ABSTRACT
Title of Thesis: IMPROVED VENTING FOR
FLAMMABILITY LIMIT TESTING
USING ASTM E681 APPARATUS
Conor G. McCoy, Master of Science, 2016
Directed by: Associate Professor Peter B. Sunderland
Department of Fire Protection Engineering
The literature on the determination of flammability limits was reviewed
and experts on the ASTM E681 standard were interviewed to identify new
means of improving the reproducibility of the ASTM E681 test. Venting was
identified as a variable of flammability limits not yet addressed. Limitations
of the current system for sealing and venting (a rubber stopper) were iden-
tified and addressed by the development of a custom burst disc. The burst
disc was evaluated for its ability to hold and maintain a vacuum, its ability
to vent at pressures of interest, and for its venting phenomena. The burst
disc was deemed to be a satisfactory alternative to the rubber stopper and
is recommended to be included in the ASTM E681 standard.
IMPROVED VENTING FOR FLAMMABILITY LIMIT
TESTING
USING ASTM E681 APPARATUS
by
Conor G. McCoy
Thesis submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Master of Science
2016
Advisory Committee:
Professor Peter B. Sunderland, Chair
Professor Emeritus James G. Quintiere
Professor Stanislav I. Stoliarov
c© Copyright by
Conor G. McCoy
2016
Acknowledgments
I have so many amazing individuals to thank for helping. To begin, I
would like to thank my family for supporting me throughout the grad school.
I also want to thank Peter Sunderland for mentoring me and giving me
this opportunity in the first place. Dr. Sunderland, you took a chance on
me and I am eternally grateful for it.
I would like to thank Vivien Le Coustre for his work in starting the
project, especially helping with the literature review and interviews of ex-
perts. I also have to thank my friends, Peter Lomax and Dennis Kim for their
help in the lab, general advice, and for working together through classes.
Finally, I must thank all of the fantastic interns who served on this
project. Niko Thomas, thank you for your help in building the ASTM E681
apparatus. Gauthier Pipard, thank you for constructing the burst disc, help-
ing run tests, and machining anything we needed. Last but not least, thank
you Angela Hermann for making all the technical drawings, acting as our
test technician, and helping in the analysis of the results.
Much gratitude to ASHRAE for sponsoring this project
to make this possible.
ii
Contents
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Purpose of this work . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.1 Common Definitions and Equations . . . . . . . . . . . 3
1.3.2 Flammability Test Methods . . . . . . . . . . . . . . . 5
1.3.3 Variable Effects on Flammability Limits . . . . . . . . 8
1.3.4 Theoretical Work . . . . . . . . . . . . . . . . . . . . . 13
1.3.5 Small-Scale Testing . . . . . . . . . . . . . . . . . . . . 16
1.3.6 Work of Shigeo Kondo et al. . . . . . . . . . . . . . . . 17
1.3.7 Literature Review Conclusions . . . . . . . . . . . . . . 20
2 Interview Results 21
3 Experimental Apparatus 24
3.1 ASTM E681 Apparatus . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Burst Disc Development . . . . . . . . . . . . . . . . . . . . . 25
4 Results and Discussion 28
4.1 Vacuum and Leak Testing . . . . . . . . . . . . . . . . . . . . 28
4.2 Venting Pressure Testing . . . . . . . . . . . . . . . . . . . . . 30
4.2.1 Burst Disc . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2.2 Rubber Stopper . . . . . . . . . . . . . . . . . . . . . . 32
4.3 Determination of the LFL of R-32 . . . . . . . . . . . . . . . . 34
4.3.1 Flammability Criteria . . . . . . . . . . . . . . . . . . 34
4.3.2 Rubber Stopper . . . . . . . . . . . . . . . . . . . . . . 35
4.3.3 Burst Disc . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3.4 Comparison With Past Data . . . . . . . . . . . . . . . 37
4.4 Burst Disc Venting Performance . . . . . . . . . . . . . . . . . 38
4.5 Direct Flame Angle Measurement . . . . . . . . . . . . . . . . 42
4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5 Conclusions 47
A Appendix: Design Drawings for Burst Disc 49
B Appendix: Supplementary Flame Images 53
iii
List of Tables
1 Performance at Vacuum Pressure . . . . . . . . . . . . . . . . 28
2 LFL values for R-32 . . . . . . . . . . . . . . . . . . . . . . . . 37
3 Stopper Venting Time . . . . . . . . . . . . . . . . . . . . . . 38
4 Burst Disc Venting Time . . . . . . . . . . . . . . . . . . . . . 41
iv
List of Figures
1 Schematic of Experimental Apparatus . . . . . . . . . . . . . . 24
2 Burst Disc Without and With Cover . . . . . . . . . . . . . . 26
3 Burst Disc Foil Sheet and Cover Plate Separated . . . . . . . . 26
4 Knife Edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 Diagram of Knife Edge and Micrometer Station . . . . . . . . 27
6 Pressure Time Histories at Vacuum . . . . . . . . . . . . . . . 29
7 Leak Rates for Different Pressures . . . . . . . . . . . . . . . . 30
8 Front View of a Successful Burst . . . . . . . . . . . . . . . . 31
9 Side View of a Successful Burst . . . . . . . . . . . . . . . . . 31
10 Burst Disc Venting Pressures . . . . . . . . . . . . . . . . . . 32
11 14.8% Test Showing Stopper Displacement . . . . . . . . . . . 33
12 Diagram showing the proper location for the apex of the angle
measurement [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 34
13 An example of a “Nonflammable” result with the stopper at
14.7% . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
14 An example of a “Flammable” result with the stopper at 14.8% 36
15 An example of a “Nonflammable” result with the burst disc
at 14.6% . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
16 An example of a “Flammable” result with the burst disc at
14.7% . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
17 Flame during venting (14.8%) . . . . . . . . . . . . . . . . . . 39
18 Audio of 14.5% Test . . . . . . . . . . . . . . . . . . . . . . . 40
19 Flame of 14.5% Test . . . . . . . . . . . . . . . . . . . . . . . 40
20 Audio of 14.8% Test . . . . . . . . . . . . . . . . . . . . . . . 40
21 Flame of 14.8% Test . . . . . . . . . . . . . . . . . . . . . . . 41
22 Direct Angle Measurement for a Flame with the Burst Disc
(14.3%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
23 Direct Angle Measurement for an Asymmetric Flame with the
Stopper (14.9%) . . . . . . . . . . . . . . . . . . . . . . . . . . 43
24 Results of Measured Flame Angles . . . . . . . . . . . . . . . 45
25 Flask Neck Clamp Drawing . . . . . . . . . . . . . . . . . . . 49
26 Nylon Block Drawing . . . . . . . . . . . . . . . . . . . . . . . 50
27 Cover Plate Drawing . . . . . . . . . . . . . . . . . . . . . . . 51
28 Assembled Burst Disc . . . . . . . . . . . . . . . . . . . . . . 52
29 14.3% Test with stopper . . . . . . . . . . . . . . . . . . . . . 53
30 14.4% Test with stopper . . . . . . . . . . . . . . . . . . . . . 53
31 14.4% Test with stopper . . . . . . . . . . . . . . . . . . . . . 54
32 14.8% Test with stopper . . . . . . . . . . . . . . . . . . . . . 54
v
33 14.9% Test with stopper . . . . . . . . . . . . . . . . . . . . . 55
34 14.3% Test with burst disc . . . . . . . . . . . . . . . . . . . . 55
35 14.5% Test with burst disc . . . . . . . . . . . . . . . . . . . . 56
36 14.8% Test with burst disc . . . . . . . . . . . . . . . . . . . . 56
vi
1 Introduction
1.1 Background
Accurate determination of flammability limits is critical for safe practices
worldwide. Flammability limits are used to design and enforce safe practices
in storage, transportation, and use of all kinds of chemicals. Flammability
limits are important for safety in domestic settings as well because of the
large presence of refrigerants in domestic appliances. After the phase out of
chlorofluorocarbons (CFC) and hydrochlorofluorocarbons (HCFC) because of
their ozone-depleting effects, refrigerants have been replaced with the more
flammable hydrofluorocarbons (HFC).
There exist a number of test standards for determining flammability lim-
its. Examples include ASTM E681 (U.S.), DIN 51649 (E.U.), EN1839T
(E.U.), EN1839B (E.U.), and GB/T12474 (China). ASTM E681, DIN 51649,
EN1839T, and GB/T12474 all use glass vessels and a visual criteria to deter-
mine if a mixture is flammable. However, ASTM E681 uses a spherical flask
while the rest use a vertical tube. EN1839B is a steel sphere or bomb-type
and uses a pressure rise criteria for the determination of flammability.
In general, there are two types of tests for the determination of flammabil-
ity limits. Tests either rely on visual propagation such as flame detachment
or traveling a certain length, or use a pressure rise criteria, often a percentage
of the initial pressure. Some researchers have also modified the standard ap-
paratuses or even created their own unique apparatus. Needless to say, there
are many different approaches, variables, and configurations in flammability
1
testing. Consequently, there can be significant differences in the results.
This work shall focus on ASHRAE Standard 34, which is a modified
version of ASTM E681 specifically for the determination of flammability
limits of refrigerants.
1.2 Purpose of this work
This work sought to understand the complications of flammability limit test-
ing and find new, unexplored ways to improve the reproducibility of the
ASTM E681 test, following the guidelines of ASHRAE standard 34. The
approach adopted was as follows.
First, it was necessary to understand the current knowledge and criticisms
on the determination of flammability limits, especially with respect to the
ASTM E681 test method and refrigerants. This was accomplished by a
literature review in conjunction with interviews of experts. Next, an ASTM
E681 apparatus was built and tested for investigating the test and for future
evaluation of solutions. Details of the development of the apparatus can be
found in the thesis of P. Lomax [2]. After reviewing the findings from the
testing of this work, the interviews, and the literature, it was determined
that the effects of venting on flammability limit testing has not yet been
addressed. Finally, a solution to the complication of venting − in the form
of a burst disc − was developed and evaluated.
2
1.3 Literature Review
A review of the literature was conducted on the determination of flamma-
bility limits and on burning velocity measurements. Burning velocity mea-
surements have been used in conjunction with flammability limits to classify
flammability. Both theoretical and experimental work have been reviewed.
A summary of the work reviewed is given below and sorted by sub-topics.
Summaries are in chronological order within their sub-topic.
Before the summaries is a short section containing common equations and
concepts used in the literature of flammability. These concepts are mentioned
in the summaries so they are included here to aid your understanding.
1.3.1 Common Definitions and Equations
Lower Flammability Limit (LFL):
From the ASTM E681 standard, the LFL is defined as “The minimum con-
centration of a combustible substance that is capable of propagating flame
in a homogeneous mixture of the combustible and a gaseous oxidizer under
the specified conditions of the test” [3].
Upper Flammability Limit (UFL):
From the ASTM E681 standard, the UFL is defined as “The maximum con-
centration of a combustible substance that is capable of propagating flame
in a homogeneous mixture of the combustible and a gaseous oxidizer under
the specified conditions of the test” [3].
3
Le Chatelier’s formula:
Le Chatelier’s formula [4]is a method to predict the lower flammability limit
of a mixture given the flammability limits of its components as well as the
fraction of each component. Much work in the literature is spent modifying
this formula to improve its accuracy. The justification for this formula is
assuming the heat of combustion per mol is constant for each component
and for the mixture itself.
1
L
=
∑ ci
Li
(1)
Where L is the lower flammability limit of the mixture, ci are the mol frac-
tions of the component gases and Li are the lower flammability limits of each
component gas.
Flame/Propagation Velocity:
Flame or Propagation velocity refers to the linear velocity of the flame front.
It can be found by tracking the distance traveled by the flame front over time
and calculated accordingly − often done via high speed video.
Burning Velocity:
Burning velocity refers to the velocity of the flame relative to the unburned
gas mixture in front of it. It can be found from the linear velocity by
Su = Ss · af
Af
(2)
Where Su is the burning velocity, Ss is the propagation velocity, af is the
cross-sectional area of the flame, and Af is flame surface area.
4
F-Substitution Rate:
F-substitution rate [5] is an important parameter of hydrofluorocarbons. It
refers to the percentage of fluorine atoms with respect to the total number
of fluorine and hydrogen atoms. The equation is given by
F − substitutionRate = nF
nF + nH
(3)
F number:
F number is an important relationship for prediction of flammability limits
for hydrofluorocarbons. It is defined as F = 1−G where G is the geometric
mean of the flammability limits G =
√
UL where U is the UFL and L is the
LFL. Making the substitution gives
Fnumber = 1−
√
UL (4)
1.3.2 Flammability Test Methods
H.F. Coward at the Bureau of Mines gave an extensive foundation for flamma-
bility limit testing of individual gases (and a few mixtures) using a long tube
apparatus [6]. Although his results were later criticized − other researchers
argue that the diameter of the tube apparatus used by Coward was too small
and led to quenching − his results were widely used for comparison and
validation of early experiments and modeling. Coward’s work was an early
demonstration of the widening of flammability limits from increased pressure
and/or increased temperature [6].
G. De Smedt et al. compared experimentally two different tests: the VDI
5
2263 (20 L bomb-type based using pressure rise) and the DIN51649 (glass
tube using visual criteria [7]. It was found that a 2% pressure rise criteria
corresponded to flame propagation to the top of the tube [7].
K. Takizawa et al. determined the burning velocity of mildly-flammable
refrigerants using both Schlieren photography and the Spherical Vessel (SV)
method [5]. They determined that these methods are valid for mildly flammable
compounds and that burning velocity strongly depends on the ratio of hydro-
gen to fluorine atoms. Takizawa et al. continued to conduct their burning
velocity measurements and later measured the burning velocity of fluoro-
propanes using their two apparatuses. They also found that fuels with lower
F-substitution rates had higher burning velocities and had higher equilibrium
concentration of H and OH species [8]. This led them to believe that the F-
atom has an inhibitory effect by reacting with the active H-species forming
the stable compound HF, terminating the chain reaction [8]. Measurements
were also later done for R-1234yf using the same method but with varied
humidity and oxygen-enriched air because of its low burning velocity with
normal air [9].
V. Schroder carefully reviewed the standards used worldwide, specifically
the DIN 51649-1 (tube), EN1839T (tube), EN1839B (bomb), and ASTM
E681(spherical flask) [10]. He concluded that the European standards are
more conservative than the American standard in their definition of flamma-
bility limits − European flammability limits use the first concentration where
flame fails to propagate as the LFL and UFL giving lower LFL and higher
UFL. It was also concluded that the visual tests agreed well enough with one
another, but the bomb type was deemed unsuitable based on its significant
6
deviations in LFL from the other tests [10].
F. Van den Schoor et al. compared two test methods: the DIN51649,
which is a tube method (d=60 mm) using visual propagation, and a closed
sphere (bomb-type) using pressure rise [11]. The flammability criteria used
are 100 mm flame detachment and 5% pressure for their respective tests.
Methane-air and hydrogen-air mixtures were tested. The tube method had
wider flammability limits than the bomb-type, suggesting that the tube
method is more conservative. Finally, numerical analysis based on either
limiting flame speed or limiting flame temperature was conducted assum-
ing either spherical or planar propagation. Their numerical models failed to
agree with the UFL measurements, but had slight success for the LFL mea-
surements. The analysis using a limiting flame velocity assuming spherical
propagation best fit the data [11].
D. Crowl et al. argued that existing pressure rise criteria in test methods
are arbitrary and not “fundamentally-based” [12]. They proposed instead
using the second-derivative of pressure rise as it represents an acceleration
signifying the reaction. An expression was derived to calculate the second-
derivative of pressure rise from the maximum pressure difference, burning
velocity, time rate-of-change of flame front area, vessel volume, and den-
sity [12]. Pressure-time data was collected using a 20 L bomb-type with
wire fusing ignition of hydrogen-air mixtures. The flammability limits were
determined as the turning points of the sign of the second derivative where
a positive sign indicates growth and a minus sign, extinction. Their results
agreed fairly well with published values, but were found to be more conser-
vative giving a wider range of flammability.
7
I. Zlowchower et al. conducted tests using two spherical bomb-type tests
(20 L and 120 L) with a 7% pressure rise criterion for flammability [13]. Fuels
tested included methane, propane, ethylene, carbon monoxide, and hydrogen.
The results of the 120 L bomb-type were in good agreement with published
results using the 12 L flask (visual criteria). This suggests that the 12 L
vessel is large enough and that the 7% pressure rise is suitable since it aligns
with the visual criteria. Additionally, Zlochower et al. found the Limiting
Oxygen Concentration (LOC) for the fuels diluted with nitrogen. Predictions
for the LOC using Le Chatelier’s rule agreed well with experimental results.
D. Clodic et al. measured ranges of burning velocities for various fuel-air
mixtures including R-32, propylene, propane, R-152a, ammonia, and R-143a
at varying concentrations [14]. A tube-type apparatus was used. Burning
velocity was determined by measuring the flame propagation velocity through
video-photography and multiplying by the ratio of effective cross-sectional
area (subtracting the dead space) to flame surface area. Flame surface area
was found by fitting points to flame images, obtaining a function, and then
rotating the function to obtain an area. Ammonia, R-32, and R-143a had
significantly lower burning velocities than the other fuels.
1.3.3 Variable Effects on Flammability Limits
M.V. Blanc et al. were some of the first to investigate the minimum ignition
energy for flammability limit testing − using methane-air mixtures − by
changing the electrode spacing and measuring the voltage and capacitance
[15]. Too large of a distance and the energy is lost to the surroundings and
if too close, there is quenching between the electrodes.
8
A.L. Furno et al. investigated the LFL for upward and downward flame
propagation for hydrogen-air and butane-air mixtures [16]. They attempted
to explain the relationship between the upward and downward limits using
preferential diffusion [16]. They found the following expression (Do/Df )
1/2 ∼
LFLup/LFLdown after comparing the calculation on the left hand side with
their collected data. Subscripts o and f refer to oxygen and fuel respectively.
A. Takahashi et al. evaluated different types of wire fusing ignition sources
by changing the wire material and supplied power [17]. Methane-air mixtures
were tested in a bomb-type vessel and ignition was determined by a pressure
rise [17]. They later investigated the effect of vessel size and shape on flamma-
bility limits, testing methane-air and propane-air mixtures in eight different
cylinders and the 12 L flask [18]. It was determined that the results of the 12
L flask agreed well enough with the 160 L cylinder. They concluded then that
the 12 L flask is sufficiently large to minimize the effects of quenching [18].
Finally, they recommended a minimum diameter of 30 mm and a minimum
height of 60 cm for tubes used in flammability testing.
R. Richard sought to determine appropriate small-scale criteria by com-
paring the flammability of refrigerant blends in large-scale testing with small-
scale testing. More specifically, flammability testing in a 200 L cylinder was
compared with testing in a 12 L vessel with the same temperature, humidity,
and ignition controls [19]. It was determined that flame propagation in the
large cylinder corresponded to a 90 ◦ fan-shape flame front [19]. This is the
criteria adopted by the ASHRAE standard 34.
Le Chatelier published later a paper on the flammability limits of car-
bon monoxide-air mixtures and the effects of tube diameter, pressure, and
9
temperature using a bell-jar apparatus [20]. His results echoed previous find-
ings that high temperature widens flammability limits, low pressure narrows
flammability limits, and larger diameters widen flammability limits [20].
G. Ciccarelli investigated the flammability limits of hydrogen-air and
ammonia-air mixtures at elevated temperatures 20− 600 ◦C in a 12 L cylin-
drical vessel [21]. Flammability criterion was very conservative, such that
any “measurable pressure rise” resulted in being deemed “flammable” [21].
More specifically, Ciccarelli found that flammability limits for these mixtures
were linearly dependent on temperature.
Y. Shoshin et al., similar to Takahashi et al., explored the effects of vessel
size on flammability limits by changing tube diameter [22]. Tested were
methane-hydrogen-air mixtures. Flame speed and position were recorded
using a PIV camera. Smaller tube diameters were found to reduce flame
speed. The authors attribute this to heat loss and viscous interaction. LFL
also decreased with larger diameters.
Z. Li et al. determined the flammability limits of methane-nitrogen-air
mixtures at temperatures below atmospheric (150-300K) [23]. The appara-
tus was a stainless-steel cylinder cooled by a cryostat. Flammability criterion
used was an absolute pressure rise of 10 kPa or more. It was found that de-
creased temperature narrows the flammability limits (higher LFL and lower
UFL). They modified Le Chatelier’s formula to include terms for dilution
and temperature. Predictions from their modified formula agreed well with
their experimental results.
In later work, Z. Li explored the diluting effect of R-134a, a nonflammable
refrigerant using a tube apparatus with a visual criterion [24]. R-134a was
10
added to propane and R-152a. It had little effect on the LFL of these mix-
tures, but significantly decreased the UFL for both mixtures, narrowing the
flammable range. Also, their modified version of Le Chatelier’s formula was
used to predict the flammability limits with dilution. Predictions agreed well
with experimental results.
Z. Yang et al. sought to determine the minimum inerting concentra-
tion (MIC): the concentration of diluent to make a flammable mixture non-
flammable (LFL=UFL) [25]. Experimental work used a vertical tube with
R-32 and air mixtures diluted by varying amounts of R-125, R-1227ea, or R-
1311. R-1311 was found to have the lowest MIC making it the most effective
diluent. The MIC was also calculated theoretically based on a limiting flame
velocity and the inhibition coefficient of the fuel calculated by the group
contribution method. Inhibition values were taken from previous references.
The theoretical predictions however, did not agree well with the experimental
results.
Z. Wu et al. conducted extensive work on the flammability limits of
ethylene including temperature effects above and below ambient as well hu-
midity effects [26]. Temperatures tested ranged from -5 to 5 ◦C . The test
method employed was based on the Chinese GB/T12474 which is a verti-
cal tube test method. Flammability criteria was propagation to the top of
the tube and was recorded by video. Echoing the results of previous work,
increased temperature widened flammability limits and decreased tempera-
ture narrows flammability limits. They also found that the geometric mean
of flammability limits was near constant with temperature. Humidity was
found to have little to no effect on the flammability limits of ethylene. Fi-
11
nally, a new expression to predict LFL based on the number of chemical
bonds and bond-energies was derived. Their new expression agreed very well
with experimental data.
X. Liu et al. investigated the effects of high pressure and temperature on
flammability limits of hydrogen-air mixtures using a 5 L bomb-type cylinder
[27]. The conditions included 20, 40, 60, 75, and 90 ◦C as well as 1, 2, 3,
4 atm pressure. The criterion for flammability was 7% pressure rise. Their
apparatus was first validated with respect to ambient conditions and was
deemed acceptable. It was found that increasing temperature and increasing
pressure widened the flammability limits, but pressure had a much greater
effect than temperature.
K. Zhang et al. explored the effects of different diluent on the flammability
limits of dimethyl ether-air mixtures using an ASHRAE type apparatus [28].
Diluents included N2, CO2, R-22, R-134a, and R-227ea. As expected, in-
creased presence of diluents narrowed the flammable range of dimethyl ether.
Le Chatelier’s formula was modified to include polynomial terms related to
the fuel volume fraction. The predictions of the modified Le Chatelier’s
formula agreed well with experimental results.
R. Tschirschwitz et al. sought to develop flammability criteria for non-
atmospheric conditions (high pressure) [29]. A windowed autoclave − made
of stainless steel 12 L in size − was used to allow visual observation at high
pressure testing. Hydrocarbon-air mixtures were tested and ignited by wire
fusing. Their visual criterion for flammability was based on the failure of
the flame to reach the vessel walls. A pressure rise of 2% and an optional
temperature rise criteria of 100 K was determined to correspond to this visual
12
criteria and became their new flammability criteria. However, it is worth
noting that their criteria is much stricter than in other standards.
1.3.4 Theoretical Work
C.K. Law et al. sought to couple the two existing mechanisms for near-limit
flames: heat loss balance and the branch chain reaction by comparing the
theoretical LFL obtained by each method [30]. For the heat loss analysis, the
mass burning rate was found for different equivalence ratio and the point of
the lowest mass burning rate to be the LFL. For the branch chain analysis, a
“flammability exponent” was defined relating the rates of chain termination
to the rate of chain branching [30]. The LFL was said to occur where the
flammability exponent is equal to one. The resulting LFL for both analysis
was found to be in good agreement with one another.
P. Westmoreland et al. presented early simulations to evaluate the sup-
pressant ability of different fluoromethanes [31]. Westmoreland used the
NIST-modified version of Sandia’s PREMIX code to evaluate flame temper-
ature and flame velocity for methane-air mixtures with varying percentage of
fluoromethanes. [31]. R-32 was found to accelerate the flame (higher burning
velocities) but lower flame temperature due to a smaller heat of combustion
than methane. CHF3 (fluoroform) had opposite behavior, it slightly raised
the flame temperature but significantly reduced the flame speed − believed
to be the result of its chain-terminating reactions. Reaction paths were also
presented in detail.
N. Kim et al. conducted numerical simulations of a 1D-flame in small
tubes [32]. A finite volume method was used. The conditions simulated
13
included adiabatic and isothermal walls and varied velocity profiles. They
were able to obtain the propagation velocity and flame shapes for the different
conditions by modeling the heat transfer, momentum of the flow and flame,
and chemical species [32].
M. Vidal et al. utilized an algebraic method to predict LFL using the
Calculated Adiabatic Flame Temperature (CAFT) [33]. The calculation is
based on the assumption that the flame temperature at the LFL is equal to
the CAFT. An enthalpy balance is then calculated to determine the mols
of oxygen to fuel to then find LFL. Vidal et al. used experimental LFL
values to find different CAFT to then use in their model to predict LFL [33].
The predictions based on CAFT agreed well with the experimental data and
CAFT were determined for different chemical families; however, to be of use,
there needs to be another way to approximate CAFT [33].
C. Chen et al. developed a theoretical model to predict LFL of a diluted
mixture given the LFL of the undiluted mixture, properties of the diluent,
and amount of diluent [34]. They found that the “inerting ability” of a diluent
is thermal in nature and depends most on the mean molar heat capacity (at
adiabatic flame temperature). Diluents evaluated included Carbon Dioxide,
Steam, Nitrogen, and Helium.
J. Rowley et al. worked to improve the predictions of the Burgess-Wheeler
law [35]. They modified the equation to account for heat loss to the products
of combustion and allowed for effects of initial temperature on flame adia-
batic temperature. Their new formula agrees much better with experimental
data than previous models. However, their formula is valid only for basic
hydrocarbons.
14
S. Zhang et al. sought to numerically predict laminar flame speeds of
hydrocarbon-air mixtures using CHEMKIN and CANTERA software [36].
They attempted to validate their model experimentally using a vertical tube
method. Flame speed was found experimentally from analysis of recorded
images. The predictions did not agree well with the experimental data nor
other published data. The authors conclude the need to refine their model.
Zlowchower et al. also conducted theoretical work, using a thermody-
namic program (CEA code developed by NASA Glenn) to predict flamma-
bility limits based on a limiting flame temperature [37]. The calculated
adiabatic flame temperature was chosen as the limiting temperature. The
program would vary the concentration of fuel (and air consequently) until the
limiting temperature is calculated. Results were inconsistent with respect to
published experimental data provided by Kondo. The flammability limits
agreed well for certain fuels but not others without a noticeable pattern.
T. Albahri et al. developed Artificial Neural Networks to predict the
LFL of hydrocarbons based on structural group contributions − 30 struc-
tural groups were employed [38]. The model had excellent agreement with
published experimental data and proved to be more accurate (smaller error)
than previous models.
A. Di Benedetto conducted numerical simulations to calculate LFL based
on an energy balance: The energy required to raise the temperature of the
burned mixture to the ignition temperature equals the heat released from the
reaction minus the heat loss [39]. The heat loss term was a variable parame-
ter and when heat loss goes to zero, the result is the adiabatic flammability
limits, they proposed. They argued that because these limits are indepen-
15
dent of heat loss, they would be independent of the test method and be a
thermodynamic property. The adiabatic flammability limits are much wider
than normal flammability limits. However, their work was limited to hydro-
carbons only and may carry limited physical significance.
H. Okamoto et al. used CFD (STAR-CD program) to simulate leak-
age of refrigerants from air conditioners to determine what scenarios would
lead to the creation of flammable mixtures [40]. Scenario variables included
room size, floor-mounted vs wall-mounted vs ceiling-mounted, different leak-
age rates, and different room air flows [40]. Cumulative volume fraction
of the fuels and flammable volume were tracked over time and used to an-
alyze the hazard of each scenario. It was found that wall-mounted units
accumulate significantly lower flammable volumes than floor-mounted units.
The concentration of refrigerant did not reach the LFL in the cases of wall-
mounted units. Concentrations for floor-mounted units were much higher,
but in many cases above the UFL. Finally, it was found that ventilation rate
has an enormous impact on the accumulation of flammable volume.
1.3.5 Small-Scale Testing
Maruta et al. examined combustion within 2 mm diameter cylindrical ves-
sel [41]. They were able to create stable hydrocarbon flames, overcoming
thermal-quenching by heating the walls with external heaters [41].
N. Kim et al. sought to model stationary flames in small-scale combustors
experimentally by using larger vessels at low pressure [42]. The justification
for the comparison is by the similarity in flame structure, specifically similar
flame sheet thickness [42]. Stationary flames were created by adjusting the
16
flow rate and pressure of pre-mixed propane-air and methane-air mixtures.
Finally, through CCD camera, flame thicknesses and dead space − distance
between the flame and the tube wall − were directly measured. They argue
that most of the heat lost is to the unburned mixture in the dead space
and not the walls directly because varied wall temperature profiles had little
effect on the observed flammability limits.
B. Bai et al. sought to numerically model the quenching of combustion in
small tubes [43]. 1D flame propagation was simulated with species and energy
conservation equations. Both thermal and kinetic quenching was evaluated.
Two parameters were defined: heat loss coefficient and radical quenching
coefficient. Because quenching is inversely related to diameter, quenching ef-
fects on negligible in sufficiently large vessels. Additionally, they found that
the quenching limits for both thermal quenching and radical quenching are
increased with higher wall temperatures. However, changing wall tempera-
ture had a much greater effect on thermal quenching than radical quenching.
Thus elevated wall temperatures can serve as a solution to quenching in small
combustors and allow for smaller diameters.
1.3.6 Work of Shigeo Kondo et al.
Shigeo Kondo et al. have contributed significantly to the study of flamma-
bility limits. Their work comprises much of the flammability limit testing
and modeling done for refrigerants.
Early work of S. Kondo et al. included investigating the effects of spark
duration and spark gap using a tube apparatus and methane-air mixtures
[44]. It was determined the optimal ignition conditions are 0.1-0.3 s duration
17
and spark gap of 6-8 mm [44].
Kondo et al., after taking new measurements of halogenated hydrocar-
bons following the ASHRAE Standard, an algebraic model was developed to
predict LFL and UFL based on molecular structure [45]. In 2006, they fur-
ther developed their model specifically for fluorinated compounds. First, they
plotted F number versus F-substitution rate and found an F number of zero
(meaning LFL=UFL or nonflammable) corresponded to an F-substitution
rate = 0.625. They developed a numerical model to predict the F num-
ber and geometric mean which could then be used to find LFL and UFL.
Both expressions consisted of a series of terms relating to the presence of
structural groups or other parameters such as saturation. There was good
agreement between the predictions of the numerical model and experimental
results, except for compounds with F-substitution rates = 0.625 where the
formula predicted flammability. Their results suggested that F-atoms are
not good chain carriers. Compounds with higher F-substitution rates had a
lower fraction of active radicals and more free F-atoms.
Kondo next investigated Le Chatelier’s formula [46]. Experimental mea-
surements using the ASHRAE method confirmed that Le Chatelier’s formula
predicts the LFL of hydrocarbon-air mixtures well. Using fitting techniques,
they were able to modify Le Chatelier’s formula to predict UFL of mixtures.
The modified formula agreed well with experimental UFL data. Kondo then
investigated the ability of Le Chatelier’s equation to predict mixtures with
diluents [47]. It was found that Le Chatelier’s equation fails for mixtures
with nonflammable compounds. The mixtures used were hydrocarbon-air
mixtures diluted with R-125. Le Chatelier’s equation was modified again,
18
this time to include polynomial terms based on the mol fraction of the dilu-
ent. The predictions of the modified equation accounting for diluents agreed
well with the experimental data.
Kondo also continued to develop the algebraic prediction of F number for
hydrofluorocarbons [48]. Added additional terms to their original expression
to account for not just the number of F-atoms, but their distribution as well
as the presence of specific structures such as −O − CF3 , O − CF = and
−O − CF2− which were found to reduce flammability, have no effect, and
have no effect, respectively.
In 2011, Kondo et al. departed from previous work to conduct tests at
non-atmospheric conditions, specifically high pressure [49]. A 5 L stainless
steel sphere bomb-type was used to test R-1234yf, R-32, and methane-air
mixtures. The flammability criterion was 20% pressure rise. The justification
of this criterion was their theoretical calculation of a 30% pressure rise for a
cone of burned gas at 1500 K. It was found that increasing pressure widens the
flammability limits. Pressure dependence of flammability limits was found
to be logarithmic. However, R-32 had a sudden change in behavior at 1500
kPa, suggesting a change in combustion mechanism.
The following years, 2011-2014, Kondo et al. thoroughly investigated
the effects of temperature and humidity on flammability limits using their
ASHRAE-based apparatus. In 2011, they first tested with simple hydrocar-
bons with dry air [50]. Temperature was controlled via an air bath. Flamma-
bility limits were found to be linearly dependent on temperature. Expressions
based on White’s rule were developed to make predictions of the limits. Their
expression was in excellent agreement with the experimental data. In 2012,
19
Kondo et al., using the same apparatus, tested 2L refrigerants [51]. However,
the air was no longer dry but with varied levels of humidity. Both tempera-
ture and humidity widened the flammability limits. In 2014, they published
their work on high humidity testing [52]. The same test apparatus was used,
but this time the temperature was kept at 60 ◦C and the humidity was var-
ied. Again, increased humidity widened the flammability limits. The fuels
tested − R-410A, R-410B, R-134a, and R-125 − were nonflammable without
sufficiently high humidity. Humidity provides hydrogen atoms to react with
the excess F-atoms.
1.3.7 Literature Review Conclusions
With respect to the goal of improving the accuracy and reproducibility of the
determination of flammability limits, the following are the key ideas gained
from the literature review.
Variables such as temperature, pressure, humidity, ignition source, ves-
sel size, and criteria, and their effects on flammability limits have already
thoroughly been investigated. The 12 L spherical flask has been found to be
of sufficient size to minimize the effects of wall-quenching, although tests in
smaller chambers can be possible if the vessel walls are heated. There exist
many algebraic methods to predict the LFL (and UFL) of mixtures from
data of its individual components and to predict the LFL and UFL of a pure
substance based on its molecular structure. These predictions have aligned
well with experimental data (for most compounds and blends). However,
these has been no work done yet on venting nor any new solutions either.
20
2 Interview Results
Experts on the ASTM E681 standard, research scientists and engineers, were
interviewed. A copy of the interview questions can be found after the below
discussion. The results of the interviews can be summarized as follows.
Experts praise the test for its simplicity and low cost; however, they
have a number of criticisms of the test, especially in the context of mildly
flammable substances such as Class 2L refrigerants. There is concern that
the 90 degree criteria − used to determine if a concentration is flammable −
is not physically justified. Additionally, results can depend on the recording
equipment, settings and positioning, and operator’s judgment. There are no
recommendations or controls for the recording of test results to address these
variables. The standard should include sample images so that recording is
consistent and interpretation of results is reproducible.
Another concern is on the creation of flammable blends. Operators can
use different mixing speeds − only a minimum RPM is specified − which
can result in different levels of turbulence. Additionally, the standard has no
verification of the homogeneity of the blend.
Finally, the majority of the experts expressed concern about the rub-
ber stopper and venting effects. The rubber stopper can have difficulties
in maintaining a seal during testing, especially at pressures close to atmo-
spheric. Venting pressure is not consistent because the weight of the stopper
with the inlet components − electrodes, sensors, gas controls, etc. − is not
standardized. Also, tipping may occur which can also hurt the consistency
of the venting pressure. Early venting interferes with the flame structure,
21
invalidating the test. Late venting poses an explosive hazard to safety and
the delayed evacuation of gaseous products can damage the flask, one of the
more expensive pieces of equipment. From what was seen in the literature,
this issue has not yet been addressed and will be the end goal of this work.
22
Interview Questions
1. What is the nature of your experience with E-681?
2. If you have an E-681 facility, how did you acquire
it?
3. What do you think are the strengths of E-681?
4. What do you think are the weaknesses of E-681?
5. Do you find E-681 to be repeatable and operator in-
dependent?
6. What changes would recommend to E-681?
7. Can you recommend other interviewees?
23
3 Experimental Apparatus
3.1 ASTM E681 Apparatus
The experimental apparatus was developed according to ASHRAE Standard
34, which is a slightly modified version of the ASTM E681 test, using a 12
L spherical flask instead of a 5 L spherical flask. Below is a schematic of
the experimental apparatus. Ignition is accomplished by an electric spark.
Figure 1: Schematic of Experimental Apparatus
It is worth noting that the duration of the spark was reduced from 0.4 s
to 0.2 s because the spark was seen during flame propagation. Tests were
conducted at room temperature and atmospheric pressure. R-32 (CH2F2)
and air mixtures were tested. Gas mixtures were created using the partial
pressure method. Flame propagation was recorded using a high-speed video
24
camera. High speed video (120 fps) was also taken of the rubber stopper
to better understand its venting. Flammability was assessed using the 90 ◦
criteria − to be discussed further in section 4.3. The element of greatest
interest is the rubber stopper which seals the flask and supports all of the
inlet components.
3.2 Burst Disc Development
To address the problems of venting, a burst disc was designed and built as an
alternative to the rubber stopper. The goal of the burst disc is for consistent
venting pressure, even across laboratories. The burst disc design has the
weight of the inlet components rest on the glass and so that they do not
affect venting pressure.
The burst disc consists of the following parts: Attached to the neck of the
flask is an aluminum clamp, lined with gasket to protect the glass. Bolts
attach a nylon block with a hole in the center to the aluminum clamp. There
is also a thin sheet of gasket between the aluminum clamp and the nylon
block to prevent leaks between the two surfaces. All the components − the
electrodes and tubing for the gases and vacuum − are supported by the nylon
block. A small 3.75 x 3” sheet of heavy-duty aluminum foil is placed on top
of the nylon block, covering the hole. Vacuum grease is also used between the
foil and the nylon block. Finally, the cover plate (also made of aluminum) is
placed on top of the foil sheet and secured by bolts into the threaded inserts
of the nylon block.
Another part of the burst disc system is the adjustable knife edge to break
25
Figure 2: Burst Disc Without and With Cover
Figure 3: Burst Disc Foil Sheet and Cover Plate Separated
the foil. The knife edge itself consists of four utility blades that are held by
a 3D-printed “blade holder”. The knife edge is connected is to a metal sheet
that is connected to a micrometer so that the height of the knife edge can be
controlled precisely.
Please see the following page for photos and diagrams of the burst disc.
Design drawings can be found in Appendix A.
26
Figure 4: Knife Edge
Figure 5: Diagram of Knife Edge and Micrometer Station
27
4 Results and Discussion
4.1 Vacuum and Leak Testing
If the burst disc design is to be an alternative to the rubber stopper, it must
meet, if not exceed all the functions of the original rubber stopper. Of critical
importance, is the ability to achieve and maintain a vacuum. Leaks of any
nature can greatly affect the mixture composition being tested and can lead
to wasted tests and invalid results.
ASTM E681 requires a vacuum of -100 kPa and a maximum leak rate
of 0.1 kPa/min or less [3]. Table 1 below compares the performance of the
rubber stopper and the burst disc at vacuum pressure (-100 kPa or less).
Although the burst disc has a slightly higher leak rate than the rubber stop-
per, it still exceeds the requirements of the test standard. The burst disc is
also able to reach a vacuum faster than the rubber stopper. So although the
rubber stopper may create a better seal, it takes longer to do so.
Type Time to Vacuum [min] Leak Rate [kPa/min]
Stopper 3:03 0.0051
Stopper 3:22 0.0049
Stopper 5:01 0.0032
Burst Disc 2:22 0.0062
Table 1: Performance at Vacuum Pressure
The leak rates of Table 1 were obtained from linear fitting of the pressure-
time history. Pressures were recorded every three minutes for a duration of
28
30 minutes.
The pressure-time history at vacuum are shown below in Figure 6. The
solid line of the pressure profile of the burst disc performs satisfactorily with
respect to the rubber stopper.
Figure 6: Pressure Time Histories at Vacuum
Additionally, leak testing was performed at other pressures because it
was determined that the rubber stopper could only be safely filled to -2 kPa
before a leak path was generated evident by a sudden gain in pressure. These
leak rates were also found by fitting tools of pressure-time data. Figure 7
shows the leak rates vs different pressures for the stopper and for the burst
disc.
As shown, the burst disc allows for filling up to and past atmospheric
pressure. Given that the burst disc does not break at vacuum pressure, it is
29
Figure 7: Leak Rates for Different Pressures
likely the burst disc could be filled to 100 kPa above atmospheric. This is
another advantage to the burst disc design.
4.2 Venting Pressure Testing
4.2.1 Burst Disc
It was found that the burst disc has two stages of venting. There is an
initial puncture pressure at which the aluminum foil deflects upwards to the
knife edge, creating a hole. If the pressure is allowed to continue to rise, the
aluminum foil will burst open and its edges unfold. See Figures 8 and 9 for
results of a completed burst test.
Most importantly, burst disc has reproducible venting. It has been found
30
Figure 8: Front View of a Successful Burst
Figure 9: Side View of a Successful Burst
to consistently puncture at 7 kPa with the use of the knife edge. As seen in
Figure 10, multiple tests all punctured near 7 kPa. Changing the position of
the knife edge (relative to the foil sheet) was unable to change this pressure,
although. However, adjusting the position of the knife edge did adjust the
burst pressure and appears to be linearly related. The burst pressure results
do have greater deviations however.
31
Figure 10: Burst Disc Venting Pressures
4.2.2 Rubber Stopper
The rubber stopper does not have a clear venting pressure. All that is known
is that the rubber stopper loses its seal between -2 kPa and 0 kPa. However,
there is venting phenomena still to discuss.
High speed images have revealed that the rubber stopper ejects violently
when it vents at the conclusion of the test. Note that the concentration
(14.8%) used to generate these images is near the LFL and not near stoichio-
metric (17.3%). R-32 is considered a mildly flammable substance and yet
produces violent ejection. Image analysis has also revealed that the stopper
can have the issue of re-sealing. After the initial ejection, the stopper can
potentially re-assume its position. This presents another advantage of the
32
burst disc: prevention of re-sealing.
Figure 11: 14.8% Test Showing Stopper Displacement
33
4.3 Determination of the LFL of R-32
The conditions of the test can be summarized as follows: relative humidity of
the supplied air was 11% − Humidity was not a concern because it has shown
to have no effect on compounds where nh = nf such as R-32. Temperature
was ambient (23 ◦C).
4.3.1 Flammability Criteria
This work used the same criteria as ASHRAE Standard 34 [1]. A mixture is
considered flammable if the flame font forms a continuous, 90 ◦ arc - measured
from the ignition point (electrodes) to the vessel walls - when the flame front
reaches the vessel walls. To aid your understanding, Figure 12 shows the
positioning of the angle measurement.
Figure 12: Diagram showing the proper location for the apex of the angle
measurement [1]
34
4.3.2 Rubber Stopper
Flammability was assessed by image analysis of the flame. The procedure
was as follows:
1. Convert the video file into a series of images
2. Select the images of interest: One frame of the ignition and one frame
of the flame front reaching the vessel walls
3. In any software with image editing (this work used MS powerpoint),
place the images side-by-side.
4. Place a horizontal line such that it passes through the point of ignition
and marks it on the other image of interest
5. Place a 90 ◦ angle with its vertex aligned with the previously placed
horizontal line
6. Examine the image and decide if the flame angle equals or exceeds the
arc.
Figures 13 and 14 show two sample images analyzed from testing using
the rubber stopper.
4.3.3 Burst Disc
The image analysis of the burst disc used the same method. Figures 15 and
16 show two analyzed images
The flames produced with the burst disc seem much more uniform and
nearly perfectly symmetrical. This is valuable for the reproducibility of
35
Figure 13: An example of a “Nonflammable” result with the stopper at 14.7%
Figure 14: An example of a “Flammable” result with the stopper at 14.8%
Figure 15: An example of a “Nonflammable” result with the burst disc at
14.6%
flammability testing using the flame angle criteria, especially because of the
requirement that the flame front be continuous.
36
Figure 16: An example of a “Flammable” result with the burst disc at 14.7%
4.3.4 Comparison With Past Data
Table 2 contains many different LFL values for R-32 as well as the test
method and the references.
LFL (%) Method Reference Notes
13 Reported [53] Airgas MSDS
13.5 ASHRAE [51] -
14 Reported [54] Linde MSDS
14 Reported [55] Hynote Gas MSDS
14.4 ASHRAE [56] -
14.7 ASTM E681-12L [57] Reduced spark power
14.7 ASHRAE, Burst Disc This work Reduced spark duration
14.8 ASHRAE, Stopper This work Reduced spark duration
Table 2: LFL values for R-32
Compared to previously published data for the LFL of R-32, the results
of this work are higher. There are two possibilities for this deviation. One
possibility is that our ignition source is weaker (reduced duration) and so a
more flammable (less lean) mixture is required to compensate. The fact that
Kul’s value of 14.7% agrees with this work also used a weaker ignition (same
duration but lower power) supports this notion. Another possibility is that
37
this work’s criteria is not conservative enough and instead of the initial flame
angle, the maximum flame angle should be used.
4.4 Burst Disc Venting Performance
One of the most important things for the burst disc (and the stopper) is
the time difference between the flame front reaching the vessel walls and the
start of venting.
Through image analysis, the delay between the flame reaching the vessel
walls and venting by displacement of the rubber stopper has been quantified.
The approach was to identify the frame where the flame reached the vessel
walls − which was the frame used to evaluate the flame angle − and to
identify the frame where venting first occurs. Then this difference in frame
number can be divided by the frame rate (120 fps) to obtain seconds, see the
below equation
∆t =
∆frame
120fps
(5)
Venting is identified by the onset of turbulence and destruction of the initial
flame cap. Below is an example of the frame when a flame vented.
In Table 3, are the data used to find the time difference.
Concentration
(%)
Flame Angle
Frame
Venting
Frame
Frame
Difference
Time Difference
(ms)
14.4 654 658 4 33.3
14.5 163 167 4 33.3
14.6 654 659 5 31.7
14.7 367 369 2 16.7
14.8 692 695 3 25
Table 3: Stopper Venting Time
38
Figure 17: Flame during venting (14.8%)
It appears that it takes about 16-34 ms for the stopper to vent, with
shorter times at higher concentrations. This is reasonable because it has
been shown by Takizawa that as the concentration of fuel increases, burning
velocity increases until a maximum value at stoichiometric before decreasing
[5].
Regarding the burst disc, image analysis is not an option to determine
time to venting because the camera does not have a direct view of the foil
disc itself. Installing a camera in the fume hood was not an option because of
the effect of the gaseous HF products (from the combustion of R-32) on glass.
However, audio analysis can be used to identify the onset of venting. The
action of the burst disc was identified by the sudden increase in sound. The
timestamp of venting was recorded and compared with the time at which the
flame struck reached the vessel walls. The smallest unit of the time stamp is
frames, which can be converted into ms as done previously.
39
Figure 18: Audio of 14.5% Test
Figure 19: Flame of 14.5% Test
Figure 20: Audio of 14.8% Test
As seen above, the burst disc starts slow venting (the first peak marked
“venting”) before the flame reaches the vessel walls, but this has been found
40
Figure 21: Flame of 14.8% Test
Concentration (%) Frame Difference Venting Time Difference (ms)
14.5 4 33.3
14.8 2 16.7
Table 4: Burst Disc Venting Time
not to affect flame structure. Immediately after the flame reaches the vessel
walls, the burst disc breaks open and vent completely shown as the largest
peak in the audio file. In Table 4 are the time differences for the two tests
with analyzed audio.
Comparing the venting time difference, the Burst Disc is able to vent as
rapidly as the rubber stopper. The time difference is equal. This presents
further proof that the burst disc is a reliable alternative to the rubber stopper.
41
4.5 Direct Flame Angle Measurement
As an alternative to the simple 90 ◦ assessment with binary “Flammable” or
“Nonflammable” checks, a method to measure the flame angle was developed
and carried out. A summary of the method is given below:
1. Identify the frame when the flame front has reached the vessel walls
2. Arrange the frame of interest next to the ignition frame to determine
ignition point.
3. Using ImageJ software, draw an angle and adjust it until:
(a) Its apex is at the ignition point
(b) The left arm is tangent to the left edge of the frame front
(c) The right arm is tangent to the right edge of the frame front
4. Record the degrees of the angle given by the ImageJ software
5. Repeat 10 times for each test
6. Calculate the average of the 10 measurements
Figures 22 and 23 show examples of the new method.
42
Figure 22: Direct Angle Measurement for a Flame with the Burst Disc
(14.3%)
Figure 23: Direct Angle Measurement for an Asymmetric Flame with the
Stopper (14.9%)
43
Figure 24 shows the results of the direct flame angle measurement. The
markers correspond to the raw data for both the stopper and burst disc.
A dashed line at 90 ◦ shows graphically the criteria for flammability used by
ASHRAE. A linear fit (shown as a solid line) was made. Flame angle appears
to linearly increase with increasing fuel concentration which is in agreement
with test observations.
However, it is important to recall that flame angles should eventually
decrease because of the existence of the UFL, past which the mixture is
no longer flammable. Thus perhaps a quadratic would better fit the data
since the behavior should be close to parabolic when the entire flammable
range is considered. Further data is needed before quadratic fitting can be
accomplished.
Also plotted are the lines that define the confidence intervals. Using the
linear fit and the confidence interval, an LFL of 14.74 ± 0.12% is reported
with 95% confidence using this new method.
44
Figure 24: Results of Measured Flame Angles
4.6 Discussion
From the feedback in the interviews and our testing experience with the
ASTM E681 standard, a number of limitations of the rubber stopper have
been identified.
Use of the rubber stopper can lead to hazardous situations. The rub-
ber stopper − even for mildly flammable substances − can eject violently,
potentially damaging the electrodes. The rubber stopper may also re-seal,
45
trapping the gas which can result in re-ignition. Re-sealing can also mean
serious etching of the glass flask, the most expensive part of the apparatus.
Additionally, the rubber stopper can interfere with the assessment of
flammability. When close to atmospheric pressures, the stopper is known
to leak, which can affect the composition of the mixture, ruining the test.
The rubber stopper also can tilt, leading to asymmetric flames or patches of
flames which interfere with flame angle interpretation.
The burst disc is a viable solution to the problems posed by the rubber
stopper. It can hold a vacuum as well as the rubber stopper and even reach a
vacuum in a shorter period of time. The burst disc does not leak when close
to, and even above atmospheric. Tilting and re-sealing is impossible with the
burst disc. The burst disc vents at a consistent pressure. Additionally, the
burst disc would allow for testing in other orientations. The rubber stopper
requires that the flask is kept upright because it is secured by gravity, but
the burst disc does not need this support. Also, the burst disc allows for the
use of audio analysis because of its audible pop. All in all, the burst disc
makes flammability testing more reproducible.
Finally, direct flame angle measurement is a more objective way of as-
sessing flammability instead of a simple “Flammable” or “Nonflammable”.
This new method of assessing flammability should be included in the ASTM
E681 standard because it will help address the variable of operator judgment
that occurs when using the 90 ◦ criteria.
46
5 Conclusions
Venting effects on flammability limits were identified as a variable not yet
addressed in research. The greatest variable affecting venting is the rubber
stopper of the ASTM E681 apparatus. Limitations of the rubber stopper
have been identified. In response to these limitations, a burst disc was de-
veloped. The burst disc has been shown to meet the leakage and vacuum
requirements, reaching below -100 kPa in less than 3 minutes and having
a leakage rate of only 0.0062 kPa/min. The burst disc also provides great
consistency in venting after being found to puncture at 7 kPa for multiple tri-
als. The burst disc avoids the complications discussed in the interviews such
as re-sealing, tilting, and leaking. The flames obtained during flammability
testing with the burst disc have less instabilities than those obtained from the
rubber stopper, which is critical for flammability assessment using a visual
criteria. Needless to say, the burst disc meets all the current requirements
of the rubber stopper and exceeds the performance of the rubber stopper
in many ways. An LFL of 14.8% for R-32 was obtained using the rubber
stopper and an LFL of 14.7% was obtained using the burst disc. However,
the investigation into the burst disc and venting effects has only just begun.
There remain many variables to couple with the burst disc such as temper-
ature, humidity, testing at UFL instead of LFL and other refrigerants and
refrigerant blends. Additionally, simulations could help better understand
the physics of the flow during venting. The burst disc has great potential in
flammability testing and should become part of the standard test method.
Finally, a new method of direct flame angle measurement for assessing
47
flammability is presented. Direct flame angle measurement will result in more
objective assessment of flammability than the traditional binary decision,
leading to more reproducible results. An LFL of 14.74 ± 0.12% for R-32 is
found when using data fitting with the new method.
48
A Appendix: Design Drawings for Burst Disc
Figure 25: Flask Neck Clamp Drawing
49
Figure 26: Nylon Block Drawing
50
Figure 27: Cover Plate Drawing
51
Figure 28: Assembled Burst Disc
52
B Appendix: Supplementary Flame Images
Figure 29: 14.3% Test with stopper
Figure 30: 14.4% Test with stopper
53
Figure 31: 14.4% Test with stopper
Figure 32: 14.8% Test with stopper
54
Figure 33: 14.9% Test with stopper
Figure 34: 14.3% Test with burst disc
55
Figure 35: 14.5% Test with burst disc
Figure 36: 14.8% Test with burst disc
56
References
[1] ASHRAE. Standard 34-2013, Designation and Safety Classification of
Refrigerants. Technical report, American Society of Heating, Refriger-
ating and Air-conditioning Engineers., Atlanta, 2013.
[2] Peter Lomax. Development of the astm e681 standard. Master’s thesis,
University of Maryland, College Park, 8 2016.
[3] ASTM. Standard Test Method for Concentration Limits of Flamma-
bility of Chemicals, Designation E681-09. Technical report, American
Society of Testing and Materials, West Conshohocken, 2009.
[4] O. Boudard H. Le Chatlier. Limits of flammability of gaseous mixtures.
Bull. Soc. Chim, 1898.
[5] K. Takizawa, A. Takahashi, K. Tokuhashi, S. Kondo, and A. Sekiya.
Burning velocity measurement of fluorinated compounds by the
spherical-vessel method. Combustion and Flame, 141(3):298–307, may
2005.
[6] H.F. Coward and G.W. Jones. Limits of flammability of gases and
vapors. Technical report, US bureau of Mines Bulletin, 1952.
[7] G. De Smedt, F. De Corte, R. Notele, and J. Berghmans. Comparison of
two standard test methods for determining explosion limits of gases at
atmospheric conditions. Journal of hazardous materials, 70(3):105–113,
1999.
[8] Kenji Takizawa, Akifumi Takahashi, Kazuaki Tokuhashi, Shigeo Kondo,
and Akira Sekiya. Burning velocity measurements of fluoropropanes by
the spherical-vessel method. Journal of Fluorine Chemistry, 129(8):713–
719, aug 2008.
[9] Kenji Takizawa, Kazuaki Tokuhashi, and Shigeo Kondo. Flammability
assessment of CH2CFCF3 comparison with fluoroalkenes and fluoroalka-
nes. Journal of hazardous materials, 172(2-3):1329–38, dec 2009.
[10] Volkmar Schroder and Maria Molnarne. Flammability of gas mixtures.
Part 1 fire potential. Journal of hazardous materials, 121(1-3):37–44,
may 2005.
[11] F. Van den Schoor, R. T. E. Hermanns, J. A. van Oijen, F. Verplaet-
sen, and L. P. H. de Goey. Comparison and evaluation of methods for
the determination of flammability limits, applied to methanehydrogenair
mixtures. Journal of hazardous materials, 150(3):573–81, feb 2008.
57
[12] Daniel A. Crowl and Young-do Jo. A method for determining the
flammable limits of gases in a spherical vessel. Process Safety Progress,
28(3):227–236, sep 2009.
[13] Isaac a. Zlochower and Gregory M. Green. The limiting oxygen concen-
tration and flammability limits of gases and gas mixtures. Journal of
Loss Prevention in the Process Industries, 22(4):499–505, jul 2009.
[14] Denis Clodic and Tony Jabbour. Method of test for burning velocity
measurement of flammable gases and results. HVAC and R Research,
17(1):51–75, feb 2011.
[15] M. V. Blanc, P. G. Guest, Guenther von Elbe, and Bernard Lewis. Igni-
tion of Explosive Gas Mixtures by Electric Sparks. I. Minimum Ignition
Energies and Quenching Distances of Mixtures of Methane, Oxygen, and
Inert Gases. The Journal of Chemical Physics, 15(11):798, 1947.
[16] A.L. Furno, E.B. Cook, J.M. Kuchta, and D.S. Burgess. Some observa-
tions on near-limit flames. Symposium (International) on Combustion,
13(1):593–599, jan 1971.
[17] Akifumi Takahashi, Youkichi Urano, Kazuaki Tokuhashi, Masahiro
Kaise, and Shigeo Kondo. Fusing ignition of various metal wires for
explosion limits measurement of methaneair mixture. Journal of Loss
Prevention in the Process Industries, 11(5):353–360, sep 1998.
[18] Akifumi Takahashi, Youkichi Urano, Kazuaki Tokuhashi, and Shigeo
Kondo. Effect of vessel size and shape on experimental flammability
limits of gases. Journal of Hazardous Materials, 105(1-3):27–37, dec
2003.
[19] RG. Richard. Refrigerant flammability testing in large volume vessels.
Technical Report March 1998, Department of Energy DOE, 1998.
[20] Henri Le Chatelier and O. Boudouard. On the flammable limits of gas
mixtures. Process Safety Progress, 24(1):3–5, mar 2005.
[21] G. Ciccarelli, D. Jackson, and J. Verreault. Flammability limits of
NH3–H2–N2–air mixtures at elevated initial temperatures. Combustion
and Flame, 144(1-2):53–63, jan 2006.
[22] Y.L. Shoshin and L.P.H. De Goey. Experimental study of lean flamma-
bility limits of methanehydrogenair mixtures in tubes of different di-
ameters. Experimental Thermal and Fluid Science, 34(3):373–380, apr
2010.
[23] Zhenming Li, Maoqiong Gong, Eryan Sun, Jianfeng Wu, and Yuan Zhou.
Effect of low temperature on the flammability limits of methanenitrogen
mixtures. Energy, 36(9):5521–5524, sep 2011.
58
[24] Zhenming Li, Maoqiong Gong, Jianfeng Wu, and Yuan Zhou. Flamma-
bility limits of refrigerant mixtures with 1,1,2,2-tetrafluoroethane. Ex-
perimental Thermal and Fluid Science, 35(6):1209–1213, sep 2011.
[25] Zhao Yang, Huanwei Liu, and Xi Wu. Theoretical and experimen-
tal study of the inhibition and inert effect of HFC125, HFC227ea and
HFC13I1 on the flammability of HFC32. Process Safety and Environ-
mental Protection, 90(4):311–316, jul 2012.
[26] Xi Wu, Zhao Yang, Xiaoming Wang, and Yulong Lin. Experimental and
theoretical study on the influence of temperature and humidity on the
flammability limits of ethylene (R1150). Energy, 52:185–191, apr 2013.
[27] Xueling Liu and Qi Zhang. Influence of initial pressure and tempera-
ture on flammability limits of hydrogen–air. International Journal of
Hydrogen Energy, 39(12):6774–6782, apr 2014.
[28] Ke Zhang, Xianyang Meng, and Jiangtao Wu. Flammability limits of
binary mixtures of dimethyl ether with five diluent gases. Journal of
Loss Prevention in the Process Industries, 29:138–143, may 2014.
[29] Rico Tschirschwitz, Volkmar Schroder, Elisabeth Brandes, and Ulrich
Krause. Determination of explosion limits – Criterion for ignition under
non-atmospheric conditions. Journal of Loss Prevention in the Process
Industries, jan 2015.
[30] CK Law and FN Egolfopoulos. A unified chain-thermal theory of funda-
mental flammability limits. Symposium (international) on combustion,
24(1):137–144, 1992.
[31] PR. Westmoreland. Fluoromethane chemistry and its role in flame sup-
pression. Symposium (International) on Combustion, pages 1505–1511,
1994.
[32] N. Kim and K. Maruta. A numerical study on propagation of premixed
flames in small tubes. Combustion and Flame, 146(1-2):283–301, jul
2006.
[33] M. Vidal, W. Wong, W. J. Rogers, and M. S. Mannan. Evaluation of
lower flammability limits of fuel-air-diluent mixtures using calculated
adiabatic flame temperatures. Journal of hazardous materials, 130(1-
2):21–7, mar 2006.
[34] Chan-Cheng Chen, Tzu-Chi Wang, Horng-Jang Liaw, and Hui-Chu
Chen. Nitrogen dilution effect on the flammability limits for hydro-
carbons. Journal of hazardous materials, 166(2-3):880–90, jul 2009.
59
[35] J. R. Rowley, R. L. Rowley, and W. V. Wilding. Estimation of the lower
flammability limit of organic compounds as a function of temperature.
Journal of hazardous materials, 186(1):551–7, feb 2011.
[36] S. Zhang, S. P. Zeppieri, M. B. Colket, K. Smith, S. M. Opalka, and
P. Verma. Laminar Flame Speeds of Hydrofluorocarbon-Air Mixtures. In
Fall Technical Meeting of the Eastern States Section of the Combustion
Institute, pages 1–7, Storrs, CT, 2011.
[37] Isaac A. Zlochower. Experimental flammability limits and associated
theoretical flame temperatures as a tool for predicting the temperature
dependence of these limits. Journal of Loss Prevention in the Process
Industries, 25(3):555–560, may 2012.
[38] Tareq A. Albahri. Prediction of the lower flammability limit percent
in air of pure compounds from their molecular structures. Fire Safety
Journal, 59:188–201, jul 2013.
[39] Almerinda Di Benedetto. The thermalthermodynamic theory of flamma-
bility The adiabatic flammability limits. Chemical Engineering Science,
99:265–273, aug 2013.
[40] Hiroaki Okamoto. Leakage of Mildly Flammable Refrigerants into a
Room. 2014.
[41] Kaoru Maruta, Takuya Kataoka, Nam Il Kim, Sergey Minaev, and
Roman Fursenko. Characteristics of combustion in a narrow channel
with a temperature gradient. Proceedings of the Combustion Institute,
30(2):2429–2436, jan 2005.
[42] Nam Il Kim, Takuya Kataoka, Shigenao Maruyama, and Kaoru Maruta.
Flammability limits of stationary flames in tubes at low pressure. Com-
bustion and Flame, 141(1-2):78–88, apr 2005.
[43] Bin Bai, Zheng Chen, Huangwei Zhang, and Shiyi Chen. Flame propa-
gation in a tube with wall quenching of radicals. Combustion and Flame,
160(12):2810–2819, dec 2013.
[44] Shigeo Kondo, Youkichi Urano, Akifumi Takahashi, and Kazuaki
Tokuhashi. Reinvestigation of Flammability Limits Measurement of
Methane by the Conventional Vessel Method with AC Discharge Ig-
nition. Combustion Science and Technology, 145(1-6):1–15, sep 1999.
[45] Shigeo Kondo, Kenji Takizawa, Akifumi Takahashi, and Kazuaki
Tokuhashi. Measurement and numerical analysis of flammability limits
of halogenated hydrocarbons. Journal of hazardous materials, 109(1-
3):13–21, jun 2004.
60
[46] Shigeo Kondo, Kenji Takizawa, Akifumi Takahashi, Kazuaki Tokuhashi,
and Akira Sekiya. A study on flammability limits of fuel mixtures.
Journal of hazardous materials, 155(3):440–8, jul 2008.
[47] Shigeo Kondo, Kenji Takizawa, Akifumi Takahashi, Kazuaki Tokuhashi,
Junji Mizukado, and Akira Sekiya. Flammability limits of olefinic and
saturated fluoro-compounds. Journal of hazardous materials, 171(1-
3):613–8, nov 2009.
[48] Shigeo Kondo, Kenji Takizawa, Akifumi Takahashi, Kazuaki Tokuhashi,
and Akira Sekiya. Flammability limits of five selected compounds each
mixed with HFC-125. Fire Safety Journal, 44(2):192–197, feb 2009.
[49] Shigeo Kondo, Akifumi Takahashi, Kenji Takizawa, and Kazuaki
Tokuhashi. On the pressure dependence of flammability limits of
CH2=CFCF3, CH2F2 and methane. Fire Safety Journal, 46(5):289–
293, jul 2011.
[50] Shigeo Kondo, Kenji Takizawa, Akifumi Takahashi, and Kazuaki
Tokuhashi. On the temperature dependence of flammability limits of
gases. Journal of hazardous materials, 187(1-3):585–90, mar 2011.
[51] Shigeo Kondo, Kenji Takizawa, and Kazuaki Tokuhashi. Effects of tem-
perature and humidity on the flammability limits of several 2L refriger-
ants. Journal of Fluorine Chemistry, 144:130–136, dec 2012.
[52] Shigeo Kondo, Kenji Takizawa, and Kazuaki Tokuhashi. Effect of high
humidity on flammability property of a few non-flammable refrigerants.
Journal of Fluorine Chemistry, 161:29–33, 2014.
[53] AirGas. Safety Data Sheet Difluoromethane (R32), 5 2015. Ver 0.04.
[54] Linde Gas. Difluoromethane (R32) Safety Data Sheet, 4 2015. Initial
Release.
[55] Hynote Gas. Material Safety Data Sheet: Difluoromethane (R-32), 2015.
[56] David P. Wilson and Robert G. Richard. Determination of refrigerant
lower flammability limits in compliance with proposed addendum p to
standard 34. ASHRAE Transactions, 108 PART 2:739–755, 2002.
[57] I. Kul, D. L. Gnann, A. L. Beyerlein, and D. D. DesMarteau. Lower
flammability limit of difluoromethane and percolation theory. Interna-
tional Journal of Thermophysics, 25(4):1085–1095, 2004.
61