxygen, especially as the ame nears extinction due to changes
in the soot volume fraction. It is well known that for res over
1 m in diameter, the radiation fraction also drops due to soot
blockage. The latter does not apply to the present application,
and the former cannot be described due to lack of detailed experimental data about the soot levels as a function of the conditions
varied in this study. So in the current application, the radiation
fraction will be taken as a constant for a given fuel.
4.2.4. Empirical analysis
As an approximation to Eq. (13), the ame energy balance can
be written as

_ 00 L  q_ 00ext  rT 4v  T 41  q_ 00f ;c q_ 00f ;r


m

23

Using simple scaling, we see that the inuence of the convective


ame heat ux on the mass loss rate scales with the square root of
external pressure, Eq. (10), and linearly with the oxygen mass fraction, Eq. (13). The convective ame heat ux can then be approximated as

q_ 00f ;c / p1=2 Y o2 ;1

24

Similarly, for optically thin radiant environments (small jfD), a


ame radiant heat ux scaling can also be derived used Eq. (21), to
see the pressure scaling effect on the ame emissivity, and Eq. (22),
to see the oxygen mass fraction effect on the ame temperature.
The overall ame radiant heat ux scaling is approximately proportional as

q_ 00f ;r / p2 Y 4o2 ;1
This means that

26

This functionality is investigated for the data as a way to simplify the inuence of pressure and oxygen on the burning rate, at
least where the above approximations are valid.

 1:02

_ 00 s2 Dhc in kW, p is pressure in psi


where Q_ m
It is seen that both pressure and oxygen affects the ame
height.
The effect of pressure also affects the ame absorption coefcient as identied by de Ris et al. [33]. They show that it depends
on pressure to the second power. Therefore, we represent

jf jo

_ 00 L  q_ 00ext  rT 4v  T 41  functionp1=2 Y o2 ;1
m

25

5. Comparison of measurements with analytical model


In order to compare the experimental measurements to the
simple analytical model proposed, the burning of the samples
needed to show a steady burning region such as in Fig. 4 which
was the case for all the runs. The main objective of the experimental study is to determine the sample ignition time and burning rate
for normal to low-pressure environments and normal to low oxygen concentrations. To better understand the measured time to
ignition of the sample, this can be compared with that predicted
from Eq. (1) in Fig. 5. The good t suggests that the scaling is valid.
The parameter TRP is determined from a t to the time to ignite
data at normal pressure. Using Eq. (1), the TRP is found to be
238 kW s1/2/m2. However, using a more unied theory that takes
into account the inuence of the CHF at long ignition delay times
(Eq. (4)), the relevant TRP is 179 kW s1/2/m2, as shown in Fig. 5.
Knowing the TRP, the only unknown parameter left in the analytical model for time to ignition in Eq. (4) is the CHF which contains
the pressure effect. This can be found using Eq. (12) and analytical
model prediction for tig can be obtained.
The analytical model for time to ignition previously described in
Eq. (4) is also shown in Fig. 6 as solid lines. Similar to the experiments, the model shows that the effect of pressure is negligible
at high heat uxes and hence can be neglected. At lower levels of
external heat ux the convective losses became important as
shown for 10 and 12 kW/m2. When comparing the analytical model to the experimental results in Fig. 6, the model works well for
test conditions at the lower heat uxes, but underestimates the
ignition times for conditions at the higher heat uxes.
The experimental mass loss rate measurements are also compared to the analytical model for the burn rate in Eq. (13) in
Fig. 7. The model can be computed by calculating all the heat ux
components from Eqs. (14)(22). The terms not know in these
equations include the soot emitter parameter, and the radiative
fraction. These can be obtained from [23,33] to give jf = 1.3 m1
and Xr = 0.34 respectively. Also, the experimental value for the
vaporization temperature Tv = 350400 C was obtained from
[35] and an average vaporization temperature was used in the
model as 375 C. Last, the heat of gasication is obtained by plotting the inverse of the slope of mass ux at ambient condition versus the external heat ux [36], resulting in a value of 2.2 kJ/g.
Having estimated values for all the terms in the right hand side
of Eq. (13), the burn rate can be analytically calculated for all the
test conditions used. The model predicts the measurements at
low external heat ux up to about 25 kW/m2 but clearly underpredicts the mass ux at higher external heat uxes where the
experiments show that the effect of pressure is negligible. The effect of pressure on the mass loss rate is also predicted by the analytical model in Fig. 8 although the decrease in mass ux is not as
pronounced as the predictions by the model. Also the effect of the
larger external heat uxes of 25 kW/m2 and above are not well predicted by the simple model used. The analytical model showing the
effect of oxygen mass fraction on the mass loss rate is presented in
Fig. 9. The values used for the model in Eq. (13) used in this gure
are cp = 1 kJ/kg K, hc = 12 W/m2 K, sf = 1.3, Tf = 1300 K, Tv = 648 K,
T1 = 300 K, L = 2.2 kJ/g, Xr = 0.34. Although the model has some
limitations at the lower oxygen mass factions, it captures the trend
of the decrease in mass ux for all pressures tested when the oxygen mass fraction is decreased.

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

The critical mass ux determined experimentally can also be


compared with the analytical model in Eq. (8) with the heat of
combustion taken at the re-point, or where aming combustion
can be sustained. The scaling of the critical mass ux at ignition
with pressure comes largely from the dependence of the convective transfer coefcient, since the LFL and re-point of the sample
do not signicantly vary with pressure over this range [4]. The
model predicts lower critical mass ux at ignition as pressure is decreased as shown in Fig 10. At the lowest pressure tests conducted
in this study, the critical mass ux data tends to correlate less well
with the theoretical re-point prediction, consistent with the
chemical kinetic timescale becoming longer relative to the convective timescale.When combining the results from Figs. 8 and 9, they
indicate that the mass loss rate scales with pressure and oxygen
concentration. Thus, the mass loss rate measurements are plotted
against the proposed scaling variable [p1=2 Y O2 ;1 ] from Eq. (26) in
Fig. 11. The scaling suggests that the mass loss rate should be proportional to the product of the square root of pressure and oxygen
concentration. The plot in Fig. 11 shows that the properly normalized mass loss measurements lie onto a single line that can be tted by a power function of the form y = axn, where a = 64, and
n = 1.3 to give

_ 00 
m

q_ 00ext  q_ 00rr
64p1=2 Y O2 ;1 1:3
L

27

where q_ 00rr rT 4v  T 41 is the re-radiation heat ux from the surface. This is a simple relation that works for the measurements of
all PMMA samples burned at different pressures, oxygen concentrations and external heat uxes. By plotting the y-axis in Eq. (27) as
_ 00  q_ 00ext  q_ 00rr =L it allow us to show that the data is independent
m
of heat ux. The only exceptions are the results for external heat
uxes above 50 kW/m2 that are not shown since this large external
heat ux dominates the mass loss rate.
The above simple scaling does not include modeling of a nite
chemical reaction timescale, which can play an important role in
ignition and extinction. A Damkohler number could be created
which combines the chemical reaction timescale with that from
the convective heat transfer. At the highest heat ux cases examined in this study, it is possible that the timescale associated with
convective heat transfer combined with the very strong buoyancy
generated in the ame oweld has decreased to be of the same
order as the fundamental chemical timescale [38]. This introduction of a Damkohler number dependence would allow for the
extension of the proposed model to higher heat ux values. At

1529

these higher externally applied heat ux values, the ame temperature may also be signicantly affected, inuencing the diffusion
coefcient and introducing a Lewis number effect [39]. These higher order effects may limit the correlation of this model with very
high external heat ux cases (as seen and discussed in Fig. 6).
Due to these signicant added complexities, we have chosen to
limit the applicability of our model to a lower externally applied
heat ux range where these additional scaling parameters are of
less import.

6. Conclusion
An experimental study of the ammability properties of PMMA
at low pressures and oxygen concentrations was performed in a
large 10 m3 pressure vessel, capable of reaching pressures as low
as 0.1 atm. The PMMA ammability was characterized by the burning rate and time to ignition of 10  10 cm2 samples. Tests were
performed at different applied external heat uxes ranging from
10 to 72 kW/m2 with burning rates obtained during steady burning
for all cases. Other test conditions included oxygen concentrations
from 1221%, and pressures from 0.18 to 1 atm. For each burn test
condition, the experiment was repeated 6 times in the mass loss
calorimeter inside the pressure vessel.
Experimental measurements allowed the observation of the effect of pressure and oxygen concentration on the burning rate. The
results show that at low pressure the burning was less intense,
which is shown by the decrease in the mass loss rate. On the other
hand, the reduction of pressure causes the sample to ignite faster
and at a lower critical mass ux, but this is only relevant at low levels of external heat ux. In general, the experimental measurements showed a good agreement with the power law t
obtained for the relation of the combined pressure and oxygen effect on the burning rate. This correlation predicts the burning rate
behavior for the full range of pressure and oxygen measurements
obtained.
Experimental measurements were compared with a simple analytical model. The results show how pressure and oxygen concentration contributed to the heat transfer from the ame. Pressure
affects the heat transfer trough the convective heat transfer coefcient and the effective soot emitter parameter. The convective heat
transfer coefcient affects the convective heat losses which are
diminished at low pressure, causing the sample to ignite faster.
Similarly it explains the decrease in the burning rate through lowering of the heat feedback mechanism to the sample surface as
well as the decrease in the critical mass ux at ignition through
the reduction in the available oxygen. The lower oxygen content
the reduces then ame temperature, thereby lowering the net heat
ux back to the sample surface.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]

Fig. 11. Measured mass loss rate plotted as a function of the proposed scaling
variable (p1=2 Y O2 ;1 ) from Eq. (26). The scaling suggests that the mass loss rate
should be proportional to the product of the square root of pressure and oxygen
concentration.

[10]
[11]
[12]
[13]

R. Friedman, Fire Mater. 20 (1996) 235243.


C.K. Law, G.M. Faeth, Prog. Energy Combust. Sci. 20 (1994) 65113.
A. Tewarson, J. Fire Sci. 18 (2000) 183214.
S. McAllister, C. Fernandez-Pello, D. Urban, G. Ruff, Combust. Flame 157 (2010)
17531759.
A. Tewarson, J.L. Lee, R.F. Pion, Proc. Combust. Inst. 18 (1981) 563570.
Y. Wang, L. Yang, X. Zhou, J. Dai, Y. Zhou, Z. Deng, Fuel 89 (5) (2010) 1029
1034.
D.L. Dietrich, H.D. Ross, Y. Shu, P. Chang, J.S. Tien, Combust. Sci. Technol. 156
(1) (2000) 124.
S.L. Olson, Combust. Sci. Technol. 76 (4) (1991) 233249.
J. Kleinhenz, I. Feier, S. Hsu, J. Tien, P. Ferkul, K. Sacksteder, Combust. Flame
154 (2008) 637643.
A.E. Frey, J.S. Tien, Combust. Flame 26 (1976) 257267.
R. McAlevy, R.S. Magee, Proc. Combust. Inst. 12 (1969) 215227.
J.B. West, Aviat. Space Environ. Med. 68 (1997) 159163.
Y. Nakamura, N. Yoshimura, H. Ito, K. Azumaya, O. Fujita, Proc. Combust. Inst.
32 (2009) 25592566.

1530

M. Zarzecki et al. / Combustion and Flame 160 (2013) 15191530

[14] J. Li, J. Ji, Y. Zhang, J. Sun Chin. Sci Bull. 54 (2009) 19571962.
[15] J.S. Goldmeer, Extinguishment of a Diffusion Flame Over a PMMA Cylinder by
Depressurization in Reduced-Gravity, NAS 1.26:198550; NASA-CR-19855,
NASA Glenn Research Center.
[16] S. Olson, Fire Saf. Sci. 10 (2011) 959970.
[17] O.V. Roditcheva, X.S. Bai, Chemosphere 42 (57) (2001) 811821.
[18] A. Beltrame, P. Porschnev, W. Merchan-Merchan, A. Saveliev, A. Fridman, L.A.
Kennedy, O. Pertrova, S. Zhdanok, F. Amouri, O.S. Charon, Combust. Flame 124
(12) (2001) 295310.
[19] R.E. Lyon, in: C.A. Harper (Ed.), Handbook of Building Materials for Fire
Protection, McGrawHill, New York, 2004, pp. 151 (Chapter 3).
[20] A. Tewarson, Generation of Heat and Chemical Compounds in Fires, SFPE
Handbook of Fire Protection Engineering, third ed., National Fire Protection
Association, 2002, Section 3, pp. 82162.
[21] M. Kindelan, F.A. Williams, Combust. Sci. Technol. 10 (1975) 119.
[22] C. Vovelle, C. Akrich, J.L. Delfau, Combust. Sci. Technol. 36 (1984) 118.
[23] A. Tewarson, R. Pion, Combust. Flame 26 (1976) 85103.
[24] A. Hamins, J.C. Yang, T. Kashiwagi, Global Model for Predicting the Burning
Rates of Liquid Pool Fires, NISTIR 6381, 1999, pp. 139.
[25] J.G. Quintiere, Fundamentals of Fire Phenomena, John Wiley & Sons, Ltd., West
Sussex, England, 1993.
[26] R.E. Lyon, J.G. Quintiere, Combust. Flame 151 (4) (2007) 551559.

[27] D. Drysdale, An Introduction to Fire Dynamics, third ed., John Wiley & Sons,
Ltd., Chichester, UK, 2011.
[28] X. Liu, J.G. Quintiere, Flammability Properties of Clay-Nylon Nanocomposites,
DOT/FAA/AR-07/29, Federal Aviation Administration, 2007.
[29] V. Babrauskas (Ed.), Ignition Handbook, Fire Science Publishers/Society of Fire
Protection Engineers, Issaquah WA, 2003.
[30] S.I. Stoliarov, R.N. Walters, Polym. Degrad. Stab. 93 (2) (2008) 422427.
[31] R. Siegel, J.R. Howell, Thermal Radiation Heat Transfer, third ed., Hemisphere,
New York, 1992.
[32] G. Heskestad, Fire Saf. J. 5 (2) (1983) 103108.
[33] J. de Ris, P.K. Wu, G. Heskestad, Proc. Combust. Inst. 28 (2000) 27512759.
[34] ASTM E 1354, Standard Test Method for Heat and Visible Smoke Release Rates
for Materials and Products Using Oxygen Depletion, American Society for
Testing and Materials, Philadelphia PA.
[35] T. Kashiwagi, A. Inaba, J.E. Brown, Fire Saf. Sci. 1 (1986) 483493.
[36] B.T. Rhodes, J.G. Quintiere, Fire Saf. 26 (3) (1996) 221240.
[37] S. Fereres, C. Lautenberger, C. Fernandez-Pello, D. Urban, G. Ruff, Combust.
Flame 158 (2011) 13011306.
[38] S. Venkatesh, A. Ito, K. Saito, I.S. Wichman, Proc. Combust. Inst. 26 (1996)
14371443.
[39] R.C. Corlett, A. Luketa-Hanlin, in: K. Saito (Ed.), Progress in Scale Modeling,
Springer, New York, 2008, pp. 8597.

You might also like